Jurnal Teknik Informatika (Jutif)

Performance Comparison of Child Stunting Prediction Support Vector Machine vs Random Forest with Grid Search Optimization

2025-09-03T02:17:14+00:00

Stunting is a serious global health problem, particularly in developing countries. Its prevalence is high in Indonesia, reaching approximately 24.4% among children under five in 2021. This condition, defined as failure to thrive due to chronic malnutrition, repeated infections, and a lack of psychosocial stimulation, has long-term impacts on an individual's cognitive development and productive capacity. This study aims to conduct a comparative analysis of the Support Vector Machine and Random Forest algorithms in predicting stunting in children, with a focus on evaluating the impact of hyperparameter optimization using Grid Search on model performance. This study used the public stunting dataset from Kaggle and included data preprocessing steps such as handling missing values, duplication, encoding, and scaling. The data was then divided into 80% for training, 10% for testing, and 10% for validation. Comprehensive evaluation metrics such as precision, recall, F1-score, and ROC-AUC were also used to assess model performance. Grid Search optimization was applied to both models to find the best hyperparameter combination. Experimental results showed that Grid Search optimization significantly improved the accuracy of the SVM model from 94.29% to 98.37%. Meanwhile, the Random Forest model demonstrated very high performance, achieving 99.59% accuracy both before and after Grid Search optimization. These findings underscore the significant potential of machine learning models in supporting stunting prevention efforts for public health intervention policies. This research contributes to the development of machine learning-based decision support systems for public health, particularly in early detection and intervention strategies for stunting.

Fine-Tuned Transfer Learning with InceptionV3 for Automated Detection of Grapevine Leaf Diseases

2025-06-18T12:43:07+00:00

Grape leaf diseases pose a major threat to vineyard productivity, making early and accurate detection essential for modern grape plantation management. Despite advancements in computer vision, challenges remain in differentiating diseases with visually similar symptoms. This study addresses that gap by developing a grape leaf disease classification system using a fine-tuned deep learning model based on the InceptionV3 architecture. Three training scenarios were conducted with fixed parameters batch size of 32 and learning rate of 0.001while varying the number of epochs (25, 50, and 75). Results showed a consistent improvement in classification accuracy with increased training epochs, reaching 98.64%, 98.78%, and 99.09% respectively. Confusion matrix analysis revealed that most misclassifications occurred between visually similar diseases such as Black Rot and ESCA, but error rates declined as the number of epochs increased. Rather than merely applying transfer learning, this research highlights the impact of systematic tuning specifically epoch count optimization in enhancing model accuracy for difficult to distinguish disease classes. These findings underscore the urgency of developing high performance, automated disease detection tools to support precision agriculture and sustainable crop health monitoring.

Design and Implementation of Kernel-Based Quantum Classification Algorithms for Data Analysis in Software Engineering using Quantum Support Vector Machine (QSVM)

2025-07-03T02:20:43+00:00

With the increasing complexity of projects and the volume of data in Software Engineering (SE), the need for efficient and accurate data analysis techniques has become crucial. Classification algorithms play a vital role in various SE tasks, such as bug detection, software quality prediction, and requirements classification. Quantum computing offers a new paradigm with the potential to overcome classical computational limitations for certain types of problems. This research proposes the design and implementation of a kernel-based quantum classification algorithm (also known as Quantum Support Vector Machine - QSVM) tailored for data analysis in the SE domain. We will discuss the fundamental principles behind quantum feature mapping and quantum kernel matrices, and demonstrate its implementation using quantum computing libraries. As a case study, the designed algorithm will be tested on a software bug detection dataset, comparing its performance with classical kernel-based classification algorithms like Support Vector Machine (SVM). The result of the comparison show that QSVM is superior in terms of accuracy, precision, recall, and F1-score compared to SVM.

Accelerating Classification For Iot Attack Detection Using Decision Tree Model With Gini Impurity Tree-Based Feature Selection Technique

2025-01-23T04:56:07+00:00

The Internet of Things (IoT) continues to expand rapidly, with the number of connected devices expected to reach billions in the near future. However, it makes IoT devices prime target for cyber-attack. Therefore, an effective Intrusion Detection System (IDS) is required to detect these attacks swiftly and accurately. This study aims to build a machine learning-based IDS to effectively detect attack on IoT network using the CIC IoT 2023 dataset. The dataset contains over 46 million data rows with 48 features, covering 33 attack types and 1 benign class. To address the dataset's complexity and enhance processing efficiency, feature selection technique was applied. Six feature selection techniques from three categories – Filter-based, Wrapper-based, and Hybrid methods – were evaluated to produce the best feature subset. Each subset was tested using a Decision Tree algorithm. Then, the model performance calculated based on accuracy, computational time, as well as macro-precision, -recall, and -F1-score. The results demonstrate that the three best feature selection from each category – Mutual Information, Genetic Algorithm, and Gini Impurity Tree-based – improved training time by average different 55 seconds from 148 seconds, which speed up by 63.06% without sacrificing accuracy. The Gini Impurity Tree-based algorithm proved to be the most efficient, producing the smallest feature subset, which is 10 features, faster processing times, which is 40 seconds, and shallower tree’s depth, which is 64 level from 73 level. In conclusion, feature selection not only enhances computational efficiency but also simplifies tree’s shape without sacrificing the accuracy of detection.

A Bluetooth-Based Attendance System for Educational Administration at SMA Muhammadiyah: Cross-Platform Development and Usability Validation

2025-06-16T02:10:48+00:00

The transformation of educational administration through technology has accelerated significantly, particularly in attendance systems, which have traditionally relied on manual roll calls. These conventional methods are time-consuming, error-prone, and susceptible to manipulation. This study presents a novel Bluetooth-based attendance system that contributes to the field by demonstrating passive MAC address detection for automated attendance recording, eliminating the need for additional software installations on student devices. The system was developed using React Native for cross-platform compatibility, with PostgreSQL for data management and NestJS for backend processing. The software engineering process followed Rapid Application Development (RAD) methodology, combined with comprehensive system validation through experimental testing. Usability evaluation with 133 participants using the System Usability Scale (SUS) yielded a score of 79.85, categorizing the system within the "Good to Excellent" usability range. The findings demonstrate significant improvements in efficiency and a reduction in attendance fraud compared to conventional methods. However, hardware quality and device proximity remain key limitations. Future research should explore the integration of Bluetooth Low Energy (BLE) technology, the implementation of machine learning algorithms for anomaly detection, or the development of hybrid validation models that combine multiple authentication factors. This system demonstrates the potential to modernize educational administration through seamless, device-level integration while maintaining high user acceptance.

Integrated Fuzzy Logic Model for Smart Water Quality Monitoring and Floating Net Cage Optimization in Barramundi Aquaculture

2025-09-29T04:01:40+00:00

Water quality and aquatic conditions are critical factors in the success of fish farming with Floating Net Cages (FNCs). However, manual monitoring is often delayed due to limited human resources, irregular measurement schedules, and dependence on manual sampling, which can result in late detection of deteriorating water quality and ultimately increase the risk of fish stress, disease outbreaks, and mortality. This study aims to develop an Internet-based water quality monitoring system, integrated with smartphones and PCs, to support rapid decision-making for FNC relocation when water conditions deteriorate. The system is equipped with sensors for temperature, dissolved oxygen (DO), pH, electrical conductivity (EC), total dissolved solids (TDS), turbidity, anemometer, and wind direction, and was field-tested for 36 days in sea-based Barramundi aquaculture. Decision-making was implemented using a Fuzzy Inference System (FIS) with input variables: temperature, DO, pH, and anemometer data, while the output variable was the FNC status: “Relocate” or “Remain.” Results indicated that water quality changes occurred across both short-term and long-term intervals, and during a 56-hour fuzzy simulation, 10 data points suggested “Relocate” while 46 data points indicated “Remain.” The novelty of this research lies in the integration of real-time IoT monitoring with fuzzy logic specifically for FNC relocation decision-making, bridging environmental sensing and intelligent decision support. These findings demonstrate that the proposed system is more effective and efficient than conventional methods, contributing to the advancement of intelligent aquaculture technologies.

Enhancing Fake News Detection on Imbalanced Data Using Resampling Techniques and Classical Machine Learning Models

2025-07-26T11:49:48+00:00

Class imbalance remains a critical challenge in fake news detection, particularly in domains such as entertainment media where class distributions are highly skewed. This study evaluates seven resampling techniques—Random Oversampling, SMOTE, ADASYN, Random Undersampling, Tomek Links, NearMiss, and No Resampling—applied to three classical machine learning models: Logistic Regression, Support Vector Machine (SVM), and Random Forest. Using the imbalanced GossipCop dataset comprising 24,102 news headlines, the proposed pipeline integrates TF-IDF vectorization, stratified 3-fold cross-validation, and five evaluation metrics: F1-score, precision, recall, ROC AUC, and PR AUC. Experimental results show that oversampling methods, particularly SMOTE and Random Oversampling, substantially improve minority class (fake news) detection. Among all model–resampling combinations, SVM with SMOTE achieved the highest performance (F1-score = 0.67, PR AUC = 0.74), demonstrating its robustness in handling imbalanced short-text classification. Conversely, undersampling methods frequently reduced recall, especially with ensemble models like Random Forest. This approach enhances model robustness in fake news detection on skewed datasets and contributes a reproducible, domain-specific framework for developing more reliable misinformation classifiers.

Cross-Temporal Generalization of IndoBERT for Indonesian Hoax News Classification

2025-08-19T03:34:16+00:00

The spread of hoaxes in digital media poses a major challenge for automated detection systems as language and topics evolve over time. Although Transformer-based models such as IndoBERT have demonstrated high accuracy in previous studies, their performance across different time periods remains underexplored. This study examines the cross-temporal generalization ability of IndoBERT for hoax news classification. The model was trained on labeled articles from 2018–2023 and tested on data from 2025 to evaluate its robustness against temporal distribution shifts. The results indicate high accuracy on similar-period data (99.67–99.89%) but a decrease on 2025 data (95.45–95.87%), with most errors occurring as false negatives in the hoax class. These findings highlight the impact of temporal distribution shifts on model reliability and underscore the importance of adaptive strategies such as periodic retraining and domain-based data augmentation. Practically, this model has the potential to assist social media platforms and government institutions in developing dynamic and time-adaptive hoax detection systems. The cross-temporal approach employed in this study also offers methodological innovation compared to conventional random validation, as it better reflects real-world conditions where misinformation patterns continually evolve.

Segmentasi Pelanggan Menggunakan K-Means Clustering Berdasarkan Data Kepribadian dan Pola Konsumsi

2025-07-19T03:47:01+00:00

In today's competitive business landscape, a deep understanding of customer behavior and preferences is crucial for strategic success. Customer segmentation emerges as a vital approach to identify distinct customer subgroups, enabling personalized and efficient marketing strategies. However, many companies still struggle to achieve this understanding due to suboptimal data utilization and inaccurate manual grouping methods. To address these challenges, this research proposes and implements a data mining approach using the K-Means Clustering algorithm for automated and measurable customer segmentation. Leveraging the "Customer Personality Analysis" dataset from Kaggle, this study aims to uncover hidden patterns in customer demographics (age, income, marital status, number of children) and purchasing behavior (number and frequency of transactions). A comprehensive data pre-processing pipeline, including handling missing values, feature engineering, irrelevant column removal, categorical transformation, and numerical scaling, ensures data quality and readiness. Using the Elbow Method, four optimal clusters were identified: "Balanced Spenders with Teenagers" (Cluster 0), "Budget-Conscious Families" (Cluster 1), "High-Value Engaged Buyers" (Cluster 2), and "Active Mature Buyers" (Cluster 3). Visualization using Principal Component Analysis (PCA) further confirms significant characteristic differences between these segments. Cluster 2, being the most valuable and responsive segment, requires premium marketing strategies, while Cluster 1, the largest segment, demands a value-oriented approach. The results of this segmentation provide deep strategic insights, enabling companies to allocate marketing resources more efficiently, craft more relevant messages, and ultimately enhance customer satisfaction and business profitability. These findings demonstrate the potential of unsupervised learning in enhancing data-driven customer profiling systems in marketing and business informatics.

Mapping Facial Expressions Based on Text for Virtual Counseling Chatbot Using IndoBERT Model

2025-07-16T00:15:04+00:00

Early marriage in Lombok remains a serious issue, with a prevalence rate of 16.59% in 2021, the second highest in Indonesia. Limited access to counseling services, especially in rural areas, poses a significant prevention challenge. This study developed a virtual counseling chatbot system capable of mapping text-based emotions to facial expressions to improve the effectiveness of counseling for early marriage prevention. The methodology involved training an IndoBERT model on a synthetic dataset to analyze conversation texts. The model was designed to classify user input into five functional emotion categories: Enthusiasm, Gentleness, Analytical, Inspirational, and Cautionary. Performance evaluation revealed that the IndoBERT model achieved an outstanding accuracy of 94% in its final phase. This result significantly surpassed other models evaluated, such as CNN (71.6%) and KNN (79%), confirming the superiority of the chosen approach The study concludes that the high-accuracy IndoBERT model is a robust foundation for empathetic virtual agents. This research provides a significant contribution to the fields of Affective Computing and Human-Computer Interaction by demonstrating an effective framework for mapping nuanced, functional emotions from Indonesian text to facial expressions. The proposed system not only offers a scalable technological solution for mental health challenges like early marriage prevention but also highlights the impact of advanced, context-aware NLP models in creating more human-like and empathetic user interactions.

From Logs to Insights in the Pulp & Paper Industry: Generating Structured Alarm Reports Using LLMs and RAG

2025-08-06T01:54:06+00:00

Effective alarm management is essential in industrial environments to ensure operational safety and minimize costly downtime. Traditional rule-based reporting systems often struggle to handle heterogeneous alarm log formats and the complexity of natural language queries, limiting their adaptability in real-world applications. To address these limitations, this study proposes a generative alarm reporting system that integrates Large Language Models (LLMs) with a Retrieval-Augmented Generation (RAG) framework. The system converts natural language queries into structured JSON filters, enabling efficient retrieval of contextual information from historical alarm logs. Three open-source LLMs—CodeLlama-7B, LLaMA 3.1-8B, and Mistral-7B—were locally deployed and evaluated using both quantitative and qualitative methods. Experimental results show that CodeLlama-7B achieved the best overall performance, with an Exact Match Accuracy of 0.80, a Field Match score of 93.8%, and a 0% Parse Failure Rate, outperforming the other models in reliability and structural consistency. Compared to conventional rule-based approaches, the proposed LLM-RAG integration demonstrates improved relevance, interpretability, and responsiveness in alarm reporting. This work represents the first systematic benchmarking of locally deployed open-source LLMs for industrial alarm management, providing a replicable framework and highlighting their potential to advance intelligent, real-time, and domain-specific reporting in the pulp and paper industry and beyond.

Random Forest Machine Learning Analysis of Generative AI’s Impact on Learning Effectiveness in Indonesian Higher Education

2025-10-02T01:07:56+00:00

Generative Artificial Intelligence (GenAI) has rapidly penetrated Indonesian higher education, creating opportunities for learning innovation while raising concerns about effectiveness and academic integrity. This study develops a machine learning–based quantitative model to analyze the impact of GenAI usage on learning effectiveness, with a particular focus on Informatics students as key digital literacy stakeholders. Data were collected from a simulated survey of 300 students, covering demographics, GPA, exam scores, GenAI usage patterns, digital literacy, motivation, self-efficacy, academic integrity, and institutional support. Preprocessing steps included normalization of continuous variables, one-hot encoding of categorical variables, and feature selection using Recursive Feature Elimination (RFE). Six machine learning algorithms—Logistic Regression, Decision Tree, Random Forest, Support Vector Machine (SVM), XGBoost, and Artificial Neural Network—were compared to identify the best predictive model. Results show that Random Forest achieved the highest performance, with 87% accuracy and an AUC greater than 0.90, significantly outperforming other algorithms. The most influential predictors were digital literacy, institutional policies, and frequency of GenAI usage, while demographic variables contributed minimally. These findings suggest that GenAI can enhance learning effectiveness in Informatics education when supported by critical digital literacy and ethical awareness. The novelty of this study lies in integrating survey-based educational data with Random Forest machine learning to empirically model GenAI’s role in Indonesian higher education. The results provide practical implications for policymakers, educators, and institutions to design AI-integrated learning strategies that maximize innovation while safeguarding academic integrity.

Data Augmentation-Driven Predictive Performance Refinement in Multi-Model Convolutional Neural Network for Cocoa Ripeness Prediction

2025-09-08T22:11:06+00:00

Timely and accurate prediction of cocoa fruit ripeness is critical for optimizing harvest schedules, improving yield quality, and supporting post-harvest processing. Conventional visual inspection methods are prone to subjectivity and inconsistencies, especially when distinguishing among multiple ripeness levels based on fruit age. This study proposes a deep learning approach that leverages multi-model convolutional neural network transfer learning combined with image data augmentation to classify cocoa fruit into four maturity stages derived from fruit age. An augmented dataset of cocoa fruit images was used to fine-tune five well-established pre-trained models: MobileNetV2, Xception, ResNet50, DenseNet121, and DenseNet169. Data augmentation techniques were employed to increase variability and improve model generalization. Model evaluation was conducted using a standard 80:20 training-to-testing split to ensure sufficient data for learning while preserving a representative test set across all ripeness classes. The results demonstrate that DenseNet169 consistently outperformed other models, achieving the highest average accuracy of 85,05%, followed by DenseNet121 84,06%. Across all models, the use of data augmentation led to notable performance gains, highlighting its importance in enhancing predictive capability and reducing overfitting. The proposed framework shows promising potential for automating ripeness classification in agricultural contexts, offering a robust, scalable, and accurate solution for intelligent cocoa harvest management. This work contributes to the growing application of deep learning in precision agriculture, particularly in addressing fine-grained classification problems using limited but enriched visual data.

Implementation of Ant Colony Optimization in Obesity Level Classification Using Random Forest

2025-05-26T03:25:30+00:00

Obesity is a pressing global health issue characterized by excessive body fat accumulation and associated risks of chronic diseases. This study investigates the integration of Ant Colony Optimization (ACO) for feature selection in obesity-level classification using Random Forests. Results demonstrate that feature selection significantly improves classification accuracy, rising from 94.49% to 96.17% when using ten features selected by ACO. Despite limitations, such as challenges in tuning parameters like alpha (α), beta (β), and evaporation rate in ACO techniques, the study provides valuable insights into developing a more efficient obesity classification system. The proposed approach outperforms other algorithms, including KNN (78.98%), CNN (82.00%), Decision Tree (94.00%), and MLP (95.06%), emphasizing the importance of feature selection methods like ACO in enhancing model performance. This research addresses a critical gap in intelligent healthcare systems by providing the first comprehensive study of ACO-based feature selection specifically for obesity classification, contributing significantly to medical informatics and computer science. The findings have immediate practical implications for developing automated diagnostic tools that can assist healthcare professionals in early obesity detection and intervention, potentially reducing healthcare costs through improved diagnostic efficiency and supporting digital health transformation in clinical settings. Furthermore, the study highlights the broader applicability of ACO in various classification tasks, suggesting that similar techniques could be used to address other complex health issues, ultimately improving diagnostic accuracy and patient outcomes.

Designing an AI-Based Village Information System Using Research and Development Approach for Public Governance Modernization in Popalia Village

2025-08-13T22:52:24+00:00

The COVID-19 pandemic has highlighted the challenges faced by village administrations in managing public services that remain highly manual, inefficient, and prone to errors. This study aims to design an Artificial Intelligence (AI)-based Village Information System to improve administrative efficiency, accuracy, and accessibility. The research employs a Research and Development (R&D) approach through requirement analysis, system design, prototype development, integration of an AI Generative Model and a Natural Language Processing (NLP) chatbot, followed by functional testing using the Black-box method and usability evaluation with the System Usability Scale (SUS). The results show that functional testing achieved a 95% pass rate and the SUS evaluation scored 87.0, placing the system in the “Excellent” category. These findings indicate that the system effectively automates document creation, validates citizen data, and supports interactive services through an NLP-based chatbot. The study contributes to the modernization of digital village governance in Indonesia by demonstrating how AI integration can reduce administrative workload, minimize errors, and enhance service quality.

Design of a Digital Platform for PAUD Child Development Monitoring Using the Dynamic Systems Development Method and Machine Learning

2025-07-10T07:44:12+00:00

This study aims to design a digital platform for monitoring early childhood development in PAUD (Pendidikan Anak Usia Dini) institutions by integrating Machine Learning (ML) into the Dynamic Systems Development Method (DSDM) framework. The research addresses persistent challenges in traditional monitoring systems, which are typically manual, fragmented, and lack real-time responsiveness. Utilizing a Research and Development (R&D) approach, the platform was developed iteratively with active involvement from teachers, parents, and administrators of PAUD institutions. System modeling employed Unified Modeling Language (UML), while ML techniques such as Decision Trees were trained on datasets sourced from PAUD Flamboyan in Tangerang. Key platform features include child data input, growth visualization, predictive analytics, and interactive dashboards. The system underwent black-box testing and usability assessments, achieving an average usability score of 4.5 out of 5. The ML model demonstrated statistically valid and reliable performance with 89% accuracy, 85% precision, and 87% recall in predicting developmental delays. The findings highlight the effectiveness of the DSDM approach in facilitating adaptive system development, and underscore the value added by ML integration in enhancing decision-making within early childhood education. The platform not only streamlines developmental monitoring but also supports early interventions. Future work is recommended to broaden data sources, enrich personalization, and scale deployment across varied PAUD contexts. This study contributes to the advancement of intelligent decision support systems in early childhood education, enabling more accurate developmental monitoring and timely interventions.

A Decision Tree Model with Grid Search Optimization for Scholarship Recipient Classification

2025-08-05T22:07:42+00:00

This study aims to classify scholarship recipients using the Decision Tree algorithm implemented in RapidMiner. The dataset consists of 1.404 records with socioeconomic and academic attributes. Preprocessing was conducted using two Replace Missing Value operators, where categorical attributes such as No. BANTUAN, No. KKS, and Prestasi were filled with "Tidak Punya," while Kepemilikan Rumah was imputed using the average value. The model was built using a Decision Tree algorithm, optimized with the Optimize Parameters (Grid) operator to determine the best values for maximal depth and confidence. Evaluation was performed using 10-fold Cross Validation to ensure reliability. The results show that the optimized Decision Tree model achieved a high accuracy of 97.72%, with strong precision, recall, and F1-score values in both the "Eligible" and "Not Eligible" classes. These findings demonstrate that the Decision Tree algorithm, when properly optimized and validated, can effectively support decision-making processes in scholarship eligibility classification. The model provides an interpretable and robust tool for educational institutions to evaluate student applications based on critical socioeconomic features, This research contributes to educational data mining by offering a validated and interpretable model that enhances fairness, transparency, and efficiency in the scholarship selection process.

Comparison of Port Scanning, Vulnerability Scanning, and Penetration Testing Combinations for Network Vulnerability Detection in GNS3 Testbed

2025-07-03T12:03:05+00:00

Network security faces significant challenges due to the increasing number and complexity of system vulnerabilities. This study aims to develop and evaluate a full combination method (ABC) integrating port scanning (Nmap), vulnerability scanning (OpenVAS), and penetration testing (Metasploit), and compare it with partial combinations (AB, BC, AC) for more effective vulnerability detection. Using a quantitative experimental approach within a controlled GNS3 TestBed, three key indicators were analyzed: number of vulnerabilities detected, detection time, and exploit validity. Experimental results show that the ABC method detected 62 potential vulnerabilities, including 11 high and medium severity CVEs, matching the AB method but significantly outperforming AC, which detected none. In terms of detection time, the ABC method achieved a balanced performance at 91 minutes, which is 31.5% faster than AB (133 minutes), while maintaining full exploit validation. Notably, the ABC method successfully validated 100% of critical vulnerabilities using Metasploit, confirming the practical applicability and reliability of the integrated approach compared to dual combinations. Overall, the findings demonstrate that the full combination method (ABC) offers superior accuracy and comprehensiveness in detecting and validating network vulnerabilities. This research contributes to cybersecurity practices by proposing an integrated detection workflow that effectively balances speed and depth of analysis, setting a practical benchmark for vulnerability detection systems applicable to both simulated and real-world network environments.

Decision Support System for Selecting Outstanding Religious Counselors in Jambi Province Using Analytical Hierarchy Process and Technique for Order Preference by Similarity to Ideal Solution

2025-10-22T11:02:47+00:00

Religious counselors play an essential role in fostering religious moderation, strengthening community cohesion, and promoting social harmony. However, the evaluation of their performance remains largely manual, leading to subjectivity, inconsistency, and limited accountability. This study develops a web-based Decision Support System that integrates the Analytical Hierarchy Process and the Technique for Order Preference by Similarity to Ideal Solution to enhance objectivity, transparency, and data-driven evaluation. The Analytical Hierarchy Process was applied to determine the importance of five criteria—portfolio, scientific paper, program video, presentation or interview, and absenteeism—through expert pairwise comparisons. The Technique for Order Preference by Similarity to Ideal Solution was then used to rank twenty-four religious counselors from the Regional Office of the Ministry of Religious Affairs in Jambi Province. The results show that portfolio (47.4%) and presentation or interview (24.4%) were the most influential criteria, while the others served as complementary factors. Counselors with comprehensive documentation and strong communication skills consistently ranked higher, validating the system’s analytical reliability. This study’s novelty lies in applying a multi-criteria decision-making framework within the religious sector, directly aligned with the 2024 Technical Guidelines for the Islamic Religious Counselor Award (Keputusan Dirjen Bimas Islam No. 352/2024). Furthermore, this research supports the Ministry of Religious Affairs’ Eight Priority Transformation Programs (Asta Protas), particularly in digitalizing governance and promoting transparent, accountable, and data-driven management. From an informatics perspective, this system demonstrates the effective implementation of decision-support algorithms in a web-based environment, highlighting the contribution of information technology to evidence-based performance evaluation.

Evaluating Synthetic Minority Oversampling Technique Strategies for Diabetes Mellitus Classification using K-Nearest Neighbors Algorithm

2025-07-31T01:44:17+00:00

Data-driven classification of Diabetes Mellitus is a crucial strategy in developing medical decision support systems that are both accurate and efficient. A major challenge in this classification task is the imbalanced class distribution, which tends to reduce the model’s sensitivity to positive cases. This research utilizes a dataset of 1,000 patient medical records obtained from the Mendeley Data repository, containing clinical attributes relevant to diabetes diagnosis. This research examines the impact of various K values on the K-Nearest Neighbors (KNN) algorithm when it is combined with the SMOTE oversampling technique to enhance classification performance. The experiment employs a 10-Fold Cross-Validation methodology with five principal assessment metrics: accuracy, precision, recall, F1-score, and Area Under Curve (AUC). Compared to prior studies, this work advances the methodology by applying SMOTE within each fold of the cross-validation process, effectively preventing data leakage and improving model generalizability. Results indicate that the K=3 configuration yields the highest F1-score of 95.13% and recall of 91.83%, while the highest AUC of 96.40% is achieved at K=9 with lower sensitivity. Applying SMOTE within each fold of the cross-validation process preserves evaluation integrity and prevents potential data leakage. The model demonstrates the ability to detect positive cases more effectively while maintaining high precision. These findings highlight that combining KNN with SMOTE and proper validation strategy is a promising approach for developing a reliable early detection system for Diabetes Mellitus that is adaptive to imbalanced clinical data.

Evaluating the Impact of Model Complexity on the Accuracy of ID3 and Modified ID3: A Case Study of the Max_Depth Parameter

2025-06-26T06:49:36+00:00

The complexity of decision tree structures has a direct impact on the generalization capability of classification algorithms. This study investigates and evaluates the performance of the classical ID3 algorithm and its modified version in the context of tree depth. The primary objective is to identify the optimal accuracy point and assess the algorithms' robustness against overfitting. Experiments were conducted across tree depths ranging from 1 to 20, with accuracy used as the main evaluation metric. The results indicate that both algorithms achieved peak performance at depth 3, followed by a notable decline. While the classical ID3 algorithm exhibited a gradual decrease in accuracy, the modified ID3 showed a sharp drop and performance stagnation between depths 11 and 20. These findings suggest that the modified ID3 algorithm enhances sensitivity in selecting informative attributes but also increases the risk of overfitting in the absence of structural regularization mechanisms. Therefore, the study recommends the implementation of regularization strategies such as pruning and cross-validation to mitigate performance degradation caused by model complexity. This research not only contributes to the theoretical understanding of how tree depth influences classification performance but also offers practical insights for developing adaptive, stable, and accurate decision tree-based classification systems.

Predicting Underweight Toddlers in Gorontalo Province Using Supervised Learning Algorithms

2025-09-28T13:47:55+00:00

Malnutrition in toddlers, notably underweight, remains a critical public health issue in Indonesia. According to the 2023 Indonesian Health Survey, the prevalence of underweight among toddlers has reached 15.9%. This condition has a significant impact on children's physical growth, cognitive development, and overall quality of life. This study aims to develop a predictive model for early detection of toddler nutritional status using three supervised machine learning algorithms: Decision Tree C4.5, K-Nearest Neighbor, and Naïve Bayes. The dataset consisted of 9,284 toddler records from Gorontalo Province, comprising eight attributes and one class label indicating nutritional status. Evaluation results showed that the Decision Tree C4.5 algorithm delivered the best performance with 98.56% accuracy. The K-Nearest Neighbor model achieved an accuracy of 97.99%, while the Naïve Bayes model obtained 96.96%. These findings demonstrate that machine learning can be an effective tool for identifying toddlers at risk of undernutrition early in their development. Beyond individual predictions, the proposed model represents a significant advancement in health informatics by providing a scalable decision-support system. This system can enhance the efficiency and precision of public health interventions, enabling faster, data-driven responses to combat malnutrition and improve child health outcomes across broader populations.

REACH: A Reinforcement Learning-Based Protocol for Adaptive Cluster Head Selection in Wireless Sensor Networks

2025-06-22T12:11:27+00:00

Wireless Sensor Networks (WSNs) are widely used in critical applications such as environmental monitoring and the Internet of Things (IoT), where energy efficiency and minimal latency are critical for network robustness and effectiveness. Conventional clustering and routing methods often struggle to adapt to fluctuating network conditions, resulting in suboptimal energy usage and increased latency. This study introduces REACH, an adaptive clustering and routing algorithm that leverages reinforcement learning to optimize energy consumption and reduce latency in WSNs. The proposed protocol dynamically selects cluster heads based on real-time network characteristics, including node density and energy levels, enhancing adaptability and robustness. Simulation results using MATLAB show significant improvements, with energy consumption reduced by 35% and latency reduced by 40% compared to traditional protocols such as LEACH and HEED. These findings suggest that reinforcement learning can significantly improve the performance of WSNs by extending the network lifetime and minimizing data transmission delay. This research contributes to the development of intelligent network protocols, offering practical insights into the integration of reinforcement learning for sustainable and scalable WSN design.

MSMEs Recommendation System using Item-Based Collaborative Filtering and LightGBM Machine Learning

2025-08-26T01:30:24+00:00

Micro, Small, and Medium Enterprises (MSMEs) face challenges in recommendation systems for digital economy growth, particularly in participatory development and sustainable revenue optimization. This study aims to develop a recommendation system using Item-Based Collaborative Filtering and LightGBM for stock prediction and item recommendation at Kedai Pesisir MSME. Based on 1,229 transaction records from January to July 2025, we performed preprocessing, feature engineering, and LightGBM training to generate daily stock predictions and monthly priorities for August 2025 to January 2026. Evaluation yielded RMSE 0.069, MAE 0.034, and MAPE 1.14%, indicating high accuracy. This advances informatics by providing a scalable AI tool for MSME inventory management and revenue enhancement, supporting strategic decisions in dynamic markets.

Comparative Analysis of CNN, SVM, Decision Tree, Random Forest, and KNN for Maize Leaf Disease Detection Using Color and Texture Feature Extraction

2025-07-17T09:34:34+00:00

Corn (Zea mays L.) is an important agricultural commodity in Indonesia, serving as the second staple food after rice and playing a crucial role in supporting national food security. However, corn production is frequently threatened by sudden outbreaks of pests and diseases, making accurate early detection essential to maintaining yield stability. This study aims to detect maize leaf diseases using five classification algorithms: Support Vector Machine (SVM), Decision Tree, K-Nearest Neighbors (KNN), Random Forest, and Convolutional Neural Network (CNN). These algorithms were tested using a combination of texture and color features, including Gray Level Co-occurrence Matrix (GLCM), Local Binary Pattern (LBP), Hue-Saturation-Value (HSV), and L*a*b*. The dataset used consists of 2,048 maize leaf images classified into four categories: Blight, Common Rust, Gray Leaf Spot, and Healthy, with 512 images per class. Each class was divided into training and testing sets to train and evaluate the classification models.

The results show that CNN achieved the highest accuracy of 93.93% when using a complete combination of color and texture features. Meanwhile, SVM also demonstrated high performance, achieving the same accuracy (93.93%) using only the combination of color features (HSV and Lab*). Random Forest and Decision Tree performed best when using color features alone, with accuracies of 89.81% and 87.14%, respectively. These findings indicate that color features have a dominant influence on classification accuracy, and that combining color and texture features can significantly enhance model performance, particularly in CNN architectures. This study contributes to the development of early disease detection systems in precision agriculture.

Accurate Skin Tone Classification for Foundation Shade Matching using GLCM Features-K-Nearest Neighbor Algorithm

2025-05-27T04:06:21+00:00

Foundation shade matching remains a significant challenge in the beauty industry, particularly in Indonesia where consumers exhibit three distinct skin tone categories: ivory white, amber yellow, and tan. Manual foundation selection often results in mismatched shades, leading to customer dissatisfaction. This study presents a novel automated skin tone classification system combining Gray Level Co-Occurrence Matrix (GLCM) feature extraction with the K-Nearest Neighbor (KNN) algorithm. The GLCM method extracts four key texture features (contrast, homogeneity, energy, and entropy) from facial images, while KNN performs classification. A comprehensive dataset of 963 facial images was used, with 770 training and 193 test samples collected under controlled lighting conditions. After testing K values from 1 to 15, the optimal K=1 achieved 75.65% accuracy. Compared to baseline color histogram methods (60% accuracy), our GLCM-KNN approach demonstrates 15.65% improvement in classification performance. This research contributes to computer vision applications in beauty technology, enabling the development of mobile applications for virtual foundation try-on and personalized product recommendations. The findings have significant implications for the cosmetics industry, particularly for automated cosmetic shade matching systems and enhanced customer experience in online beauty retail. Further research is recommended to explore deep learning approaches and expand dataset diversity to improve accuracy.

Mapping Gestures Based on Text Emotion Classification for a Virtual Chatbot for Early Marriage Consultation in Lombok Using RoBERTa Model

2025-07-15T15:03:13+00:00

To address the persistent issue of early marriage among Indonesian adolescents, this study proposes a virtual counseling chatbot that classifies emotional cues in text using a fine-tuned IndoRoBERTa model. The emotion classification framework is designed to support counseling-based prevention efforts by moving beyond basic sentiment analysis and adopting five functional emotional categories such as ‘Enthusiastic’, ‘Gentle’, ‘Analytical’, ‘Inspirational’, and ‘Cautionary’ to align with psychological counseling styles. Built on fine-tuned IndoRoBERTa architecture, the model was trained in two phases: first with 2,500 manually validated samples yielding 92.8% accuracy, and then with 12,500 auto-labeled entries, resulting in 91.3% accuracy. Performance was assessed using accuracy, precision, recall, and F1-score. A gesture mapping layer was also integrated to enhance empathetic response generation. Each emotion label was paired with a predefined body gesture, grounded in counseling theory, to support future development of multimodal virtual agents capable of expressing emotions both textually and physically. The novelty lies in combining context-aware emotion classification with gesture mapping, enabling future development of expressive, culturally relevant, and empathetic virtual chatbot agents.

Designing AI - IoE Precision Farming to Create Sustainable Eco-Friendly Hydroponic Greenhouses

2025-08-13T00:57:01+00:00

Conventional greenhouses, while boosting crop yields, face critical sustainability challenges due to high energy consumption and resource inefficiency, particularly in developing nations where manual management prevails. This research addresses these limitations by designing a comprehensive AI-IoE system architecture to create a smart, resource-efficient, and sustainable operational model for eco-friendly greenhouses. The development methodology involved a systematic process of requirements analysis, integrated hardware and software design, prototype assembly, and functional testing. The system utilizes an ESP32 microcontroller as its central control unit, integrating a suite of six sensors comprising light intensity, temperature, humidity, pH, Total Dissolved Solids (TDS), and CO₂ to monitor critical environmental parameters in real-time. This integration utilizes the extensive dataset for AI based predictive analysis, enabling the intelligent forecasting of environmental trends and proactive resource management. The research resulted in a complete system blueprint, including a detailed electronic circuit design, a production-ready Printed Circuit Board (PCB) layout, defined operational control logic, and an intuitive web-based dashboard for remote monitoring and management. This integrated AI-IoE architecture provides a tangible solution that surpasses previous fragmented approaches by offering holistic environmental control. The findings present a significant contribution to precision farming, establishing a scalable and efficient framework to enhance greenhouse productivity and ecological sustainability.

A General-Purpose Web-based TOPSIS tool for Accessible Multi-Criteria Decision Making

2025-08-04T16:08:10+00:00

Multi-Criteria Decision Making (MCDM) is a crucial framework for evaluating alternatives based on diverse and often conflicting criteria. However, most existing MCDM tools still require technical skills or software installation, which limits their accessibility for non-technical users. This study introduces Topsisku, a general-purpose web-based decision support system that implements the TOPSIS method. Topsisku enables users to upload datasets, assign weights and criterion types, and obtain ranking results directly through an intuitive web interface. The system was evaluated using three case studies from prior literature, each representing a different domain. Results indicate that the system’s computations are identical to manual calculations, with Spearman rank correlation coefficients approaching perfection (ρ ≈ 1.0; p < 0.001). A usability test involving 10 respondents yielded an average System Usability Scale (SUS) score of 76.5, placing Topsisku in the Good usability category. These findings confirm that Topsisku is both mathematically accurate and user-friendly. The primary contribution of this study lies in democratizing the application of MCDM for non-technical users while maintaining the reliability of the TOPSIS method. Future research directions include the development of advanced sensitivity analysis modules, integration of collaborative multi-user features, and the incorporation of artificial intelligence techniques to enhance system adaptability and decision-making support.

Optimizing Data Augmentation Parameters in YOLOv11 for Enhanced Rip Current Detection on Small Datasets from Depok-Parangtritis Coastline

2025-09-30T12:44:14+00:00

Rip currents are powerful ocean currents that can suddenly pull swimmer offshore and are often difficult to recognize visually. However, automatic monitoring technology for detecting rip currents is still limited, while small datasets often lead to overfitting problems and reduce detection accuracy. This study aims to optimize data augmentation parameters in YOLOv11 to improve the mean Average Precision (mAP) value and enable rip current detection even with limited data. The dataset was collected from Google Earth and aerial photographs from the Depok-Parangtritis coastline. Preprocessing includes manual labelling, cropping, and resizing to 640 x 640 pixels. Four augmentation techniques were applied, namely crop (0-10%), rotation (-10% to +10%), brightness adjustment (-10% to +10%), and 1 pixel blur using Roboflow. The dataset was split into 70% training and 30% validation. The YOLOv11 model was then trained and evaluated with precision, recall, and mAP metrics. Results show that data augmentation significantly improves model performance. Dataset 2 without augmentation achieved only 31.8% precision, 32.8% recall, and 23.8% mAP50, while the best model from a combination of the original Dataset 1 and the augmented Dataset 3 reached 90.6% precision, 85.7% recall, and 90.4% mAP50. The integration of YOLOv11 into a web application enables automatic detection in both images and videos with bounding box and confidence score. This study emphasizes the importance of visual variation in the dataset for improving the model generalization and provide a practical foundation of real-time coastal monitoring system.

Data-Driven Student Group Formation for Group Investigation: A K-Medoids Clustering Approach in Cooperative Learning

2025-06-21T07:28:43+00:00

Group Investigation (GI) is a widely used cooperative learning strategy in higher education, but challenges such as large class sizes and diverse student profiles complicate manual group formation. Previous studies have applied clustering algorithms like K-Means, yet K-Medoids, which is robust to noise, remain underexplored for group formation, especially GI. This study proposed a data-driven approach using the K-Medoids clustering algorithm to create student groups that are both interest-aligned and heterogeneous in profile, which enhancing the effectiveness of GI activities. Employing the Knowledge Discovery in Databases (KDD) framework, the process included data selection, preprocessing, transformation, three grouping processes, and evaluation were performed. In grouping process students were initially grouped by interest, clustered using K-Medoids with various distance measures tested, and finally, groups were adjusted to balance homogeneity and diversity. In grouping stage 2, clustering with Euclidean distance and PCA achieved the highest Silhouette Score, indicating superior grouping quality. The result of heterogeneity group of students evaluated with Gower dissimilarity shows that the method produces internally diverse yet cohesive interest groups, supporting GI goals.

Performance Comparison of AdaBoost, LightGBM, and CatBoost for Parkinson's Disease Classification Using ADASYN Balancing

2025-05-26T04:16:43+00:00

Parkinson's disease is a neurodegenerative condition identified by the decline of neurons that produce dopamine, causing motor symptoms such as tremors and muscle stiffness. Early diagnosis is challenging as there is no definitive laboratory test. This study aims to improve the accuracy of Parkinson's diagnosis using voice recordings with machine learning algorithms, such as AdaBoost, LightGBM, and CatBoost. The dataset used is Parkinson's Disease Detection from Kaggle, consisting of 195 records with 22 attributes. The data was normalized with Min-Max normalization, and class imbalance was resolved with ADASYN. Results show that ADASYN-LightGBM and ADASYN-CatBoost have the best performance with 96.92% accuracy, 97.10% precision, 96.92% recall, and 96.92% F1 score. This improvement suggests that combining boosting methods and data balancing techniques can improve the accuracy of Parkinson's diagnosis. These results demonstrate the effectiveness of ADASYN in addressing data imbalance and improving the performance of boosting algorithms for medical classification problems. The findings contribute to the development of intelligent diagnostic systems in the field of medical informatics and computer science. These findings are essential for developing more accurate and efficient diagnostic tools, supporting early diagnosis and better management of Parkinson's disease.

Deep CNN for Wetland Mapping from Satellite Imagery

2025-08-31T08:46:05+00:00

Wetland loss endangers the ecosystem through loss of biodiversity, carbon sequestration and flood regulation potential. A precise determination of wetlands status is necessary to safeguard for their conservation and ensure sustainable management. Implementation This study aims to assess the performance of deep CNNs in wetland detection using high-resolution Google Earth image data in South Kalimantan province, Indonesia. The work adopts the Chopped Picture Method (CPM) and the use of sliding windows for data augmentation to improve the diversity of the dataset and reduce the computational cost. Two CNN models, VGG-16Net, and LeNet-5, were trained using a dataset comprising 220 satellite images, which we converted into 89,100 patches of 56×56 pixels. Performance was compared using accuracy, precision, recall, and F1-score. Experimental results show good levels of accuracy for the two architectures, but LeNet-5 provided more stable results between test locations, having a F1-score closer to 100% and spending less computational time (≈10s per epoch) than VGG-16Net (≈40s per epoch). These results validate that CPM significantly increases the variety of training data, making it possible for a CNN to correctly identify the vague and irregular shapes of wetlands with high accuracy. In addition to advancing environmental conservation strategies, the study highlights the contribution of informatics to large-scale, automated environmental monitoring, particularly in supporting wetland conservation, sustainable land-use planning, and climate adaptation efforts.

Modification Of Yolov11 Nano And Small Architecture For Improved Accuracy In Motorcycle Riders Face Recognition Based On Eye

2025-06-13T07:58:36+00:00

Face recognition still faces challenges in identifying faces covered by masks and helmets with open visors, such as those commonly used by motorcyclists, especially when entering parking areas. To improve the accuracy of face recognition in these conditions, this study proposes nano and small versions of the YOLOv11 modification, which is an internal version. Modifications are made to the neck section and the DySample module is added in place of the UpSample module to improve the model's capabilities. Experiments were conducted using a self-generated dataset consisting of 50 classes. The results show that the modified nano version achieves 99.3% accuracy at the same mAP50 as YOLOv11n and YOLOv12n. At mAP50-95, it shows a 1.6% accuracy improvement compared to YOLOv11n and YOLOv12n with 75% accuracy. Meanwhile, the modified small version achieved an accuracy improvement of 1.3% and 1.2% compared to YOLOv11n and YOLOv12n, respectively, reaching 76.1% on mAP50-95, although the accuracy on mAP50 remained the same as YOLOv11n and 0.1% superior to YOLOv12n. However, recall and precision did not show significant improvement in both as well as the increase in model parameters. However, the model is still in the nano and small versions. Therefore, the model can be implemented on edge devices. This research is important for the field of computer vision, especially in the context of face recognition. The contribution of this research is the improvement of the accuracy of the mAP50-95 metric in eye-based face recognition, which is relevant for intelligent security systems with limited resources.

Development of a Strategy for Independent Village Development Based on IDM Predictions Using Linear Regression: A Study of Bi'ih Village, South Kalimantan

2025-08-12T02:13:46+00:00

Village development in Indonesia still faces inequality due to uneven utilization of technology, resulting in many villages lagging behind economically despite their strong social and environmental potential. This study aims to predict the status of the Village Development Index (IDM) of Bi’ih Village in 2025 and formulate development strategies toward an independent village through the integration of information technology and local wisdom. The method used is linear regression analysis of IDM data from 2017 to 2024, followed by a gap analysis comparing the achievements of Bi’ih Village with those of 11 self-reliant villages in Karang Intan Subdistrict. The prediction results show that the IDM value of Bi’ih Village in 2025 will be 0.7744, placing it in the Advanced category, 0.0411 points below the Independent threshold (>0.8155). The Social Resilience Index (IKS = 0.8782) and Environmental Resilience Index (IKL = 0.8388) have exceeded the Independent threshold, while the Economic Resilience Index (IKE) remains low at 0.6061. The main constraints include limited access to digital markets, manual financial record-keeping, and low digital literacy among village entrepreneurs. The novelty of this research lies in the formulation of a development strategy based on the integration of ICT and local wisdom, with a focus on strengthening digital BUMDes, implementing integrated financial systems, developing upstream–downstream partnerships, and branding local products. This approach demonstrates that digitalization rooted in local wisdom can enhance economic productivity, strengthen resilience, and support sustainable progress toward independent-village status.

Clustering and Modeling of Daily Weather Pattern Distribution in Makassar City Using Hybrid DBSCAN-Gaussian Mixture Model

2025-09-03T02:18:56+00:00

Dynamic and irregular daily weather changes present major challenges in understanding seasonal patterns. Data uncertainty, outliers, and inter-season variability further complicate weather analysis using conventional methods. To address this issue, this study integrates Density-Based Spatial Clustering of Application with Noise (DBSCAN) and Gaussian Mixture Model (GMM) to analyze daily weather patterns in Makassar City. A total of 2,192 daily records from 2019 to 2024, including rainfall, specific humidity, atmospheric pressure, and wind speed, were examined. DBSCAN detected one dominant cluster (2019 data) and 173 outliers. The main cluster was further partitioned by GMM into three sub-clusters representing the wet (511 records, 13.39 mm rainfall), dry (633 records, 0.15 mm), and transition (875 records, 2.53 mm) seasons. GMM identified 1,764 fixed clusters and 255 ambiguous data points, with a log-likelihood of 5091.22 and the highest Silhouette Score of 0.188. Comparative evaluation demonstrated that the hybrid DBSCAN-GMM achieved superior performance (Silhouette Score = 0.1434) compared to DBSCAN or GMM individually. The novelty of this research lies in applying the DBSCAN-GMM integration, which is rarely used in tropical weather analysis, to capture seasonal structure and anomalies adaptively. This study contributes methodologically to clustering-based weather modeling and practically supports applications such as agricultural planning, disaster mitigation, and adaptive climate strategies in tropical regions.

Data Augmentation Techniques on the Accuracy of Fertile and Infertile Egg Classification Using Convolutional Neural Networks

2025-08-11T07:12:22+00:00

The classification of fertile and infertile chicken eggs is crucial in the poultry industry to ensure optimal incubation efficiency and hatchability. However, the visual similarity between both egg types under candling conditions poses a significant challenge for manual inspection. This study aims to develop a convolutional neural network (CNN) model using the EfficientNetB4 architecture to automatically classify egg fertility based on image data. The dataset comprises candling images of chicken eggs, which underwent preprocessing steps such as resizing, normalization, and histogram stretching to enhance contrast. To improve model generalization, aggressive data augmentation techniques were applied, including rotation, flipping, zooming, and brightness adjustment. The model was trained in two phases—feature extraction and fine-tuning—using transfer learning and class balancing strategies. Evaluation results demonstrated high performance with an F1-score of 0.95 and balanced classification across both classes. The model's interpretability was further enhanced using Grad-CAM visualization, showing relevant activation regions. These findings indicate that the proposed method is effective in automating egg fertility classification and has potential for broader application in agricultural image diagnostics.

Multi-Class Brain Tumor Segmentation and Classification in MRI Using a U-Net and Machine Learning Model

2025-10-07T22:40:57+00:00

Brain tumor diagnosis remains a critical challenge in medical imaging, as accurate classification and precise localization are essential for effective treatment planning. Traditional diagnostic approaches often rely on manual interpretation of MRI scans, which can be time-consuming, subjective, and prone to variability across radiologists. To address this limitation, this study proposes a two-stage framework that integrates machine learning (ML) based classifiers for tumor type recognition and a U-Net architecture for tumor segmentation. The classifier was trained to distinguish four tumor categories: glioma, meningioma, pituitary, and no tumor, while the U-Net model was employed to delineate tumor regions at the pixel level, enabling volumetric assessment. The novelty of this research lies in its dual focus that combines classification and segmentation within a single framework, which enhances clinical applicability by offering both diagnostic and spatial insights. Experimental results demonstrated that among the evaluated classifiers, XGBoost achieved the highest accuracy of 86 percent, surpassing other models such as Random Forest, SVC, and Logistic Regression, while the U-Net model delivered consistent segmentation performance across tumor types. These findings highlight the potential of hybrid ML and deep learning solutions to improve reliability, efficiency, and objectivity in brain tumor analysis. In real-world practice, the proposed framework can serve as a valuable decision-support tool, assisting radiologists in early detection, reducing diagnostic workload, and supporting personalized treatment strategies.

Improving Model Capability for Sentiment Trend Analysis in Hotel Visitor Reviews with Bi-LSTM Multistage Approach

2025-07-28T14:23:10+00:00

This study focuses to improve the sentiment analysis of hotel reviews using Multistage mechanism of two-stage approach based on the Bidirectional Long Short-Term Memory (Bi-LSTM) architecture with 53,000 data from 28 hotels in Yogyakarta that captured from google maps review for hotel in Yogyakarta. Hotel customer reviews often contain mixed sentiment expressions, making it crucial to filter out only sentences with a single dominant sentiment to avoid ambiguity. In the first stage, the model detects sentiment at the token level and counts the number of sentiment expressions in each sentence. Only sentences with a single polarity are passed to the final classification stage. In the second stage, the overall sentiment is classified as positive, negative, or neutral using pooled contextual representations. Experimental results from 30 iterations demonstrate consistently high performance, with precision, recall, and F1-scores above 0.95, and overall accuracy exceeding 96%. The confusion matrix analysis shows strong model performance, although some challenges remain in distinguishing between positive and neutral sentiment. Additionally, sentiment trend analysis of hotel reviews from properties such as Lafayette Boutique Hotel and The Westlake Resort Jogja reveals dynamic shifts in guest perception over time. This multistage mechanism approach proves effectiveness of improving sentiment classification accuracy by avoid the bias on sentiment and also in providing valuable temporal insights for monitoring customer satisfaction.

Predicting Hypnotherapy Effectiveness Using Ensemble Learning: A Case Study at The Mind Solution Hypnotherapy Clinic

2025-06-17T02:12:15+00:00

Mental health is recognized as a universal human right, yet effective interventions for psychological disorders like anxiety and phobias remain challenging. Hypnotherapy shows promise but suffers from variable effectiveness across individuals, compounded by limited data-driven tools for outcome prediction in clinical settings, particularly in Indonesia where social stigma impedes accessibility. This study aims to (1) identify demographic/clinical factors influencing hypnotherapy success, (2) develop an ensemble learning-based predictive model, and (3) evaluate its performance against conventional methods. Using retrospective data from 276 patients at Mind Solution Hypnotherapy Clinic, we implemented preprocessing (missing values imputation, label encoding) and trained Decision Tree and Random Forest models via Orange Data Mining, validated through *5-fold cross-validation*. Results demonstrate Random Forest superiority (accuracy: 92.7%; precision: 94.2%; AUC: 0.918) over Decision Tree, with key predictors being gender (32.54% gain ratio), occupation (31.75%), and birth order (15.58%). Notably, 71.5% of patients achieved improvement in just one session. These findings confirm ensemble learning’s efficacy in personalizing hypnotherapy protocols, offering clinicians a decision-support tool to optimize resource allocation. The study bridges AI and mental health practice, providing empirical evidence to reduce societal stigma while advancing predictive analytics in psychotherapy.

Systematic Optimization of Ensemble Learning for Heart Failure Survival Prediction using SHAP and Optuna

2025-09-24T11:02:52+00:00

Heart failure (HF) stands as a major global health problem where precise and early prediction of patient prognosis is essential for improving clinical management and patient care. A common obstacle for standard machine learning models in this domain is the prevalent issue of class imbalance within clinical datasets. To overcome this challenge, this study introduces a systematically optimized ensemble learning model for the accurate classification of patient survival. The methodology was applied to a publicly accessible clinical dataset of 299 heart failure patients. Its comprehensive framework included logarithmic transformation, stratified data splitting (80:20), SHAP-based selection of eight key features, and hyperparameter tuning with Optuna over 75 trials, with the specific objective of maximizing the F1-score using 10-fold cross-validation. The performance of three ensemble models (Random Forest, XGBoost, and LightGBM) was refined using decision threshold tuning. The results revealed that the fully optimized Random Forest model yielded superior outcomes, attaining an accuracy of 96.67%, an F1-score of 0.9474, and precision and recall values of 0.95, demonstrating high reliability with only a single instance of a False Negative and False Positive. The study concludes that the systematic application of SHAP, SMOTE, and Optuna within an ensemble framework substantially improves classification performance for imbalanced HF data, surpassing existing benchmarks. This work thus provides a replicable and systematic framework for developing reliable machine learning models from complex, imbalanced medical datasets, contributing a valuable methodology to the field of computational science.

The Role of Deep Learning in Cancer Detection: A Systematic Review of Architectures, Datasets, and Clinical Applicability

2025-07-15T16:03:14+00:00

Early cancer detection continues to be a significant challenge in clinical practice due to limitation of conventional diagnostic technique that often takes time and error prone. This systematic review evaluates the efficacy of deep learning (DL) architecture and datasets to improve cancer detection and diagnosis. We performed a structural analysis on 40 high-impact research paper published in Q1 journals between 2014 and 2025, considering DL model performance, datasets, and clinical relevance. Results indicate that fundamental architectures such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs) consistently report high diagnostic accuracy (>90%) on radiology- and histopathology-based imaging datasets. Conversely, DL performance on non-imaging clinical data, including electronic medical records (EMDs), is more varied. Evaluation metrics such as AUC and DICE shows the trade-off between classification precision and segmentation accuracy. Despite their potential, DL models have significant limitations in terms of generalization, interpretability, and integration within real-world clinical workflows. This review highlights the need for standardized evaluation, implementation of ethical models, and multi-modal data fusion to facilitate wider and more equitable clinical uptake of DL in cancer diagnostics.

Development of a Church Information Management System Using Scrum at HKBP Sola Gratia Kayu Mas Jakarta

2025-08-06T22:02:52+00:00

The rapid growth of the congregation at HKBP Sola gratia Kayu Mas Church in Jakarta has posed challenges in managing member data efficiently and effectively. The previous data management system, which relied on Microsoft Excel, showed significant limitations in data retrieval, family grouping, and presenting birthday or elderly member information. This study aims to develop a web-based church congregation management information system using the Scrum methodology as an iterative and flexible software development approach. The research methodology includes observation, interviews, literature review, and black box testing. The results indicate that the developed system successfully meets user needs, simplifies congregation data management, and enhances the effectiveness of church administrative services. The implementation of Scrum has proven to be effective in accelerating development processes, accommodating changing requirements, and increasing user involvement through continuous evaluation. This system is expected to be replicable in other churches with similar needs as an integrated digital solution for congregation management.

Forecasting Bitcoin Price Prediction with Long Short-Term Memory Networks: Implementation and Applications Using Streamlit

2025-08-05T04:23:46+00:00

The rapid growth of cryptocurrency markets, particularly Bitcoin, has highlighted the need for accurate price prediction models to support informed decision-making. While existing studies primarily evaluate machine learning models for price forecasting, few have implemented these models in real-world applications. This paper addresses this gap by developing a Bitcoin price prediction system using Long Short-Term Memory (LSTM) networks, integrated into a user-friendly web-based application powered by Streamlit. The model forecasts Bitcoin prices at 5-minute, 1-hour, and 1-day intervals, demonstrating strong predictive performance. For the 5-minute interval, the model achieved a Mean Squared Error (MSE) of 53,479.86, Mean Absolute Error (MAE) of 150.58, Root Mean Squared Error (RMSE) of 231.26, and Mean Absolute Percentage Error (MAPE) of 0.144%. At the 1-hour interval, errors increased moderately with an MSE of 423,198.24, MAE of 499.93, RMSE of 650.54, and MAPE of 0.505%. For the 1-day interval, the model faced greater variability, reflected in an MSE of 3,089,699.07, MAE of 1,058.88, RMSE of 1,757.75, and MAPE of 2.027%. These results indicate that while predictive precision decreases over longer horizons, the model maintains strong performance across all timeframes. By embedding LSTM predictions into an interactive, real-time forecasting platform, this study demonstrates the practical integration of deep learning into complex financial systems. Beyond cryptocurrency, the approach highlights the potential of intelligent computational models to enhance decision-making processes in data-intensive domains, reinforcing the role of informatics in bridging advanced algorithms with usable technological solutions.

Comparison of Information Technology Governance Maturity Levels Based on COBIT 2019 at PT Kereta Commuter Indonesia in 2023 and 2024

2025-07-31T01:53:38+00:00

This study aims to analyze and compare the maturity level of Information Technology (IT) governance at PT Kereta Commuter Indonesia (KCI) between 2023 and 2024 using the COBIT 2019 framework. The background of this study is based on the operational complexity of KCI which serves a high daily passenger volume, so that the information system becomes the backbone of the smooth transportation service. The method used is a descriptive-comparative case study with a mixed approach, through interviews, Likert scale questionnaires, and internal document reviews such as IT audit reports and government regulations. The results of the analysis show a significant and consistent increase, where the level of IT governance maturity which was previously at level 2 (Managed) and 3 (Defined) in 2023, increased to level 3 and 4 (Quantitatively Managed) in 2024. The most prominent improvements were seen in the strategic domain EDM01 (Ensure Governance Framework Setting) and the operational domain DSS01 (Manage Operations), which successfully reached level 4. This success reflects top management's commitment and ongoing internal evaluation in strengthening IT governance strategically and operationally. The research findings confirm that annual evaluations serve as an objective benchmark for identifying governance gaps, developing digital strategies, and determining future IT investment priorities. Overall, this study confirms that regular assessments can improve the effectiveness of data-driven IT transformation and ensure alignment of IT implementation with the company's business objectives.

Sentiment Analysis Of Indihome Service Based On Geo Location Using The Bert Model On Platform X

2025-08-05T04:45:09+00:00

The rapid growth of internet usage in Indonesia has led more people to express their feelings, whether positive or negative, about online services, including IndiHome, through social media platforms such as X (formerly Twitter). This study aims to analyze public sentiment toward IndiHome services based on geographic location using the IndoBERT natural language processing model. The data consists of 3.307 Indonesian tweets that are geo-tagged and categorized into three sentiment groups: good, okay, and bad. The research process involves collecting the data, cleaning it (organizing and splitting words), and testing the IndoBERT model with a confusion matrix and classification scores. The findings reveal that negative feelings are more prevalent in most locations, especially in Java. The IndoBERT model achieved its highest accuracy of 80% in detecting negative sentiment. However, there is still room for improvement in distinguishing between positive and neutral sentiments, possibly due to data imbalance. The study shows how combining sentiment analysis with geo-location can provide strategic insights to service providers. In practical terms, these insights can help IndiHome prioritize infrastructure upgrades, improve customer support in areas with high dissatisfaction, and assist policymakers in promoting fairer digital access across regions. Beyond these practical implications, this study also contributes to the field of informatics by demonstrating the application of a transformer-based NLP model (IndoBERT) combined with geo-tagged data for regional sentiment mapping- a relatively unexplored approach in the Indonesian context. The integration of geospatial analysis with sentiment classification offers methodological advances for NLP-based service evaluation beyond business applications.

Classification of Worker Productivity and Resource Allocation Optimization with Machine Learning: Garment Industry

2025-08-16T05:46:20+00:00

This study presents an integrated predictive–prescriptive framework for improving workforce management in the garment industry by combining machine learning classification with linear programming optimization. Using a publicly available dataset of 1,197 production records, productivity levels were categorized into low, medium, and high classes. Data preprocessing included handling missing values, one-hot encoding of categorical variables, and class balancing using SMOTE. Eleven classification algorithms were evaluated, with LightGBM achieving the highest performance (accuracy 78.3%, weighted F1-score 78.3%, Cohen’s Kappa 63.4%) after hyperparameter tuning via Bayesian Optimization. The optimized model’s predictions were then incorporated into a linear programming model, implemented with PuLP, to maximize the allocation of high-productivity workers across production departments under capacity constraints. The results yielded an allocation plan assigning 117 high-productivity workers, significantly enhancing potential operational efficiency. The novelty of this work lies in integrating an optimized ensemble learning model with mathematical programming for end-to-end productivity classification and resource allocation, a combination rarely explored in labor-intensive manufacturing contexts. This framework offers a scalable decision-support tool for data-driven workforce planning and could be adapted to other manufacturing domains with similar operational structures.

Brain Tumor Auto Segmentation On 3D MRI Using Deep Neural Network

2025-07-20T09:00:08+00:00

Accurate and automated segmentation of brain tumours from Magnetic Resonance Imaging (MRI) is crucial for clinical diagnosis and treatment planning, yet it remains a significant challenge due to tumour heterogeneity and data imbalance. This research investigation examines the effectiveness of a 3D UNet architecture for the segmentation of brain tumours utilizing MRI imaging modalities. The research employs the BRATS 2021 dataset, which consists of 675 MRI datasets across four distinct imaging modalities (FLAIR, T1-Weighted, T1-Contrast, and T2-Weighted) and encompasses four distinct segmentation label classes. The employed model integrated soft dice loss and dice coefficient as its loss functions, with the objective of achieving convergence despite the presence of imbalanced data. While constraints related to resources limited the training process, the model yielded promising outcomes, exhibiting high accuracy (99.43%) and specificity (99.5%), The model aids medical professionals in understanding tumor growth and enhances treatment planning via segmentation predictions in surgery. Nevertheless, the sensitivity, particularly concerning non-enhancing tumour classes, persists as a significant challenge, underscoring the necessity for future research to concentrate on data-centric methodologies and enhanced pre-processing techniques to improve model efficacy in critical medical applications such as the segmentation of brain tumours.

Single-Image Face Recognition For Student Identification Using Facenet512 And Yolov8 In Academic Environtment With Limited Dataset

2024-11-11T18:05:00+00:00

Face recognition has become one of the most significant research areas in image processing and computer vision, mainly due to its wide applications in security, identity verification, and human and machine interaction. In this study, FaceNet512 and YOLOv8 models are used to overcome the challenges in face recognition with a limited dataset, which is only one formal photo per individual. The application of image augmentation to the model achieved 90% accuracy and ROC curve of 0.82, while the model without augmentation achieved 89% accuracy and ROC curve of 0.79. FaceNet512 showed superiority in producing more accurate and detailed facial representations compared to other models, such as ArcFace and FaceNet, especially in handling minimal facial variations. Meanwhile, YOLOv8 provides efficient face detection across various lighting conditions and viewing angles. The main challenge in this research is the low quality of the original image, which can reduce the accuracy of face recognition. These results show the great potential of using deep learning-based face recognition systems in the real world, especially for automatic attendance applications in academic environments.

Improving the Performance of Machine Learning Classifiers in Sentiment Analysis of Jenius Application Using Latent Dirichlet Allocation in Text Preprocessing

2025-08-11T07:14:43+00:00

Sentiment analysis aims to classify a person’s opinion into a specific sentiment, such as positive or negative. The choice of preprocessing used can influence the performance of a sentiment analysis model. The Latent Dirichlet Allocation (LDA) method, commonly used for topic modelling, can be employed as an additional preprocessing step to identify relevant words associated with a particular sentiment label. This study aims to assess whether the LDA method, implemented in the preprocessing stage, can enhance the performance of machine learning models, including Naïve Bayes, Decision Tree, KNN, Logistic Regression, and SVM. This study utilized a dataset comprising 1,800 reviews, with 900 labelled as positive and 900 as negative. Words with an LDA score of at least 0.15 were given additional weight in the TF-IDF stage before model training. After the model was developed, evaluation was carried out by calculating accuracy, precision, recall, and F1-score. The use of LDA in preprocessing improved the performance of all classification models by 1-3% across most evaluation metrics. Specifically, the Logistic Regression model achieved the best performance, followed by SVM and KNN. This performance improvement is aligned with the use of LDA to reduce semantic noise and improve feature representation. Furthermore, this research is also helpful for monitoring customer opinions in the digital banking sector, enabling the rapid and accurate identification of priority issues. Further research could explore the comparison of performance with other topic modelling and feature extraction methods, as well as expanding the dataset and utilizing multiclass models.

IoT-Enabled Real-Time Monitoring and Tsukamoto Fuzzy Classification of Mandar River Water Quality via Web Integration for Sustainable Resource Management

2025-08-06T11:31:10+00:00

This study presents the design and implementation of a real-time water quality monitoring system that utilizes pH, Total Dissolved Solids (TDS), and turbidity sensors, integrated with an ESP32 microcontroller. Sensor data are processed using the Tsukamoto fuzzy logic method to classify river water suitability into two categories: Suitable and Not Suitable. This approach effectively addresses imprecise and uncertain data, thereby producing more reliable classifications compared to conventional threshold-based methods. System validation was conducted through field testing over seven consecutive days at four different times of the day (morning, midday, afternoon, and evening), with results demonstrating stable performance. Recorded pH values ranged from 7.02 to 9.96, TDS values from 140 to 176 ppm, and turbidity levels between 4.00 and 5.15 NTU, indicating that the Mandar River remains within safe limits for daily use. The novelty of this study lies in the direct implementation of the Tsukamoto fuzzy logic method on a resource-constrained IoT device (ESP32), enabling edge-level classification with low latency and without full reliance on cloud computing. The system is designed to maintain decision reliability even under fluctuating sensor data, thus offering a practical and integrated solution for real-time monitoring. The main contribution of this work to computer science is the demonstration of lightweight embedded intelligent algorithms capable of running on constrained devices, the reinforcement of Explainable AI through transparent linguistic rules, and the integration of IoT with edge computing to support sustainable resource management in real-time.

An Integrated Pipeline with Hierarchical Segmentation and CNN for Automated KTP-el Data Extraction on the e-Magang Platform

2025-08-31T08:48:18+00:00

In alignment with Indonesia's digital transformation agenda, this research addresses the inefficiencies and error-prone nature of manual data entry on the Foreign Policy Strategy Agency's (BSKLN) e-magang platform. This study introduces a comprehensive, end-to-end Optical Character Recognition (OCR) pipeline, specifically designed for structured identity documents and real-world government platform integration. The proposed methodology features a robust workflow, including image preprocessing with histogram matching, hierarchical segmentation using vertical projection, and intelligent postprocessing to structure the output. To overcome the limitations of a small dataset, three specialized Convolutional Neural Network (CNN) models were rigorously trained and validated using a stratified 5-fold cross-validation technique. The final system was successfully integrated, connecting a Flask-based model engine with the existing Laravel and React platform. End-to-end testing demonstrated strong performance, achieving an average character-reading accuracy of 93.31% with a mean processing time of 14.48 seconds per image. The primary contribution of this research to the field of informatics is the development of a complete and deployable system architecture that ensures data interoperability and reliability, providing a practical blueprint for integrating intelligent automation into digital public services.

Enhancing Chronic Kidney Disease Classification Using Decision Tree And Bootstrap Aggregating: Uci Dataset Study With Improved Accuracy And Auc-Roc

2025-08-14T11:26:33+00:00

Chronic Kidney Disease (CKD) is a progressive medical disorder that requires timely and precise identification to avoid permanent impairment of kidney function. However, Decision Tree models, although widely used in clinical applications due to their transparency, ease of implementation, and ability to handle both categorical and numerical data, are prone to overfitting and instability when applied to small or imbalanced datasets. The purpose of this study is to optimize CKD classification by integrating Bootstrap Aggregating (Bagging) with Decision Tree to enhance accuracy and robustness. The methodology involves testing two model variants a standalone Decision Tree and a Bagging-supported Decision Tree using 10-fold cross-validation and evaluating performance with accuracy, precision, recall, F1-score, and the area under the ROC curve (AUC-ROC). Findings reveal that Bagging enhances model accuracy from 0.980 to 0.987, raises precision from 0.976 to 1.000, and improves recall from 0.954 to 0.954, and increases F1-score from 0.965 to 0.976. These results demonstrate that Bagging significantly improves the reliability and generalizability of Decision Tree classifiers, making them more effective for CKD prediction.

Enhanced U-Net Cnn For Multi-Class Segmentation And Classification Of Rice Leaf Diseases In Indonesian Rice Fields

2025-08-16T05:47:57+00:00

Rice is a strategic food crop whose productivity is often threatened by leaf diseases and pests. This study aims to develop an Enhanced U-Net CNN model for multi-class segmentation and classification of rice leaf conditions from field images to support early detection and plant health management. The methodology includes direct field image acquisition of rice leaves, preprocessing for image quality enhancement, expert data labeling, segmentation using a U-Net architecture, and classification using CNN. The dataset was divided into training and testing data with balanced distribution across four classes: Healthy, BrownSpot, Hispa, and LeafBlast. Evaluation results show that the model can identify rice leaf conditions with high accuracy, although signs of overfitting were observed from the performance gap between training and validation data. The implementation of this model is expected to accelerate automatic disease detection in the field, reduce reliance on manual inspection, and support precision agriculture. This study achieved a testing accuracy of 76.36% with a macro-average F1-score of 0.34. While the results indicate limitations in generalization, the proposed Enhanced U-Net CNN demonstrates the feasibility of combining segmentation and classification in field conditions. This research contributes to agricultural informatics by supporting scalable deployment in precision agriculture systems, reducing reliance on manual inspection, and providing a foundation for further optimization studies.

Automated Property Valuation with Multi-Hazard Risk: Jakarta Metropolitan Area Study

2025-08-15T02:28:12+00:00

This study crafts a machine learning framework that systematically integrates multi-hazard disaster risk assessments into automated property valuation for the Jakarta Metropolitan Area. The framework addresses 25–30% MAPE typically observed in disaster-prone regions, providing more reliable valuation results. We made 114 prediction features from 42 input variables by using 14,284 property data from Indonesian markets, physical risk data from the Think Hazard platform, and socio-economic data from Central Bureau of Statistics. Elastic Net model performed superior compared to other models which had R² = 0.7922 and a MAPE of 28.27%. We found that some disaster risks had unexpected beneficial effects on property prices. We expected that risks related to the earth (+40.5%) and water (+19.2%) would have positive effects, while risks related to the weather (-66.9%) would have negative effects. These conflicting results suggest that in complex urban markets, the quality of infrastructure, location premiums, and differences in risk perception may outweigh simple risk penalties. The idea gives realistic ideas for property valuation that takes risks into account, but it also points out big problems with how the market judges how likely a disaster is to happen.

Image-Based Classification of Rice Field Conversion: A Comparison Between MLP and SVM Using Multispectral Features

2025-07-31T21:25:53+00:00

The conversion of farmland into non-agricultural purposes has emerged as a pressing concern in many urban regions, including Koto Tangah District, Padang City. In this area, agricultural land experienced a 4% shift in land use between 2022 and 2024. If this trend continues, it could lead to a notable decline in rice production and ultimately threaten food security. This research focuses on examining spatial transformations of rice fields from 2022 to 2024 by utilizing Sentinel-2 satellite imagery along with advanced classification techniques. Vegetation and moisture features were extracted using NDVI, NDWI, texture analysis through GLCM, and Principal Component Analysis (PCA). To classify land cover changes and assess model accuracy, two machine learning approaches were applied: Multilayer Perceptron (MLP) and Support Vector Machine (SVM). The findings reveal a considerable reduction in dense vegetation, indicated by the downward shift of NDVI values in 2024. MLP achieved an accuracy of 82%, outperforming SVM, which reached 71%. Furthermore, MLP obtained a higher F1-score for non-rice field detection (0.75 vs. 0.74) and produced more realistic delineations of rice field boundaries during spatial validation. These outcomes highlight the potential of MLP in monitoring land use conversion, supporting agricultural land conservation, and guiding sustainable urban planning. Moreover, the study contributes to computer science by advancing the use of machine learning for spatio-temporal analysis and reinforcing the role of non-linear models in satellite image classification.

A BiLSTM-Based Approach For Speech Emotion Recognition In Conversational Indonesian Audio using SMOTE

2025-07-31T16:21:48+00:00

Speech Emotion Recognition (SER) identifies human emotions through voice signal analysis, focusing on pitch, intonation, and tempo. This study determines the optimal sampling rate of 48,000 Hz, following the Nyquist-Shannon theorem, ensuring accurate signal reconstruction. Audio features are extracted using Mel-Frequency Cepstral Coefficients (MFCC) to capture frequency and rhythm changes in temporal signals. To address data imbalance, Synthetic Minority Over-sampling Technique (SMOTE) generates synthetic data for the minority class, enabling more balanced model training. A One-vs-All (OvA) approach is applied in emotion classification, constructing separate models for each emotion to enhance detection. The model is trained using Bidirectional Long Short-Term Memory (BiLSTM), capturing contextual information from both directions, improving understanding of complex speech patterns. To optimize the model, Nadam (Nesterov-accelerated Adaptive Moment Estimation) is used to accelerate convergence and stabilize weight updates. Bagging (Bootstrap Aggregating) techniques are implemented to reduce overfitting and improve prediction accuracy. The results show that this combination of techniques achieves 78% accuracy in classifying voice emotions, contributing significantly to improving emotion detection systems, especially for under-resourced languages.

Sentiment Analysis and Topic Modeling for Discovering Knowledge in Indonesian Mobile Government Applications

2025-07-21T08:52:26+00:00

The accelerated rate of government applications development in Indonesia has introduced new opportunities and challenges in delivering digital public services. While thousands of apps have been developed, systemic issues ranging from usability flaws to authentication failures persist, as reflected in user reviews on platforms like the Google Play Store. This study adopts a knowledge discovery approach to extract actionable insights from more than 17,000 user-generated reviews across three major government applications: Satusehat, Digital Korlantas, and M-Paspor. A hybrid methodology is applied, combining RoBERTa-based sentiment classification, BERTopic-based topic modeling, cosine similarity analysis, and qualitative user validation. The findings reveal recurring issues in authentication, interface design, and system responsiveness that span across organizational boundaries. Cross-app topic correlation highlights critical shared pain points such as login failures and unintuitive UI that undermine user trust in e-government services. Mapping these insights onto the SECI knowledge management model, this research contributes both practical recommendations and a replicable analytical framework for public agencies seeking to institutionalize user feedback. By transforming fragmented digital feedback into organizational knowledge, this study supports continuous service improvement and strengthens the foundation for user-centric e-government.

RNN-Based Intrusion Detection System for Internet of Vehicles with IG, PCA, and RF Feature Selection

2025-08-30T12:36:27+00:00

Cyberattacks in the Internet of Vehicles (IoV) threaten road safety and data integrity, requiring intrusion detection systems (IDS) that capture temporal patterns in vehicular traffic. This study develops a Recurrent Neural Network (RNN)-based IDS and evaluates three feature-selection strategies—Information Gain (IG), Principal Component Analysis (PCA), and Random Forest (RF)—on the CICIoV2024 dataset. Features are normalized using Min–Max scaling before being fed into the RNN classifier. The models achieve perfect classification on held-out tests (accuracy/precision/recall/F1 = 1.00). However, probabilistic evaluation reveals low ROC–AUC scores (IG: 0.572, PCA: 0.429, RF: 0.415), indicating limited discriminative margins and potential overfitting or calibration issues despite flawless confusion matrices. PCA and RF further reduce computational overhead during inference compared to IG. These findings highlight that relying solely on accuracy can be misleading for IDS evaluation; temporal RNNs should be complemented with probability-aware training, calibration, or hybrid architectures. This work contributes a temporal-aware IDS framework for IoV and motivates future research on real-time deployment, hybrid RNN-CNN/LSTM models, and adversarial robustness to improve generalization and safety of connected vehicles

Web-Based Diabetes Risk Prediction System Using K-NN on Kaggle Early Stage Diabetes Dataset

2025-08-18T03:58:31+00:00

Diabetes mellitus affects approximately 537 million adults globally, and its rising prevalence poses serious health and economic burdens. Early detection is crucial to reduce risks of complications and improve patient outcomes. This study aims to design and implement a web-based diabetes risk prediction system using the K-Nearest Neighbors (K-NN) algorithm to support early detection based on symptoms. The system utilizes the Kaggle Early Stage Diabetes Risk Prediction Dataset containing 520 records with 17 symptom attributes and one class label. Data preprocessing includes converting categorical data into numerical values, discretizing age into predefined ranges, and applying min-max scaling to normalize feature values. K-NN classification was conducted with K values of 1, 3, and 5, using the PHP Machine Learning (PHP-ML) library and MySQL database integration. The system achieved its highest accuracy of 93.46% at K = 1. Manual testing confirmed that the system processes symptom inputs correctly and provides predictions consistent with training data. This web-based tool offers an accessible platform for early diabetes risk screening, supporting self-assessment and triage. It demonstrates that PHP-ML can effectively implement machine learning in a web environment and can be further enhanced through parameter optimization and integration with larger, more diverse datasets to strengthen generalization.

PROTEGO: Improving Breast Cancer Diagnosis with Prototype-Contrastive Autoencoder and Conformal Prediction on the WDBC Dataset

2025-09-02T09:46:58+00:00

Breast cancer remains one of the leading causes of mortality among women, making accurate and trustworthy early detection a critical challenge in healthcare. To address this, we propose PROTEGO, a Prototype-Contrastive Autoencoder with integrated Conformal Prediction, designed to achieve both high diagnostic accuracy and reliable uncertainty quantification. The framework combines dual-head autoencoding, supervised contrastive learning, prototype-based regularization, and conformal calibration to generate discriminative yet interpretable representations. Using the Wisconsin Diagnostic Breast Cancer (WDBC) dataset, PROTEGO was trained and evaluated through stratified data splits, with performance measured by AUROC, AUPRC, F1-score, Balanced Accuracy, Brier score, calibration error, and conformal coverage metrics. The results show that PROTEGO achieves highly competitive performance with an AUROC of 0.992 and an AUPRC of 0.995, while uniquely providing conformal coverage guarantees with an average set size close to one and more than 92% decisive predictions. Ablation studies confirm the complementary role of each component in enhancing both accuracy and calibration. These findings demonstrate that integrating prototype-guided representation learning with conformal prediction establishes a clinically meaningful diagnostic framework. PROTEGO highlights the importance of unifying precision and reliability in medical AI, offering a step toward more interpretable, safe, and clinically trustworthy systems for breast cancer detection.

Comparison of Transfer Learning Strategies Using MobileNetV2 and ResNet50 for Ecoprint Leaf Classification

2025-08-13T23:07:04+00:00

This research focuses on the classification of leaf types used in ecoprint production through the steaming technique by applying transfer learning on two widely recognized convolutional neural network (CNN) architectures, MobileNetV2 and ResNet50. Leaves have diverse applications in various sectors such as medicine, nutrition, and handicrafts. The study utilized a total of 600 leaf images from 15 species were collected from the surrounding environment and divided into 80% training and 20% testing sets. The aim of this study is to classify leaf types suitable for ecoprint quickly and efficiently, based on transfer learning with two CNN architectures, while incorporating fine-tuning. MobileNetV2 was selected for its computational efficiency, while ResNet50 was chosen for its ability to address the vanishing gradient problem and deliver high accuracy. Fine-tuning was employed to optimize model performance. Experimental results demonstrate that both architectures achieved strong performance, with MobileNetV2 reaching 94.12% accuracy and ResNet50 slightly outperforming it at 94.96%. Confusion matrix evaluation further confirmed these results, yielding accuracy, precision, recall, and F1-score values of 0.94, 0.95, 0.95, and 0.94, respectively. These findings highlight ResNet50’s superior performance over MobileNetV2 while affirming the effectiveness of both models in ecoprint leaf classification.

Air Quality Index Classification: Feature Selection for Improved Accuracy with Multinomial Logistic Regression

2025-09-02T02:27:04+00:00

Air pollution is a major public health concern, creating the need for accurate and interpretable Air Quality Index (AQI) classification models. This study aims to classify AQI into three categories—Good, Moderate, and Unhealthy—using Multinomial Logistic Regression (MLR) with feature selection. The dataset, obtained from public monitoring stations in Jakarta between 2021 and 2024, initially contained 4,620 daily records. After cleaning and outlier removal, 3,586 valid samples remained, from which 900 balanced records (300 per class) were selected for modeling. Key features included PM₁₀, PM₂.₅, SO₂, CO, O₃, and NO₂, which were standardized using Max Normalization to ensure uniform feature scaling. The classification process applied k-fold cross-validation (k = 2–5), and performance was assessed using accuracy and Macro F1-score. Results show that including PM₂.₅ improves performance by about 10%, with the best outcome at k = 5 (accuracy = 91.67%, Macro F1 = 91.45%). These findings confirm PM₂.₅ as a decisive feature for AQI prediction and demonstrate that MLR provides a lightweight, transparent, and computationally efficient solution. Beyond environmental health, the contribution of this work lies in advancing data-driven decision support systems in Informatics, particularly for real-time monitoring and policy applications.

Implementation of Clustering on Packaged Coffee Sales Data for Simulating Goods Entry in Sole Proprietorship Businesses

2025-08-31T08:42:40+00:00

In retail businesses operating under the sole proprietorship structure, decision-making regarding partnerships with beverage distributors—especially those offering packaged coffee—remains a challenge. Store owners often face uncertainty about the profitability of accepting product offerings, which can lead to suboptimal inventory decisions. This study addresses that issue by simulating goods entry scenarios and applying clustering techniques to historical packaged coffee sales data, enabling data-driven insights into product performance and distributor value. Studies focusing on clustering within retail include segmenting customer behavior and stock management strategies, yet many lacked specific application to single owner businesses and product-centric simulations. This research is novel in its contextual focus on packaged coffee distribution within sole proprietorship environments, integrating real sales metrics and clustering algorithms to empower store owners with actionable evaluation tools. Results demonstrate that clustering reveals patterns of profitable product categories and distributor consistency, offering scalable insights for micro-retail optimization. The findings provide a framework that differs from prior studies by emphasizing the intersection between small business dynamics and algorithmic decision support.

IT Governance through Mathematical Modeling: A Quantitative Assessment of Maturity Using Factor Analysis and Structural Equation Modeling

2025-07-31T16:17:38+00:00

IT Governance (ITG) ensures an organization's technological capabilities align with its business strategy. Although frameworks like COBIT 2019 offer structured guidelines, many assessment techniques rely on qualitative measures, which can compromise objectivity. This paper proposes a novel quantitative approach that integrates Factor Analysis (FA) and Structural Equation Modeling (SEM) to measure IT Governance maturity. By mapping each COBIT 2019 domain—EDM, APO, BAI, DSS, and MEA—onto a latent construct, organizations gain empirical insights into their governance status. Exploratory and confirmatory factor analyses validate these domains, while SEM reveals the magnitude and significance of each domain's impact on overall IT Governance maturity. A real-world example from a financial services company, "FinServEU," demonstrates how this framework can prioritize improvements, enhance regulatory compliance, and promote continuous monitoring. The results highlight that quantitative ITG modeling provides a reliable basis for informed decision-making and optimal resource allocation, bridging the gap between broad qualitative assessments and actionable strategies. This approach is crucial for the field of informatics and computer science, as it offers a robust, reproducible, and objective framework for evaluating a key aspect of digital transformation, ensuring that technological progress is guided by sound, data-driven principles.

Stacked Random Forest-LightGBM for Web Attack Classification

2025-08-11T07:17:50+00:00

The rapid expansion of web services in the digital era has intensified exposure to increasingly complex and imbalanced cyber threats. This study proposes a stacking hybrid ensemble framework for web attack classification, integrating Random Forest as the base learner and LightGBM as the meta-learner, enhanced by the SMOTE technique for data balancing. The Web Attack subset of the CICIDS-2017 dataset serves as a case study, with a focus on detecting minority attacks such as SQL Injection, XSS, and Brute Force. The preprocessing pipeline includes data cleaning, removal of irrelevant features, normalization, extreme value imputation, and ANOVA F-test-based feature selection. Evaluation results indicate that the proposed model outperforms baseline models in both multiclass classification (98.7% accuracy, 0.634 macro F1-score) and binary classification (99.41% accuracy, 99.47% F1-score), while maintaining high sensitivity to minority classes. These results contribute to informatics and cybersecurity scholarship through a generalizable stacking baseline and well-specified evaluation procedures for web-attack detection, facilitating replicability, fair comparison, and dataset-agnostic insights.

Comparison of IndoNanoT5 and IndoGPT for Advancing Indonesian Text Formalization in Low-Resource Settings

2025-07-16T00:01:03+00:00

The rapid growth of digital communication in Indonesia has led to a distinct informal linguistic style that poses significant challenges for Natural Language Processing (NLP) systems trained on formal text. This discrepancy often degrades the performance of downstream tasks like machine translation and sentiment analysis. This study aims to provide the first systematic comparison of IndoNanoT5 (encoder-decoder) and IndoGPT (decoder-only) architectures for Indonesian informal-to-formal text style transfer. We conduct comprehensive experiments using the STIF-INDONESIA dataset through rigorous hyperparameter optimization, multiple evaluation metrics, and statistical significance testing. The results demonstrate clear superiority of the encoder-decoder architecture, with IndoNanoT5-base achieving a peak BLEU score of 55.99, significantly outperforming IndoGPT's highest score of 51.13 by 4.86 points—a statistically significant improvement (p<0.001) with large effect size (Cohen's d = 0.847). This establishes new performance benchmarks with 28.49 BLEU points improvement over previous methods, representing a 103.6% relative gain. Architectural analysis reveals that bidirectional context processing, explicit input-output separation, and cross-attention mechanisms provide critical advantages for handling Indonesian morphological complexity. Computational efficiency analysis shows important trade-offs between inference speed and output quality. This research advances Indonesian text normalization capabilities and provides empirical evidence for architectural selection in sequence-to-sequence tasks for morphologically rich, low-resource languages.

Digital Landscape and Behavior in Indonesia 2024: A National Survey Analysis of Internet Penetration, Cybersecurity Risks, and User Segmentation Using K-Means Clustering and Logistic Regression

2025-07-31T21:23:36+00:00

Digital transformation in Indonesia reveals significant disparities in internet access, digital behavior, and cybersecurity vulnerabilities. This study analyzes the digital landscape using national survey data from 8,720 respondents across 38 provinces. This research employs a quantitative approach, utilizing chi-square tests, logistic regression for risk analysis, and K-Means clustering for user segmentation, supported by Principal Component Analysis (PCA) for dimensionality reduction. The results show a national internet penetration rate of 79.5%, with significant disparities across regions and socio-economic segments. Logistic regression analysis reveals that higher education, greater income, and the use of fixed broadband are negatively correlated with cybersecurity risks. Furthermore, K-Means clustering identifies three distinct user profiles: 'Digital Savvy', 'Pragmatic Users', and the 'Vulnerable Segment', each with unique characteristics regarding digital access and literacy. This research provides a critical empirical basis for understanding digital transformation in a developing nation. The findings underscore the necessity of data-driven, segmented policies to foster digital inclusion and enhance national cybersecurity, offering actionable insights for policymakers and service providers.

Hybrid Model for Speech Emotion Recognition using Mel-Frequency Cepstral Coefficients and Machine Learning Algorithms

2025-09-02T03:00:20+00:00

Speech Emotion Recognition (SER) is a subfield of affective computing that focuses on identifying human emotions through voice signals. Accurate emotion classification is essential for developing intelligent systems capable of interacting naturally with users. However, challenges such as background noise, overlapping emotional features, and speaker variability often reduce model performance. This study aims to develop a lightweight hybrid SER model by combining Mel-Frequency Cepstral Coefficients (MFCC) as feature representations with three machine learning algorithms: Support Vector Machine (SVM), Decision Tree (DT), and K-Nearest Neighbors (KNN). The methodology involves audio data preprocessing, MFCC-based feature extraction, and classification using the selected algorithms. The RAVDESS dataset, consisting of 1,440 English-language audio samples across four emotions (happy, angry, sad, neutral), was used with an 80/20 train-test split to ensure class balance.. Experimental results show that the KNN model achieved the highest performance, with an accuracy of 78.26%, precision of 85.09%, recall of 78.26%, and F1-score of 77.06%. The Decision Tree model produced comparable results, while the SVM model performed poorly across all metrics. These findings demonstrate that the proposed hybrid approach is effective for recognizing emotions in speech and offers a computationally efficient alternative to deep learning models. The integration of MFCC features with multiple machine learning classifiers provides a robust framework for real-time emotion recognition applications, especially in environments with limited computing resources.

Automated Video Recognition of Traditional Indonesian Dance Using Hyperparameter-Tuned Convolutional Neural Network

2025-07-30T23:16:36+00:00

Traditional Indonesian dances serve as a vital expression of cultural identity and regional heritage, yet their preservation through intelligent video recognition remains limited due to technical challenges in motion complexity, costume variation, and the lack of annotated datasets. Prior research commonly employed Convolutional Neural Networks (CNNs) with manually defined hyperparameters, which often resulted in overfitting and poor adaptability when applied to dynamic and real-world video inputs. To overcome these limitations, this study proposes a robust and adaptive classification framework utilizing a hyperparameter-tuned CNN model. The approach automatically optimizes key training parameters such as learning rate, batch size, optimizer type, and epoch count through iterative experimentation, thereby maximizing the model’s ability to generalize across both static and temporal data domains. The model was trained using image datasets representing three traditional dances (Gambyong, Remo, and Topeng), and subsequently tested on segmented frames extracted from YouTube videos. Results indicate strong model performance, achieving 99.67% accuracy on the training set and 100% accuracy, precision, recall, and F1-score across all testing videos. The proposed method successfully bridges the gap between still-image learning and real-world motion recognition, making it suitable for practical applications in digital archiving and cultural documentation. This study’s contribution lies not only in the model’s technical effectiveness but also in its support for preserving intangible cultural assets through intelligent and automated video-based recognition. Future work may incorporate temporal modelling or multi-camera perspectives to further enrich motion understanding and extend the system to broader performance domains.

Predicting Smartphone Addiction Levels with K-Nearest Neighbors Using User Behavior Patterns

2025-06-19T06:29:19+00:00

Smartphones have become an integral part of everyday life, but their ever-increasing popularity has raised growing global concerns about excessive use (nomophobia), which impacts quality of life, mental health, and academic performance. Existing research often relies on subjective questionnaires, limiting scalability and objectivity. This study addresses this gap by developing a machine learning model to predict smartphone addiction levels through an objective analysis of user behavior patterns. This research evaluates the effectiveness of the K-Nearest Neighbor (KNN) algorithm, identifies the most influential behavioral features, and assesses the model's classification performance. Using a dataset of 3,300 user behavior entries with 11 features, a waterfall-based framework was employed for data preprocessing, model design, and evaluation. The KNN model achieved 95% accuracy in classifying addiction levels. Permutation Feature Importance analysis confirmed ‘App Usage Time’ and ‘Battery Drain’ as the two most influential predictive features. This study demonstrates that KNN is a powerful and viable method for objectively classifying smartphone addiction. The findings provide a strong foundation for developing scalable, AI-driven early detection and intervention systems, offering significant contributions to the fields of computer science and digital well-being.

Comparative Evaluation of Decision Tree and Random Forest for Lung Cancer Prediction Based on Computational Efficiency and Predictive Accuracy

2025-06-16T02:23:54+00:00

Early detection of lung cancer is essential for improving treatment outcomes and patient survival rates. This paper presents a comparative evaluation of two classification algorithms: Decision Tree and Random Forest, focusing on both predictive performance and computational efficiency. The models were tested using 10-fold cross-validation to ensure robustness. Both algorithms achieved the same accuracy of 93.3%. However, Random Forest slightly outperformed Decision Tree in recall (88.8% vs. 87.9%), F1-score (92.2% vs. 92.1%), and AUC (0.94 vs. 0.91), while Decision Tree obtained higher precision (97% vs. 95.9%). In terms of computational efficiency, Decision Tree demonstrated faster training and testing times, lower memory usage, and reduced energy consumption compared to Random Forest. The results reveal a clear trade-off between prediction quality and resource usage, highlighting the importance of selecting algorithms not only for their accuracy but also for their practicality in real-world healthcare scenarios. This comprehensive evaluation provides valuable insights for developing intelligent decision support systems that are both effective and resource-efficient, especially in environments with limited computing capacity. These findings contribute to the advancement of resource-aware intelligent systems in the field of medical informatics.

Implementation of Extra Trees Classifier and Chi-Square Feature Selection for Early Detection of Liver Disease

2025-02-19T06:51:22+00:00

The imbalanced distribution of medical data poses challenges in accurately detecting liver disease, which is crucial as symptoms often remain unnoticed until advanced stages. This study examines the application of the Extra Trees Classifier algorithm and chi-square feature selection for early detection of liver disease. Compared to traditional methods like Random Forest and SVM, the Extra Trees Classifier offers enhanced computational efficiency and better handling of imbalanced datasets, while chi-square feature selection helps identify the most relevant medical indicators. The data consists of five medical variables likely to be laboratory test results from patient samples, with labels indicating classes A and B. The data is randomly divided with a ratio of 80% for each class. To address data imbalance, SMOTE technique was applied before the data was randomly split into a ratio of 80% for training and 20% for testing to ensure effective learning and testing of the model's performance. The results showed that with the help of chi-square feature selection, the Extra Trees Classifier algorithm could provide fairly accurate predictions in liver disease classification, with an accuracy of 82.6%, sensitivity of 85.5%, precision of 78.3%, and F1-Score of 81.7%. These results demonstrate significant improvement over existing methods, and the proposed approach can aid healthcare practitioners in making timely diagnostic decisions, potentially reducing mortality rates through early intervention in liver disease cases.

Comparing BERTBase, DistilBERT and RoBERTa in Sentiment Analysis for Disaster Response

2025-06-18T12:46:23+00:00

Social media platforms are vital for real-time communication during disasters, providing insights into public emotions and urgent needs. This study evaluates the performance of three transformer-based models—BERTBase, DistilBERT, and RoBERTa—for sentiment analysis on disaster-related social media data. Using a multilingual dataset sourced from the Social Media for Disaster Risk Management (SMDRM) platform, the models were assessed on classification metrics including accuracy, precision, recall, and weighted F1-score. The results show that RoBERTa consistently outperforms the others in classification performance, while DistilBERT offers superior computational efficiency. The analysis highlights the trade-offs between model accuracy and runtime, emphasizing RoBERTa's suitability for scenarios prioritizing accuracy, and DistilBERT's potential in time-sensitive or resource-constrained applications. These findings support the integration of sentiment analysis into disaster response systems to enhance situational awareness and decision-making.

Hyperparameter Optimization Of IndoBERT Using Grid Search, Random Search, And Bayesian Optimization In Sentiment Analysis Of E-Government Application Reviews

2025-07-02T14:07:57+00:00

User reviews on Google Play Store reflect satisfaction and expectations regarding digital services, including E-Government applications. This study aims to optimize IndoBERT performance in sentiment classification through fine-tuning and hyperparameter exploration using three methods: Grid Search, Random Search, and Bayesian Optimization. Experiments were conducted on Sinaga Mobile app reviews, evaluated using accuracy, precision, recall, F1-score, learning curve, and confusion matrix. The results show that Grid Search with a learning rate of 5e-5 and a batch size of 16 provides the best results, with an accuracy of 90.55%, precision of 91.16%, recall of 90.55%, and F1-score of 89.75%. The learning curve indicates stable training without overfitting. This study provides practical contributions as a guide for improving IndoBERT in Indonesian sentiment analysis and as a foundation for developing NLP-based review monitoring systems to enhance public digital services.

Brain Tumor Segmentation From MRI Images Using MLU-Net with Residual Connections

2025-05-28T02:36:58+00:00

Brain tumor segmentation plays an important role in medical imaging in assisting diagnosis and treatment planning. Although advances in deep learning such as Unet already perform image segmentation, many challenges exist in segmenting brain tumors with tumor spread boundaries. This paper proposes a model that combines CNN and MLP (MLU-Net) techniques enhanced by the addition of residual connections to improve segmentation accuracy called ResMLU-Net. This architecture combines 2D covolution layers, block MLP and residual connections to process MRI images with the dataset used is BraTS 2021. The residaul connection helps reduce gradient degradation which ensures smooth information flow and better feature learning. The performance of ResMLU-Net will be evaluated using Dice and IoU metrics and will also be compared with several models such as Unet, ResUnet and MLU-Net. The experimental scores obtained from ResMLU-Net for segmenting brain tumors are 83.43% for IoU and 89.94% for Dice. These results show that adding residual connections can improve the accuracy in segmenting brain tumors which can be seen that there is an increase in the Dice and Iou scores. The proposed ResMLU-Net model is a valuable contribution to medical imaging and health informatics. With its provision of a standard and computationally viable solution to brain tumor segmentation, it offers incorporation into Computer-Aided Diagnosis (CAD) systems and support to clinical decision-making protocols.

Prediction of Life Expectancy of Lung Cancer Patients After Thoracic Surgery Using Decision Tree Algorithm and Adaptive Synthetic Sampling

2025-05-26T04:14:45+00:00

This research focuses on predicting the life expectancy of lung cancer patients after undergoing thoracic surgery, using a decision tree classification algorithm (C4.5) combined with adaptive synthetic sampling to handle data imbalance. Data imbalance in the lung cancer patient dataset is a major obstacle in obtaining accurate prediction results, especially in identifying minority classes. Data imbalance in the lung cancer patient dataset is a major obstacle in obtaining accurate prediction results, especially in identifying minority classes. By applying ADASYN, the data distribution becomes more even, thus improving the performance of the C4.5 model. The results showed that combining these methods increased the prediction accuracy from 67% to 87%. In addition, the precision, recall, and f1-score for minority classes have significantly improved, which were previously difficult to identify by the model. Thus, combining the C4.5 algorithm and the ADASYN technique proved effective in dealing with the challenge of data imbalance and resulted in better prediction in the case of lung cancer. This study is expected to contribute to the field of medical classification and serve as a reference for further research on similar cases.

Analysis of Technology Adoption Factors in Learning among Vocational Students using UTAUT2 Model

2025-06-30T11:33:24+00:00

Technology acceptance in vocational education is a key factor in supporting the effectiveness of teaching and learning processes in the digital era. This study aims to analyze the factors influencing technology acceptance among students of the Computer and Network Engineering (TKJ) Department at SMK Ma'arif 1 Kroya using the Unified Theory of Acceptance and Use of Technology 2 (UTAUT2) framework. The model includes the variables Performance Expectancy, Effort Expectancy, Social Influence, Facilitating Conditions, Hedonic Motivation, Price Value, Habit, Behavioral Intention, and Actual Usage. The results reveal that five key variables—Performance Expectancy, Effort Expectancy, Social Influence, Hedonic Motivation, and Price Value—significantly influence Behavioral Intention, while Habit, Facilitating Conditions, and Behavioral Intention directly affect Actual Usage. All constructs in the model meet validity and reliability criteria, and no multicollinearity was detected (VIF < 3.3). The coefficient of determination (R²) values of 0.612 for Behavioral Intention and 0.673 for Actual Usage indicate strong predictive power of the model. These findings confirm the relevance of the UTAUT2 framework for understanding and enhancing technology acceptance in vocational education settings and provide valuable insights for improving technology integration in technical learning environments.

Comparative Analysis of LSTM and GRU for River Water Level Prediction

2025-07-09T06:19:42+00:00

Accurate river water level prediction is essential for flood management, especially in tropical areas like Palembang. This study systematically analyzes the performance of two deep learning models, Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU), for real-time water level forecasting using hourly rainfall and water level data collected from automatic sensors. A series of experiments were conducted by varying window sizes (10, 20, 30) and the number of layers (1, 2, 3) for both models, with model performance assessed using RMSE, MAE, MAPE, and NSE. The results demonstrate that both window size and network depth significantly influence prediction accuracy and computational efficiency. The LSTM model achieved its highest accuracy with a window size of 30 and a single layer, while the GRU model performed best with a window size of 20 and two layers. This work contributes by systematically analyzing hyperparameter configurations of LSTM and GRU models on hourly rainfall and water level time series for flood-prone regions, offering empirical insight into parameter tuning in recurrent neural architectures for hydrological forecasting. These findings highlight the importance of careful parameter selection in developing reliable early warning systems for flood risk management.

Bayesian Optimized Pretrained CNNs for Mango Leaf Disease Classification: A Comparative Study

2025-07-23T08:17:53+00:00

Mango leaf diseases pose a major threat to crop productivity, causing significant economic losses for farmers. Accurate and early detection is essential, yet manual diagnosis remains subjective and inefficient. This study aims to evaluate and compare the performance of five pretrained Convolutional Neural Network (CNN) architectures—DenseNet121, ResNet50V2, MobileNetV3 Small, MobileNetV3 Large, and InceptionV3—by systematically optimizing their hyperparameters to identify the most effective model for mango leaf disease classification. The public MangoLeafBD dataset, containing 4,000 images from eight balanced classes, was used. Bayesian Optimization was applied to fine-tune each model, and their performances were assessed before and after optimization. Results show that optimization substantially improved all models, with MobileNetV3 Large achieving the highest accuracy of 100% on the test set, followed by DenseNet121 (99.75%), ResNet50V2 (99.63%), MobileNetV3 Small (99.50%), and InceptionV3 (98.50%). The findings highlight that a well-tuned lightweight model can outperform more complex architectures, offering a practical and efficient solution for developing mobile-based diagnostic tools to support precision agriculture in resource-constrained settings.

Enhancing Customer Purchase Behavior Prediction Using PSO-Tuned Ensemble Machine Learning Models

2025-07-01T09:08:55+00:00

Predicting customer purchase behavior remains a significant challenge in e-commerce and marketing analytics due to its complex and nonlinear patterns. This study introduces a machine learning framework that integrates ensemble learning models with Particle Swarm Optimization (PSO) for hyperparameter tuning to improve classification accuracy and class discrimination. Several ensemble algorithms, including CatBoost, XGBoost, LightGBM, AdaBoost, and Gradient Boosting, were compared against a baseline Logistic Regression model, both with default and PSO-optimized configurations. Experiments on a real-world e-commerce dataset containing behavioral and demographic variables showed that ensemble methods substantially outperformed traditional models across accuracy, F1-score, and ROC AUC metrics. Notably, the PSO-tuned Gradient Boosting model achieved the highest ROC AUC of 0.9547, improving the AUC by approximately 0.0076 compared to its default configuration, while CatBoost obtained the highest overall accuracy and F1-score. PSO optimization was especially effective in enhancing simpler models such as Logistic Regression but showed marginal gains and some convergence instability in more complex ensemble models. Feature importance analyses consistently identified variables such as time spent on the website, discounts availed, age, and income as key drivers of purchase intent. These findings demonstrate the benefit of combining ensemble learning with metaheuristic optimization, offering actionable insights for developing robust, data-driven marketing strategies.