Journal of Computer Networks, Architecture and High Performance Computing
Not a member yet
473 research outputs found
Sort by
Evaluation of the Quality of Academic Information Services on the FEB UNJ Website Based on the WebQual and EUCS Models
Get PDFThis study aims to evaluate the quality of academic information services provided through the website of the Faculty of Economics and Business, State University of Jakarta (www.feb.unj.ac.id). The research approach used was quantitative descriptive using the WebQual 4.0 model which was modified through the addition of the Eas of Use variable from the End User Computing Satisfaction (EUCS) model. A total of 124 respondents consisting of students and lecturers were involved in this study through the distribution of questionnaires. Data were analyzed using multiple linear regression with the help of SPSS. The results showed that the variables Information Quality and Interaction Quality had a significant influence on user satisfaction, while Usability and Content did not show a statistically significant partial effect. However, simultaneously, the four variables contributed 70.8% to user satisfaction. This study emphasizes the importance of accuracy, trust, and ease of access as the main factors in supporting user satisfaction with website-based academic information services. This research contributes to the development of academic information systems by offering a comprehensive evaluation framework that integrates the WebQual and EUCS models, while also emphasizing the critical roles of information accuracy, user interaction, and ease of access in enhancing user satisfaction thus providing a practical reference for improving the quality of academic websites in higher education institutions in Indonesia
Comparative Analysis of Incoming Goods Patterns Using FP-Growth and Apriori Algorithms: A Case Study in Retail
Get PDFThis study aims to analyze consumer purchasing patterns in minimarkets using the Apriori and Fp Growth association algorithms based on transaction data, where the data consists of 10 goods receipt transactions with 7 variable items such as Ultra Milk UHT 250ml, Indomie Goreng Spesial, Beras Ramos 5kg, Teh Cup Sariwangi 25's, Minyak Goreng Bimoli 1L, Soap Bar Lifebuoy 75g, and Mie Lemonilo Goreng 70g. The analysis process is carried out through the preprocessing stage, transformation to binary format, and application of the algorithm with minimum support parameters of 20% and confidence of 50%. The results show that Ultra Milk UHT 250ml has the highest support (0.5) followed by Indomie Goreng Spesial (0.4), while the combination of UHT Milk with Indomie has a support of 0.2; in terms of confidence, a number of rules even reach a perfect value of 1.0, for example the relationship between Teh Cup Sariwangi and Ultra Milk which always appear together. Quantitatively, Apriori produces 25 association rules with a processing time of approximately 2.1 seconds, while Fp Growth produces the same number of rules but is more efficient with a processing time of 1.3 seconds and lower memory usage, so it can be concluded that although both are equal in terms of rule quality, Fp Growth is superior in computational efficiency. This finding has important practical implications for minimarket management, especially to support shelf arrangement strategies, more targeted stock planning, and the preparation of bundling promotions based on product combinations with high confidence, while also showing a scientific contribution in the form of comparing the performance of two association algorithms on incoming goods data that is relatively rarely used in previous studies
Analysis of C4.5 Algorithm Performance for Predicting Student Achievement Based on Socio-Economic Status, Motivation, Discipline, and Past Achievement
Get PDFLearning outcomes are changes that a person undergoes when learning. Learning outcomes can be in the form of changes in cognitive, affective, and psychomotor abilities depending on the learning objectives Student learning outcomes vary depending on each student's individual circumstances. Various factors such as family conditions, school environment, interests, motivation, and past achievements determine the success of learning outcomes. The problem occurring at SMK Negeri 4 Kota Metro is that the students come from various villages in the city. They mostly come from underprivileged families with low education levels. In addition, they are less motivated to study due to factors related to their family environment and the surrounding community. Most of the students at this school do not have good achievements in academic or non-academic fields.This research aims to predict student academic performance based on the socioeconomic status of parents, motivation, student discipline, and past achievements using data mining methods with the C.45 algorithm. For comparison, the research data was also analyzed with. The research approach used is quantitative. The subjects of this research are 606 students from the 10th grade of SMK Negeri 4 Kota Metro. The data collection techniques used were documentation and questionnaires. The research results show that the prediction analysis using decision tree has an accuracy of 98.02%, precision of 94.44%, recall of 77.27%, and AUC of 0.96
Enhanced Plant Disease Detection Using Computer Vision YOLOv11: Pre-Trained Neural Network Model Application
Get PDFThis study investigates the application of YOLOv11, a cutting-edge deep learning model, to enhance the detection of plant diseases. Leveraging a comprehensive dataset of 737 images depicting tomato leaves affected by various diseases, YOLOv11 was trained and evaluated on key performance metrics such as precision, recall, and mAP. Experimental results the model was trained and evaluated on key metrics including accuracy (75.6%), precision (0.80), recall (0.77), and [email protected] (75.6%). Experimental through base architectural such as enhanced feature extraction with C2 modules, improved multi-scale detection using SPPF layers, and optimized non-maximum suppression techniques. These improvements enable the model to achieve stable precision and recall for each class, even in challenging scenarios with overlapping objects and diverse environmental conditions. By addressing practical usability challenges, this system offers a scalable, accessible, and impactful solution for precision agriculture, paving the way for sustainable with this pretrained model. This study underscores the potential of deep learning-based models, particularly YOLOv11, in transforming the way monitoring and disease management are approached, demonstrating its ability to stable accuracy and operational efficiency in real-world applications. Furthermore, the practical usability of the YOLOv11-based system addresses challenges in the domain of precision plant detection desease. By providing a scalable, accessible, and highly efficient solution, the model offering a significant advancement toward sustainable agricultural practices
Analysis of Outpatient Patient Visit Prediction at Muntilan Regional General Hospital Using Linear Regression Method
Get PDFHospitals play a crucial role in public health, and understanding patient visit patterns is essential for effective service delivery. Thus, accurate predictions are vital for resource planning, service improvement, and addressing challenges like long wait times and overcrowding. This study focuses on predicting outpatient visits at RSUD Muntilan, a regional general hospital in Magelang, Indonesia. The method used was the linear regression method. The research involved data collection from the hospital's information system, pre-processing to prepare the data, dataset formation, model creation using linear regression, and model evaluation. The study utilized historical outpatient visit data FROM 2021 TO 2024 to develop a linear regression model that predicts the number of visits for the next three months. The model's performance was evaluated using the Mean Absolute Percentage Error (MAPE), which yielded a value of 15.33%. This indicates that the model's predictions were, on average, within 15.33% of the actual values, demonstrating an accuracy of 84.67%. The successful application of the linear regression method in this study highlights its potential for improving resource allocation, enhancing service efficiency, and ultimately enhancing the overall quality of healthcare services provided by RSUD Muntilan. The findings emphasize the significance of data-driven approaches and predictive analytics in optimizing healthcare operations and meeting the evolving needs of the community.
Web-Based E-Commerce Using Up Selling and Safety Stock Methods
Get PDFThis research examines the application of the e-commerce platform at the An-Nur Muslim Shop which sells Hajj and Umrah equipment. In the rapidly developing digital era, e-commerce has become essential for companies to expand market share and increase operational efficiency. E-commerce is changing the traditional way of commerce, allowing customers to shop easily and safely. The principle of fair and legal transactions in Islam is an important basis for the development of e-commerce, as emphasized in QS An-Nisa: 29. The Up Selling method is applied to increase sales by offering similar products that have a higher value, which is expected to encourage customers to buy more products. Up Selling works by sorting products from lowest to highest price and offering discounts and special promos on products that meet the criteria. In addition, the Safety Stock concept is used to manage inventory more efficiently. Safety Stock helps anticipate demand uncertainty and delays in receiving raw materials, maintaining optimal stock levels. With this approach, An-Nur Moeslim Shop’s can keep stock levels low to reduce inventory management costs and match new orders only when stock is low. The proposed e-commerce system will allow users to search for products, review descriptions and prices, and make purchases easily through a user-friendly web interface. It is hoped that this e-commerce platform will strengthen An-Nur Moeslim Shop’s position in the market, provide significant added value, and improve the overall customer shopping experience
The Comparison of the K Mean Algorithm with the C 45 Algorithm in Dataming Applications: Balancing Precision and Speed in Data Mining Solutions
No full textThis research topic discusses the comparison of the K-Means and C4.5 algorithms in the application of data mining to predict aquarium sales in a company. K-Means is a clustering algorithm that functions to group data based on similarity, for example grouping customers based on frequency or type of purchase. This helps companies understand market segments and design marketing strategies accordingly. Meanwhile, C4.5 is a classification algorithm that builds decision trees based on important attributes that influence sales, such as price, season, or promotions. This algorithm is able to predict sales categories, such as increases or decreases, based on historical data. By comparing these two algorithms, the research sought to find out which algorithm is more effective in helping companies predict sales and make strategic decisions. A combination of the two can also be used, with K-Means grouping the data first, then C4.5 classifying each segment formed. These results can provide more accurate sales predictions and more effective marketing strategies. This research is important to understand the effectiveness of algorithms in data mining to improve business decision making
SHORT-TERM ELECTRICITY LOAD FORECASTING SEASONAL PATTERN USING TIME SERIES REGRESSION (TSR) MODEL IN PT.PLN (PERSERO) MEDAN CITY
No full textElectricity is a crucial component of modern life, where daily consumption fluctuates significantly. Uncertain electricity demand can lead to imbalances between supply and consumption, potentially causing energy wastage or power outages. To address this issue, a forecasting method capable of accurately predicting electricity load is essential. The Time Series Regression (TSR) model is applied for short-term electricity load forecasting by considering daily and weekly seasonal patterns. The forecasting results indicate that Monday and Tuesday have the highest electricity load, while Sunday has the lowest. When the Kolmogorov-Smirnov test is used to analyse the model, the p-value is 0.9608, which shows that the residuals have a normal distribution. The model's accuracy is assessed with a Root Mean Square Error (RMSE) value of 378.0069 MW, which is relatively high for a small dataset. Given the considerable forecasting error, further improvements such as hybrid models are recommended to enhance accuracy. The implementation of these forecasting results can help optimize electricity management and improve power distribution efficiency
PROTOTYPE OF INTERNET OF THINGS (IOT) IMPLEMENTATION IN WASTE MANAGEMENT TO SUPPORT SMART CITY MONITORING WITH ANDROID-BASED MOBILE APPLICATION USING FORWARD CHAINING METHOD
Get PDFEfficient waste management is one of the main challenges in supporting the implementation of the smart city concept. This research aims to develop a prototype of an Internet of Things (IoT)-based waste management system capable of monitoring the condition of waste bins in real-time through an Android-based mobile application. The system uses the forward chaining method to perform inference processes in decision making, such as identifying the status of the bin (empty, almost full, or full) based on integrated sensor data. The results show that the system is able to detect the volume of waste with high accuracy, send automatic notifications to operators or users when the bin reaches a certain condition, and provide practical solutions to optimise the waste collection process. With these features, the system not only improves operational efficiency but also supports cost reduction and environmental impact. The resulting prototype is expected to be the first step in the application of IoT technology in urban waste management to support the realisation of smart cities
APPLICATION OF KNN METHOD FOR CLASSIFICATION OF ARRHYTHMIA TYPES BASED ON ECG DATA
Get PDFWorld Health Organization (WHO) data from June 2024 shows that 31% of adults worldwide or 1.8 billion people do not do physical activity. With that, adults are at higher risk of developing cardiovascular disease and causing an economic and social burden on people with heart disease. K-Nearest Neighbor (KNN) is a machine learning method that can be used to classify or predict heart disease conditions. KNN works by finding the closest data point in the training dataset and then using the class labels of those neighbors to classify new data points. In the context of heart disease, this can be used to predict the likelihood of someone having heart disease. Recording the electrical activity of the heart using a 3-led ECG to determine heart health as well as being material for classification. Exploring the use in the diagnosis of heart disease by focusing on screening and classification of heart disease. By utilizing the KNN method, it has the potential to produce a model that can assist in clinical decision making. Improving the prevention of heart disease and accelerating diagnosis through more sophisticated and technology-based analysis of patient health data