Jurnal Teknik Informatika (Jutif)

ANALYSIS OF THE EFFECTIVENESS OF POLYNOMIAL FIT SMOTE MESH ON IMBALANCE DATASET FOR BANK CUSTOMER CHURN PREDICTION WITH XGBOOST AND BAYESIAN OPTIMIZATION

2024-05-24T03:23:56+00:00

The case of churn in the banking industry, namely customers who leave or no longer use bank services, is a serious problem that requires an appropriate solution. The aim of this research is to predict churn and take appropriate preventive actions using machine learning. The dataset contains 10,000 bank customer data with 14 relevant features. Only about 20% of customers experience churn, creating a data imbalance problem in classification. To overcome data imbalances, the SMOTE oversampling technique was applied. Also introduced was the development of the SMOTE technique, namely, Polynomial Fit SMOTE Mesh (PFSM). PFSM works by combining each point in the data with a linear function and producing synthetic data at each connected distance. Experimental results show that the model developed using PFSM and optimized with Bayesian Optimization for the XGBoost algorithm achieved 86.1% accuracy, 70.87% precision, 53.81% recall, and 61.17% F-score. This indicates that the approach is successful in improving predictive capabilities and identifying potential customers for churn earlier. This research has significant relevance in the banking industry, helping banks to safeguard their customers and improve banking business performance..

COMPARISON PERFORMANCE OF WORD2VEC, GLOVE, FASTTEXT USING SUPPORT VECTOR MACHINE METHOD FOR SENTIMENT ANALYSIS

2024-05-24T03:24:18+00:00

Spotify is a digital audio service that provides music and podcasts. Reviews received by the application can affect users who will download the application. The unstructured characteristic of review text is a challenge in text processing. To produce a valid sentiment analysis, word embedding is required. The data set that is owned is divided by a ratio of 80:20 for training data and testing data. The method used for feature expansion is Word2Vec, GloVe, and FastText and the method used in classification is Support Vector Machine (SVM). The three word embedding methods were chosen because they can capture semantic, syntactic, and contextual meanings around words when compared to traditional engineering features such as Bag of Word. The best performance evaluation results show that the GloVe model produces the best performance compared to other word embeddings with an accuracy value of 85%, a precision value of 90%, a recall value of 79%, and an f1-score of 85%.

IMPLEMENTATION OF RSA AND AES-128 SUPER ENCRYPTION ON QR-CODE BASED DIGITAL SIGNATURE SCHEMES FOR DOCUMENT LEGALIZATION

2024-05-24T03:24:38+00:00

Maintaining the confidentiality and integrity of electronic documents is essential in the modern digital age. In the contemporary digital world, digital signatures are essential for safeguarding and legalizing electronic documents. The current issue, however, goes beyond digital signatures and instead centers on enhancing security and data integrity. Therefore, RSA and AES-128 super-encryption is required in QR-code-based digital signature techniques for document legalization. This research stage entails constructing a super encryption algorithm, testing it experimentally for security and performance, and designing a digital signature system using RSA and AES-128 super encryption. The results of this research show that the use of RSA and AES super encryption has been proven to have better performance in data security, where the encryption and decryption process time is relatively close to the RSA encryption time, and the comparison of entropy values is better than RSA and AES-128. So, the combination of Super RSA and AES-128 encryption can increase the security level of electronic documents and reduce the risk of hacking. Moreover, the proposed QR-code-based digital signature scheme is also very efficient regarding file size and processing time.

OPTIMIZING BUTTERFLY CLASSIFICATION THROUGH TRANSFER LEARNING: FINE-TUNING APPROACH WITH NASNETMOBILE AND MOBILENETV2

2024-05-24T03:24:57+00:00

Butterflies play a significant role in ecosystems, especially as indicators of the state of biological balance. Each butterfly species is distinctly different, although some also show differences with very subtle traits. Etymologists recognize butterfly species through manual taxonomy and image analysis, which is time-consuming and costly. Previous research has tried to use computer vision technology, but it has shortcomings because it uses a small distribution of data, resulting in a lack of programs for recognizing various other types of butterflies. Therefore, this research is made to apply computer vision technology with the application of transfer learning, which can improve pattern recognition on image data without the need to start the training process from scratch. Transfer learning has a main method, which is fine-tuning. Fine-tuning is the process of matching parameter values that match the architecture and freezing certain layers of the architecture. The use of this fine-tuning process causes a significant increase in accuracy. The difference in accuracy results can be seen before and after using the fine-tuning process. Thus, this research focuses on using two Convolutional Neural Network architectures, namely MobileNetV2 and NASNetMobile. Both architectures have satisfactory accuracy in classifying 75 butterfly species by applying the transfer learning method. The results achieved on both architectures using fine-tuning can produce an accuracy of 86% for MobileNetV2, while NASNetMobile has a slight difference in accuracy of 85%.

IMPLEMENTATION OF NATURAL LANGUAGE PROCESSING (NLP) IN CONSUMER SENTIMENT ANALYSIS OF PRODUCT COMMENTS ON THE MARKETPLACE

2024-05-24T03:25:11+00:00

Market product reviews are invaluable information if processed carefully. The process of analyzing product reviews is more than just considering star ratings; Comprehensive examination of the overall content of review comments is essential to extracting the nuances of meaning conveyed by the reviewer. The problem currently occurring in analyzing reviews of product purchases in the marketplace is the large number of abbreviations and non-standard language used by commenters, making it difficult for the system to understand. Therefore, a Natural Language Processing (NLP) approach is needed to improve the language in the content of review comments so as to achieve maximum performance in sentiment analysis. This research utilizes the KNN and TF-IDF algorithms, coupled with NLP techniques, to categorize Muslim fashion product reviews into two different groups that is positive and negative. The NLP-enhanced classification achieved 76.92% accuracy, 80.00% precision, and 74.07% recall, surpassing the results obtained without NLP, which had 69.23% accuracy, 80.00% precision, and 64.52 recall. %. Frequently appearing words in reviews serve as a description of collective buyer sentiment regarding the product. Positive reviews indicate customer satisfaction with the quality, speed of delivery, and price of the goods, while negative reviews indicate dissatisfaction with factors such as color differences and differences in the number of items received.

NAIVE BAYES AND PARTICLE SWARM OPTIMIZATION IN EARLY DETECTION OF CHRONIC KIDNEY DISEASE

2024-05-27T05:24:26+00:00

Chronic Kidney Disease (CKD) is a global health problem that requires early detection to reduce the risk of complications and disease progression. The Naïve Bayes (NB) algorithm has been proven effective in detecting CKD but its accuracy still varies. The problem with previous research is that it has not fully optimized existing algorithms in terms of accuracy and efficiency. This research aims to develop a more accurate and efficient early detection method for CKD using the NB algorithm and Particle Swarm Optimization (PSO). The NB method is known for its speed and ease of implementation, with global search capabilities and PSO for parameter optimization. Dataset from the UCI repository, which includes data pre-processing, NB implementation, performance evaluation, and enhancement with PSO. The results of NB+PSO show a significant increase in accuracy of 95.75% from 95.00% and Area Under Curve (AUC) value of 0.910% from 0.802% compared to the use of NB alone. The conclusion of this study is that the combination of NB+PSO increases effectiveness in early detection of CKD. This research opens up opportunities for further development in the medical field, especially in improving the diagnostic accuracy of other diseases.

SENTIMENT ANALYSIS OF ICT SERVICE USER USING NAIVE BAYES CLASSIFIER AND SVM METHODS WITH TF-IDF TEXT WEIGHTING

2024-05-27T05:27:28+00:00

Pusintek is one of the government units in Indonesia responsible for managing Information and Communication Technology (ICT), providing various ICT services to users in central and regional offices through the ICT Service Catalog. The level of service fulfillment in Pusintek's IT Service Catalog significantly influences the effectiveness and efficiency in meeting service agreements, providing accurate information, and handling disruptions promptly. User satisfaction is measured through surveys to plan improvements to ICT services, but there is currently no method to classify sentiment from survey comment data. This research aims to classify sentiment and understand customer opinions and satisfaction trends regarding ICT services. The study applies the Naïve Bayes Classifier and Support Vector Machine (SVM) methods to classify positive and negative comments in user satisfaction surveys of ICT services. The data used consists of comments from the 2022 ICT user satisfaction survey results. Based on the test results, it is observed that the SVM algorithm provides higher accuracy compared to the Naïve Bayes algorithm. Utilizing the existing dataset with established opinion values, classification modeling using Naïve Bayes Classifier and Support Vector Machine (SVM) proves capable of classifying ICT user sentiment into 3 sentiment classes: Positive, Neutral, and Negative. From the data above, it is concluded that the SVM algorithm achieves the highest accuracy of 88.76%, highest precision of 89.68%, recall of 88.76%, and an f1-score of 89.12%.

PERFORMANCE TESTING OF ACADEMIC WEBSITE USING LOAD TESTING METHOD SUPPORTED BY APACHE JMETERTM AT XYZ UNIVERSITY

2024-05-27T05:29:13+00:00

This study aims to evaluate the performance quality of the academic website of University XYZ through load testing using the Load Testing Method supported by Apache JMeterTM. The main issue addressed is how the website's performance can be measured and assessed in the context of normal, peak, and stress usage. The research methodology involves a qualitative approach to understand the meaning, interpretation, and context of the phenomenon, coupled with a quantitative approach to measure, analyze, and organize data in numerical or statistical forms. The research findings indicate that in the basic testing scenario, the website successfully met the test criteria with an average response time of approximately 0.855 seconds for GET requests, below the established maximum limit. POST requests required an average time of around 0.273 seconds with no response failures. In the peak testing scenario simulating high traffic conditions, the website remained optimal with average response times for both GET and POST requests below the 3-second limit, without response failures. Stress testing scenarios demonstrated the efficient operation of the website, even though the average response time for GET requests increased to approximately 2.564 seconds.The test results affirm that the University XYZ website functions well under various service usage conditions, including heavy loads. The overall average response time for GET requests across all scenarios is approximately 1.558 seconds, while POST requests have an average response time of around 0.355 seconds. Special attention is given to the impact of the number of threads or users and the number of students on the website's performance.

COMPARISON OF NAÏVE BAYES ALGORITHM AND SUPPORT VECTOR MACHINE IN SENTIMENT ANALYSIS OF BOYCOTT ISRAELI PRODUCTS ON TWITTER

2024-05-27T05:31:23+00:00

The Israeli-Palestinian conflict has captured the attention of Indonesians and even the world for decades, with the death toll reaching 17,000 Palestinians. Indonesians have expressed various opinions, including a proposed boycott of products that allegedly support Israel as a form of protest against the ongoing conflict. This study explores the opinions and sentiments of the Indonesian people regarding the Israel-Palestine conflict and the efforts to boycott Israeli products on social media twitter. This study aims to compare the accuracy of the two algorithms in classifying sentiment towards boycotting Israeli products. A total of 2288 comment data were processed using the Naïve Bayes and Support Vector Machine (SVM) algorithm classification methods. The results show that the Naïve Bayes algorithm has higher accuracy with a data division ratio of 70:30 and 80:30 for training data and testing data. Accuracy results with 70:30 data division reached 84% using the Naïve Bayes algorithm model, while the SVM algorithm model reached 78%. And the accuracy results with 80:20 data division reached 85% using the Naïve Bayes algorithm model, with the SVM algorithm model reaching 84%. This study provides an understanding of the concept of text mining and data mining and can be a reference for similar research.

DECISION TREE OPTIMIZATION IN HEART FAILURE DIAGNOSTICS: A PARTICLE SWARM OPTIMIZATION APPROACH

2024-05-27T05:33:31+00:00

The rapid advancement of technology has made the implementation of accurate diagnostic methods for serious diseases like heart failure extremely important. Heart failure, being a leading cause of death worldwide, necessitates precise and accurate diagnostic techniques. The problem with conventional diagnostic methods is that they often fail to effectively accommodate the complexity of clinical data, leading to an increase in mortality rates due to heart failure. Previous research has employed various data analysis methods, but there are still fluctuations in the accuracy of results. The aim of this study is to enhance the accuracy of heart failure diagnosis by integrating the Decision Tree (DT) method with Particle Swarm Optimization (PSO) optimization. This research involves collecting and preprocessing heart failure data, followed by the development of a DT model. This model is then optimized using the PSO technique. The study uses a dataset from the UCI Repository, involving testing and validation processes to measure the model's effectiveness. The results show a significant improvement in accuracy and the Area Under Curve (AUC) after applying PSO. Accuracy increased from 79.92% to 85.29%, and AUC from 0.706% to 0.794%. The conclusion is that the integration of DT and PSO successfully improved the accuracy and reliability of the model in diagnosing heart failure. This innovation offers potential for further research in integrating optimization techniques in health data analysis, with the possibility of application in various clinical scenarios.

COMPARISON OF JENKINS AND GITLAB CI/CD TO IMPROVE DELIVERY TIME OF BASU DAIRY FARM ADMIN WEBSITE

2024-05-28T02:17:51+00:00

The Basu Dairy Farm admin website is a web-based information system developed using monolithic architecture. The delivery process of source code changes from the GitLab repository on the "main" branch (development) to the main server (production) takes a long time because the build and deploy process is done manually. This causes the delivery time to be long. To overcome this, this research applies Continuous Integration/Continuous Deployment (CI/CD) as a solution. The CI/CD tools used are Jenkins and GitLab CI/CD because they are open source and the most popular. In this study, a comparison of the delivery time of the two tools was carried out. Delivery time is obtained when the build process starts to run until the deploy process is completed. The analysis includes the time required to run the build and deploy process of the CI/CD tool. The results of this research show that Jenkins and GitLab CI/CD are successfully implemented and can automate the build and deploy process. In terms of implementation, Jenkins requires in-depth configuration, so it looks complicated, while GitLab CI/CD offers simple and easy configuration. In the three experiments conducted, Jenkins showed a faster average time in completing the build and deploy process, so Jenkins has a better delivery time than GitLab CI/CD in the context of the Basu Dairy Farm admin website development process.

REUSE OF THE EFSM MODEL OF PEDULILINDUNGI APPLICATION IN SATUSEHAT APPLICATION TESTING WITH MBT METHOD

2024-05-28T03:20:39+00:00

On 2023, the Government of Indonesia announced the change of PeduliLindungi application to SatuSehat, with the addition of features that have been integrated with Electronic Medical Records (RME). In this research, the concept of model reuse is applied to facilitate the creation of test models on the same features between PeduliLindungi and SatuSehat, namely Linked Profile and Covid-19 Vaccine. In applying the reuse model, the method template and edge template strategies are used to adjust to the evolution of the model that occurs in the SATUSEHAT application, in the edge template or second iteration there are additional vertices and edges on the Linked Profile and Vaccine features. By combining the number of vertices and edges, the overall similarity percentage is around 79.81% on the Linked Profile feature, showing the efficiency of modeling with a reuse model of around 20.19%. Testing on SatuSehat using Altwalker tools with Random and Weighted Random algorithms shows high coverage achievements, especially on vertex, these achievements show the effectiveness of the reuse model. Comparison with previous research on PeduliLindungi shows an increase in coverage rate, especially on features that apply the reuse model. This research illustrates the success of the reuse model concept in accelerating the development of test models and increasing coverage in applications where changes occur.

APPLICATION OF ENSEMBLE METHOD FOR EMPLOYEE TURNOVER PREDICTIONS IN FINANCIAL SERVICES COMPANY

2024-05-28T04:00:23+00:00

High employee turnover is a challenge for every company, considering that employees are a valuable asset for the company. A high employee turnover rate indicates the high frequency of employees leaving a company. This will harm the company in terms of time, costs, human resources, and reduce the company's reputation. Low employee turnover is an objective for every company in its efforts to achieve its vision and mission, the employee turnover rate is high at 78.97% at PT. HCI operating in the financial services sector can have a negative impact on the company's reputation. Therefore, there is a need to analyze and predict employee turnover so that company management can take preventive and persuasive actions so as to reduce employee turnover rates. Therefore, a tool is needed to predict whether an employee will leave the company. This paper aims to predict the possibility of employees out of the company using the ensemble method, which is a method that uses a combination of several algorithms consisting of base learners and individual learners, algorithms with the ensemble method used are stacking, random forest, and adaboost, then comparing the result to get the best accuracy. The test results prove that the Stacking algorithm technique is the best model with the highest score in terms of accuracy with a value of 86.84%, while the Random Forest and AdaBoost algorithm techniques have a value of 81.04% and 80.30%. With this high accuracy value, the Stacking model is proven to have better individual performance in analyzing employee turnover predictions in human resource applications in companies.

COMPARISON OF ACCURACY LEVELS OF SVM, DECISION TREE AND RANDOM FOREST ALGORITHMS IN SENTIMENT ANALYSIS OF USER RESPONSES OF THE GOPAY APPLICATION

2024-05-28T07:08:38+00:00

The development of technology from time to time makes all work or activities easier, one of which is online money transactions which are called e-wallets or digital wallets. One of the digital wallet applications that is often used is GoPay, which is a platform and tool created for making digital payments. Not long ago, GoPay was separated into one application, which previously existed in the Gojek application. However, every application certainly has a negative side, such as GoPay, where to use the application you have to be connected to the internet, which creates dependence on smartphones. Based on this problem, the company needs to know the response of users of the GoPay application which has been launched using the SVM, Decision Tree and Random algorithms. Forest. Therefore, the aim of this research is to carry out sentiment analysis on the responses of GoPay application users after being separated from Gojek and to find out the comparison of evaluation results or accuracy produced by the three algorithms. The results of this research show that of the three algorithms used, Positive sentiment is more than Negative sentiment, where in SVM Positive 89% and Negative 85%, Decision Tree class Positive 89% and Negative 76% while in Random Forest class positive 93% and Negative 86 %. Apart from that, the Random Forest algorithm has a high level of accuracy, namely 90%, then the SVM algorithm 88% and the Decision Tree algorithm 84%.

APPLICATION OF CANNY OPERATOR IN BATIK MOTIF IMAGE CLASSIFICATION WITH CONVOLUTIONAL NEURAL NETWORK APPROACH

2024-05-28T08:08:35+00:00

Batik, as Indonesia's cultural heritage, has high artistic value and has a variety of unique motifs.. The main focus of this research is to solve the problem of the complexity and diversity of motifs found in Indonesian batik culture. The Canny operator is used as a first step to extract the edges of batik motifs, with the aim of improving the quality of feature extraction before entering the classification stage using CNN, specifically by using the DenseNet121 model. The dataset of this study was obtained through the Kaggle platform, published by Dionisius Darryl Hermansyah. The platform consists of 983 images (.jpg) with 20 different Indonesian batik motifs. Pre-processing includes the use of Canny for edge detection and data augmentation to increase the diversity of the dataset. Next, variations in the number of epochs and batch size were used to train the model. The results show that in the first test, the use of the Canny operation gives a higher confidence level in the model. In the model with Canny, there is a 1.6% increase in accuracy (33.57% with Canny and 31.97% without Canny). In addition, there are differences in the level of confidence in some batik classes. For example, the "batik mega mendung" class shows an increase in confidence of 66.57% with Canny (88.53% with Canny and 21.96% without Canny), while the "batik sekar" class shows a decrease in confidence of 12.09% with Canny.

PROTOTYPE OF CONTACTLESS PAYMENT SYSTEM WITH RFID AND BLOCKCHAIN TECHNOLOGY INTEGRATED WITH MOBILE APPLICATION

2024-06-02T14:13:18+00:00

The COVID-19 pandemic has led to a change in payment methods, with a shift towards cashless payments to avoid germs and viruses. Contactless payments, especially those using RFID technology, have become popular due to their convenience and no need to enter security codes. However, there is a risk of data manipulation in transaction records, leading to decreased trust and transparency. To solve this problem, this research develops an RFID contactless payment system connected with blockchain technology as the main goal of the research. Blockchain is known for its security and transparency, making it suitable for minimizing data manipulation that often occurs. This research will use the Sepolia Test Network of the Ethereum base network for development in terms of blockchain to serve as a security layer in this research. The Waterfall method will be used for application development, focusing on structured and linear stages such as requirements analysis, system design, implementation, testing, and maintenance. The application development process has shown positive results, with successful black box testing and the ability to track and validate transactions stored in the database and blockchain. This validation process is critical to ensure the integrity of transactions and detect any data manipulation.

VISUAL ENTITY OBJECT DETECTION SYSTEM IN SOCCER MATCHES BASED ON VARIOUS YOLO ARCHITECTURE

2024-06-03T05:28:03+00:00

In this study, a performance comparison between the YOLOv7, YOLOv8, and YOLOv9 models in identifying objects in soccer matches is conducted. Parameter adjustments based on GPU storage capacity were also evaluated. The results show that YOLOv8 performs better, with higher precision, recall, and F1-score values, especially in the "Ball" class, and an overall accuracy (mAP@0.5) of 87.4%. YOLOv9 also performs similarly to YOLOv8, but YOLOv8's higher mAP@0.5 value shows its superiority in detecting objects with varying degrees of confidence. Both models show significant improvement compared to YOLOv7 in overall object detection performance. Therefore, based on these results, YOLOv8 can be considered as the model that is close to the best performance in detecting objects in the dataset used. This study not only provides insights into the performance and characteristics of the YOLOv7, YOLOv8, and YOLOv9 models in the context of object detection in soccer matches but also results in a dataset ready for additional analysis or for training deep learning models.

IMPROVING HEART DISEASE PREDICTION ACCURACY USING PRINCIPAL COMPONENT ANALYSIS (PCA) IN MACHINE LEARNING ALGORITHMS

2024-06-04T05:22:46+00:00

This study aims to improve the accuracy of heart disease prediction using Principal Component Analysis (PCA) for feature extraction and various machine learning algorithms. The dataset consists of 334 rows with 49 attributes, 5 classes and 31 target diagnoses. The five algorithms used were K-nearest neighbors (KNN), Logistic Regression (LR), Random Forest (RF), Support Vector Machine (SVM), and Decision Tree (DT). Results show that algorithms using PCA achieve high accuracy, especially RF, LR, and DT with accuracy up to 1.00. This research highlights the potential of PCA-based machine learning models in early diagnosis of heart disease.

PENETRATION TESTING OF A COMPUTERIZED PSYCHOLOGICAL ASSESSMENT WEBSITE USING SEVEN ATTACK VECTORS FOR CORPORATION WEBSITE SECURITY

2024-06-04T08:41:50+00:00

Websites, being dynamic platforms, undergo regular updates and continuous usage. Consequently, methods employed in website attacks evolve in tandem with increased security measures implemented in website systems, aiming to exploit both the website itself and its users. Website systems and features must remain prepared for potential future attacks at all times. To ensure this, penetration testing needed to be done consistently to keep up with security standards. This research aims to prove the various vulnerabilities that can be found from penetration testing in order to create recommendations on what to improve within a website. This research involves black box penetration testing of a computerized psychological testing website, developed by PT Dwi Purwa Teknologi hereinafter referred to as the client. The penetration testing simulated attacks by a foreign entity unfamiliar with the website's structure. The assessment focused on seven attack vectors: SQL injection, RCE, URL manipulation, CSRF, SSRF, XSS, and Broken Authentication and Session. Vulnerabilities resulted from poorly sanitized input forms, leading to SQL injection and RCE risks. Inadequate input validation enabled cross-site scripting attacks, while missing CSRF tokens exposed the website to CSRF threats. The research underscores the importance of penetration testing to identify and address security weaknesses, empowering the client to fortify their website against potential cyber threats.

COMBINATION OF LOGARITHMIC PERCENTAGE CHANGE-DRIVEN OBJECTIVE WEIGHTING AND MULTI-ATTRIBUTIVE IDEAL-REAL COMPARATIVE ANALYSIS IN DETERMINING THE BEST PRODUCTION EMPLOYEES

2024-06-04T13:59:32+00:00

The problem that occurs in the selection of the best production employees is the lack of transparency and objectivity in the selection process. Without clear procedures and well-defined criteria, employee selection decisions can be influenced by subjective preferences or irrelevant non-performance factors. This can result in injustice in employee selection and lower the morale and motivation of unselected employees. The purpose of the combination of LOPCOW and MAIRCA in determining the best production employees is to provide a holistic and adaptive framework in the employee performance evaluation process. LOPCOW allows decision makers to dynamically adjust the weight of criteria according to the level of volatility or change in the relevant environment or situation. LOPCOW offers an adaptive and responsive approach in determining the weight of criteria, enabling decision makers to respond quickly to changes occurring in the relevant environment or situation. MAIRCA is an analytical method used to assist decision makers in evaluating and selecting alternatives based on several relevant criteria or attributes. MAIRCA provides a strong framework for decision makers to make more informed and informed decisions. Combining these two methods results in a more comprehensive and accurate understanding of production employee performance, thus enabling managers to identify the most effective employees and provide rewards or development accordingly. The final results of the ranking of the best production employees obtained by JR employees get 1^st place, YP employees get 2^nd place, and AJL employees get 3^rd place.

ANALYSIS AND IMPLEMENTATION OF AES-128 ALGORITHM IN SUKAHARJA KARAWANG VILLAGE SERVICE SYSTEM

2024-06-05T05:07:02+00:00

Data security in databases is needed in the industrial era 4.0 to prevent attacks and unwanted things from happening, one of the biggest cases that has been widely reported is data leakage, in this study aims to implement and analyze the Advanced Encryption Standard Algorithm, one of the data security algorithms with a block chiper type that has 4 transformations (SubByte, ShiftColumn, MixColumn, AddRoundKey), or what we usually call the Cryptography method. Cryptography is a method that is often used to secure important data in databases, in this article the Advanced Encryption Standard Algorithm is used to secure citizen data and family card data in the Sukaharja Karawang Village service system. The method in this research is the observation method, the data is obtained from each head of the neighborhood in Sukaharja Karawang Village with the permission of the head of Sukaharja Karawang Village. Citizen data and family cards were encrypted and analyzed for resource requirements in storing encryption results and time in returning and displaying original data. The results of the analysis obtained the amount of resources required 1.5MB to store family card data, which before encryption required 352KB. Citizen data requires a resource of 6.5MB, before encryption it takes 1.5MB. As for the AES resilience test stage using the Bruteforce attack method with the help of Hashcat software version 6.2.5 with 4 trial processes, One encrypted address data was taken for this test, but out of 4 attempts none of them showed that the data could be cracked.

IMPLEMENTATION OF THE YOLOV8 METHOD TO DETECT WORK SAFETY HELMETS

2024-06-11T05:49:59+00:00

Work safety helmets are an important tool in OHS (Occupational Health and Safety) that must be used by workers. Workers who work with heavy equipment must wear work safety helmets as an obligation. Unfortunately, there are still many workers who do not comply with this rule. They will only wear helmets if there is supervision from a supervisor. However, if the supervisor is not on site, many workers will remove their helmets. The need for supervision of workers is important in reducing work accidents. From these problems, a work safety helmet detection model was created using the YOLOv8 method. This implementation aims to increase the accuracy values obtained and can reduce workload and increase efficiency in checking violations of the use of work safety helmets among workers. The method used consists of several stages, namely image acquisition of 670 images, image labeling, preprocessing, augmentation in roboflow, YOLOv8x model training with 100 epochs, image testing with a distance of 1, 3, 5 meters between the object and the camera, evaluation of test results. Based on the results of training with 467 images, the mAP50 reached 99.5%. Meanwhile, the test results with 100 images showed an accuracy of 99%.

APPLICATION OF PROCEDURAL CONTENT GENERATION SYSTEM IN FORMING DUNGEON LEVEL IN DUNGEON DIVER GAME

2024-06-05T08:05:01+00:00

Developers face numerous challenges in game development, one of which is the lack of games replayability due to the limited variety of levels created. The absence of level variety can lead to player boredom. The Procedural Content Generation (PCG) method provides an effective solution to address this challenge. PCG is applied with a focus on the Cellular Automata method by implementing the Von Neumann Neighborhood rule. The objective of this paper is to apply the Procedural Content Generation System method to create levels in game development. The game development process utilizes Luther's MDLC method. Testing is conducted using tiles of 32x32 units and 64x64 units, with three different test parameters: a fill percentage of 25%, 45%, and 65%. Each fill percentage is tested with three different smooth amount parameters of 2, 4, and 6, with a randomly selected seed. Performance testing results indicate that creating dungeon levels with 32x32 and 64x64 tiles yields short and relatively similar times, around 0.08 to 0.3 seconds. Functional testing reveals that a 25% fill percentage results in nearly empty rooms with no footholds, a 45% fill percentage produces levels with space and footholds, while a 65% fill percentage generates small unconnected rooms. Based on these percentages, a 45% fill percentage is considered the most appropriate for creating dungeon levels because it provides suitable space and footholds for players. Implementing PCG in game level creation not only saves time compared to manual level creation but also offers more efficient variations in dungeon shapes and difficulty levels.

FISH FRESHNESS PREDICTION WITH CONVOLUTIONAL NEURAL NETWORK METHOD BASED ON FISH EYE IMAGE ANALYSIS

2024-06-05T08:38:44+00:00

The potential for fish resources in Bengkulu waters is abundant, but quality must be maintained for safety and selling value. Changes in the skin, eyes, gills and flesh of fish indicate a decrease in quality due to enzyme, chemical and bacterial activity. The process of sorting fish by fishermen or sellers is still often done manually, which is sometimes inaccurate due to limited vision. With advances in computing technology, classification algorithms are needed that can identify and differentiate between fresh fish and non-fresh fish. This research uses a Convolutional Neural Network with DenseNet201, VGG16, and InceptionV3 architecture. The dataset contains 880 Belato Alepes Djedaba fish eye images, with a ratio of 80:15:5 for train, validation, and test. DenseNet201 has the best performance compared to VGG16 and InceptionV3. Accuracy on DenseNet201 test data 98%, InceptionV3 95%, and VGG16 91%. The classification results of the best model using 8 images with various scenarios show that all images were successfully classified 100% correctly. This research makes a contribution to the field of fishery product processing technology which allows fish quality classification to be carried out quickly and accurately, as well as increasing efficiency in ensuring the quality of fish for consumption.

EXPERT SYSTEM WITH DEMPSTER-SHAFER METHOD FOR EARLY IDENTIFICATION OF DISEASES DUE TO COMPLICATIONS SYSTEMIC INFLAMMATORY RESPONSE SYNDROME

2024-06-05T09:13:41+00:00

Systemic Inflammatory Response Syndrome (SIRS) is a generalized inflammatory condition, triggered by various factors such as infection or trauma, which can lead to serious complications if not treated quickly. This condition is characterized by symptoms such as fever or hypothermia, tachycardia, tachypnea, and changes in white blood cell count. Complications that can arise from SIRS include Acute Respiratory Distress Syndrome (ARDS), which results in fluid in the alveoli and requires mechanical ventilation; acute encephalopathy, which leads to brain dysfunction; Asidosis Metabolik, indicating liver damage; hemolysis, which results in the breakdown of red blood cells; and Deep Vein Thrombosis (DVT), which is at risk of causing pulmonary embolism. To overcome this diagnostic challenge, this study implements the Dempster-Shafer method in an expert system, where it allows the aggregation and combination of various sources of evidence to produce degrees of belief and degrees of plausibility for each diagnostic hypothesis. By accounting for uncertainties and contradictions in the data, the system improves diagnostic accuracy through dynamically weighting and updating beliefs based on available evidence. This process allows early and accurate identification of SIRS complications, supporting appropriate medical intervention. System evaluation showed diagnostic accuracy of 93%, confirming the potential of expert systems in supporting rapid and precise clinical decision-making in managing SIRS complications.

OPTIMIZATION OF MACHINE LEARNING MODEL ACCURACY FOR BRAIN TUMOR CLASSIFICATION WITH PRINCIPAL COMPONENT ANALYSIS

2024-06-11T07:46:55+00:00

The main issue in brain tumor classification is the accuracy and speed of diagnosis through medical imaging. This study aims to improve the accuracy of machine learning models for brain tumor classification by using Principal Component Analysis (PCA) for dimensionality reduction. The research methods include image preprocessing, feature scaling, PCA application, and the implementation of machine learning algorithms such as Logistic Regression, Random Forest, Support Vector Machine (SVM), K-Nearest Neighbor (KNN), and Naive Bayes. The dataset consists of 3,264 images divided into training and testing sets. The results show that the use of PCA has varying impacts on different algorithms. PCA increases the accuracy of the SVM algorithm from 81% to 83% and KNN from 68% to 71%, but decreases the accuracy of Logistic Regression from 77% to 69% and Naive Bayes from 49% to 42%. Evaluation is performed using the Confusion Matrix and AUC-ROC to measure model performance. In conclusion, selecting the appropriate algorithm and preprocessing method is crucial in medical image classification, and the use of PCA should be considered based on the characteristics of the data and the algorithms used. This study also encourages the exploration of alternative dimensionality reduction methods for medical image analysis.

TEMPORAL SPATIAL PROPERTY PROFILING AND IDENTIFICATION OF EARTHQUAKE PRONE AREAS USING ST-DBSCAN AND K-MEANS CLUSTERING

2024-06-18T10:21:08+00:00

Indonesia is a country located at the confluence of three major tectonic plates, namely Indo-Australia, Eurasia, and the Pacific so that earthquakes often occur, one of which is in West Nusa Tenggara Province. One way to accelerate the disaster mitigation process is to analyze earthquake occurrence based on spatial temporal aspects. This study uses data from BMKG NTB Province during 2018 with a total of 3,699 earthquake events which are then analyzed using ST-DBSCAN and K-Means. ST-DBSCAN analysis was used to determine earthquake prone areas based on the date and location of the event, while k-means used the depth and magnitude of the earthquake. The results show that the distribution pattern of earthquakes in the NTB region has a stationary pattern and there are similar prone areas based on the location and time of occurrence as well as the strength and depth of the earthquake. The ST-DBSCAN method using latitude and longitude attributes produces one cluster that covers 96.33% of the total data. Meanwhile, K-Means using the depth and magnitude attributes produced four clusters. The four clusters were obtained from the cluster density using the silhouette score value between -1 and 1. The K-means analysis used a silhouette score result of 18.527 which was found in cluster 1. Earthquake prone areas in the distribution of earthquakes or types of earthquakes are located in Gangga and Bayan sub-districts of North Lombok and in Sambelia and Sembalun sub-districts of East Lombok. The sub-district with the most frequent earthquakes is Sambelia sub-district with 112 earthquakes. Then the strength of the largest earthquakes on average occurred in Gangga sub-district with magnitudes of 4 to 6.2 SR with shallow earthquake types. The prone area is located at the foot of the mountain and directly adjacent to the ocean.ith shallow earthquake types. The Prone area is at the foot of a mountain and directly adjacent to the ocean.

COMPARISON OF DEEP LEARNING ARCHITECTURES FOR ANEMIA CLASSIFICATION USING COMPLETE BLOOD COUNT DATA

2024-06-22T06:03:02+00:00

Anemia is a common condition marked by a deficiency in red blood cells or hemoglobin, affecting the body's ability to deliver oxygen to tissues. Accurate and timely diagnosis is essential for effective treatment. This study aims to classify different types of anemia using complete blood count (CBC) data through the application of deep learning models. We evaluated the performance of four deep learning architectures: Multi-Layer Perceptron (MLP), Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), and Fully Connected Network (FCN). The dataset included CBC parameters such as hemoglobin, platelet count, and white blood cell count, labeled with anemia types. Our results indicate that CNN and FCN models achieved the highest test accuracies of 0.85, outperforming MLP and RNN models. This superior performance is due to the ability of CNN and FCN to capture complex patterns and spatial relationships within CBC data. Techniques like data augmentation and weighted loss functions were employed to address class imbalance. These findings demonstrate the potential of deep learning models to automate anemia diagnosis, thereby enhancing clinical decision-making and patient outcomes.

IMPLEMENTATION OF LOW-CODE PROGRAMMING TECHNOLOGY WITH AGILE METHOD IN DEVELOPING A PETTY CASH TRANSACTION MANAGEMENT APPLICATION (CASE STUDY: PT BANK CENTRAL ASIA TBK)

2024-07-02T03:18:18+00:00

Conventional application development often faces challenges like extensive code writing, long development times, high costs, and difficulties in maintenance and customization. Low-code programming offers an innovative solution by minimizing manual coding and enabling application creation through visual interfaces and drag-and-drop logic. This research explains the application of low-code programming technology in developing a petty cash transaction recording application at PT Bank Central Asia Tbk, specifically in the Corporate Communication & Social Responsibility Division. The low-code approach allows for faster, more efficient, and easier-to-maintain application development. The research uses the agile method, covering plan, design, develop, test, deploy, review, and launch stages. This case study, using the OutSystems platform, shows significant benefits such as increased development time efficiency, ease of maintenance, and flexibility in meeting dynamic business needs. The developed application can be integrated into the company's existing IT environment, improving the accuracy of petty cash transaction recording and reporting, and providing easy user access. In conclusion, low-code programming technology proves to be an effective solution for developing complex business applications efficiently in terms of time and cost.