Journal of Computer Networks, Architecture and High Performance Computing
Not a member yet
473 research outputs found
Sort by
Enhancing Water Quality Early Warning System Accuracy in Pangasius Aquaculture Using Machine Learning
No full textIntensive catfish (Pangasius sp.) aquaculture faces significant economic risks driven by mass mortality events linked to unstable water quality, particularly toxic ammonia spikes and pH fluctuations. Although Internet of Things (IoT) technology enables real-time monitoring, the resulting time-series data presents complex challenges, including high sensor noise, asynchronous transmission, and severe class imbalance, which compromise standard reactive monitoring methods. This study aims to enhance diagnostic accuracy by comparing Support Vector Machine (SVM), Random Forest (RF), and XGBoost algorithms to construct a robust Early Warning System (EWS). A quantitative experimental methodology was applied to real-world sensor data, with temporal aggregation preprocessing to reduce noise. To ensure rigorous validation simulating real-world deployment, the dataset utilized a strict chronological split (80% training, 20% testing) and was further tested using 5-Fold Time-Series Cross-Validation. The results demonstrated the definitive superiority of ensemble-based models; Random Forest and XGBoost achieved 100.00% accuracy on the test set, successfully eliminating the critical false negatives exhibited by the SVM model (99.80%). Stability analysis further confirmed the robustness of Random Forest (98.35%) and XGBoost (98.32%) compared to SVM (97.02%). Additionally, feature importance analysis unequivocally identified ammonia as the dominant predictor of critical conditions. Crucially, the study detected a “concept drift” phenomenon in which “Safe” conditions disappeared during the final cultivation phase. These findings conclude that ensemble models provide the optimal architecture for EWS. However, the presence of concept drift necessitates adaptive retraining strategies to ensure long-term reliability in dynamic pond environments
Performance Analysis of an Offline Text Detection System Based on Edge AI A Case Study of DokuScan Pro
No full textThe growing use of mobile document scanning applications has increased the demand for text detection systems that can operate reliably in offline and on-device environments. Although Edge AI enables local inference without network dependency, system-level empirical evidence regarding its performance under real-world mobile usage conditions remains limited. This study presents a system-level evaluation of an offline Edge AI–based text detection system for mobile document scanning, using DokuScan Pro as a case study. The evaluation was conducted on 40 document images captured under varying lighting conditions, capture angles, and background characteristics. System performance was assessed using precision, recall, F1-score, and inference time to characterize on-device behavior rather than algorithmic novelty. Experimental results show that the system achieved a precision of 1.00, a recall of 0.975, and an F1-score of approximately 0.98, with an average inference time of 63.8 ms per image during fully offline execution on mobile devices. These results indicate stable system-level performance under real-world document scanning conditions with controlled computational overhead. This study provides empirical system-level insights into the feasibility and practical limitations of deploying Edge AI–based text detection in offline mobile document scanning applications, thereby complementing existing model-centric research with evidence from real-world, on-device evaluation
IoT-Based Smart Air Quality System: A Real-Time Monitoring Solution for Indoor Air Quality
No full textIndoor Air Quality (IAQ) plays a crucial role in maintaining human health and comfort. This study aims to design and implement an Internet of Things (IoT)-based indoor air quality monitoring system integrated with a mobile application for real-time observation. The system employs sensors to measure environmental parameters such as temperature, humidity, and carbon dioxide (CO?) levels, with data transmitted wirelessly and visualized through the mobile app. The applied method includes hardware design, IoT-based software development, and system testing in several rooms with different activity conditions. The implementation results show that the system can accurately display air quality data and provide automatic notifications when pollutant levels increase. Based on seven days of measurement, the kitchen area indicated a “Poor Air” category, while the living room and bedroom were classified as “Fresh Air.” This system effectively delivers fast and accurate air quality information, enabling users to take preventive actions to maintain healthy indoor air condition
Minimizing Subjectivity in Esports Adjudication: A Decision Support System for Indonesia Sim Racing League Using C4.5 Algorithm
No full textThe adjudication of racing incidents in the Indonesia Sim Racing League (ISL) currently faces challenges due to inherent subjectivity, inconsistency, and the time-consuming nature of decisions that rely solely on race stewards’ interpretations. This study develops a Decision Support System (DSS) for penalty recommendation in ISL racing incidents by applying the Decision Tree C4.5 algorithm. Historical incident data were collected directly from Indonesia Sim Racing League Seasons 1 to 3, and an additional synthetic dataset was generated based on predefined incident attributes to support model training. All data were processed using Python in the Google Colab environment to train and evaluate the C4.5 model. Experimental results show that the proposed DSS achieved an overall accuracy of 90%, indicating strong predictive capability in recommending appropriate penalties under the given dataset configuration. Further evaluation using class-sensitive metrics yielded a macro-average precision of 0.71, a recall of 0.73, and an F1-score of 0.72, reflecting a more balanced performance across penalty classes despite the presence of class imbalance in racing incident data. These results indicate that the model is able to capture relevant decision patterns while maintaining robustness across both majority and minority penalty classes. Overall, this study demonstrates that the proposed DSS can assist race stewards at an early stage of decision-making by narrowing the decision space and reducing subjective bias, thereby supporting fairer and more consistent adjudication processes. The main contribution of this paper lies in presenting one of the first empirical implementations of a DSS for esports racing adjudication using an interpretable C4.5-based approach, providing a transparent and practical foundation for future research on intelligent decision-support systems in competitive sim racing environments
Evaluation of Machine Learning Algorithms for an Early Warning System of Student Graduation in a Python Programming Course
No full textThe high failure rate in Python programming courses has become a serious issue for educational institutions. This study aims to evaluate the performance of four machine learning algorithms as the basis of an Early Warning System for predicting student graduation, namely Random Forest (RF), Support Vector Machine (SVM), Logistic Regression (LR), and K-Nearest Neighbors (KNN). The dataset consists of 3,000 records with 15 features, including demographic data, programming experience, and students’ learning activities. Performance evaluation was conducted using accuracy, precision, recall, F1-score, and ROC-AUC metrics after optimal hyperparameter tuning through GridSearchCV with 5-fold cross-validation. The evaluation results indicate that Random Forest achieved the best performance with an accuracy of 89.33%, precision of 87.50%, recall of 46.23%, F1-score of 60.49%, and ROC-AUC of 94.40%, outperforming SVM (accuracy 86.33%, F1-score 55.43%), Logistic Regression (accuracy 86.50%, F1-score 53.71%), and KNN (accuracy 84.83%, F1-score 44.17%). Feature importance analysis identified experience_encoded, hours_spent_learning_per_week, and projects_completed as the three strongest predictors of student graduation. These findings provide empirical evidence that Random Forest is the most effective algorithm for implementing an Early Warning System in Python programming courses, enabling instructors to identify at-risk students early and provide timely interventions to improve learning success rates
Sentiment Analysis of Shopee User Reviews Using Recurrent Neural Network with LSTM for Real-Time Web-Based Prediction
No full textSentiment analysis has become an important approach for understanding user opinions on e-commerce platforms. Shopee user reviews provide valuable information that can be utilized to evaluate service quality and customer satisfaction. This study aims to analyze the sentiment of Shopee user reviews using a Recurrent Neural Network with Long Short-Term Memory (RNN-LSTM) architecture. The research method includes data collection, text preprocessing, model training, and performance evaluation. The experimental results show that the proposed RNN-LSTM model achieved an accuracy of 97%, indicating its effectiveness in classifying user sentiment. The developed model is further implemented in a web-based application to provide real-time sentiment prediction. The findings of this study demonstrate that the RNN-LSTM approach is suitable for sentiment analysis in e-commerce environments and can support decision-making based on user feedback
The Website-Based Information System Design At Cahaya Harapan Jaya Building Shop
No full textBuilding material stores play an important role in providing building materials, but many still rely on conventional management systems that are less efficient in recording stock, transactions, and employee data. This results in delays in recording transactions, errors in stock monitoring, and the lack of an effective system in managing employee data. Therefore, this study aims to design a web-based information system to improve the operational efficiency of building material stores. This study uses the Waterfall method which consists of needs analysis, design, programming, testing and implementation of this system also uses blackbox testing and system usability scale (SUS) testing. This system is designed to make it easier for administrators and employees to manage products, transactions, employee data, and financial reports in real-time. The results of the study show that the web-based information system improves the efficiency of transaction recording, minimizes stock monitoring errors, and simplifies the management of employee data and financial reports. This system allows building material stores to optimize operational performance, improve data accuracy, and provide better customer service
Sentiment Analysis on Short Social Media Texts Using DistilBERT
Full text linkSentiment analysis on short texts from social media, such as tweets, presents unique challenges due to their brevity and informal language. This study explores the effectiveness of transformer-based models, particularly DistilBERT, in performing sentiment analysis on short texts compared to traditional machine learning approaches including Support Vector Machine, Logistic Regression, and Naive Bayes. The objective is to assess whether DistilBERT not only enhances sentiment classification accuracy but also remains efficient enough for quick social media analysis. The models used in this study were trained and evaluated on stratified samples of 10,000, 30,000, and 50,000 tweets, drawn from the Sentiment140 dataset while preserving the original class distribution. The methodology involved data collection and sampling, data splitting, data cleaning, feature extraction, model training, and evaluation using accuracy and F1-score. Experimental results showed that DistilBERT consistently outperformed traditional models in both accuracy and F1-score, and demonstrated competitive results against BERT while requiring significantly less training time. Specifically, DistilBERT trained approximately 1.8 times faster than BERT on average, highlighting its computational efficiency. The best result was achieved by DistilBERT trained on the 50k subset, reaching an accuracy of 85% and an F1-score of 84%. These findings suggest that lightweight transformer models like DistilBERT are highly suitable for real-world sentiment analysis tasks where both speed and performance are critical
Facial Expression Recognition Using Fused Features: A Comparison of Deep and Machine Learning
Full text linkFacial expression recognition (FER) is a highly active field with applications in computer vision, human-computer interaction, security, and computer graphics animation. Recent advancements in deep learning and machine learning have increased interest in utilizing these techniques for accurate facial expression classification. This paper presents a comparative study that evaluates the performance of deep learning and machine learning as classifiers in FER systems, specifically after data fusion. Data fusion techniques combine and integrate multiple sources of information, aiming to enhance the overall classification accuracy by extracting two types of features using geometrical and appearance features trained using two types of convolutional neural networks. The feature outputs of these networks are fused to create a final feature vector for the classification process. The study evaluates the performance of deep learning on two benchmark datasets, the extended Cohn-Kanade (CK+) and Oulu-CASIA datasets, to assess the performance of deep learning. As a point of comparison, the traditional machine learning approach based on the support vector machine (SVM) is also evaluated on the same datasets. Performance metrics such as classification accuracy, precision, recall, and F1-score are utilized. The results obtained from the study highlight the strengths and limitations of both deep learning and machine learning techniques when employed as classifiers in FER systems. Notably, the experimental results demonstrate that the deep learning approach significantly outperforms the baseline methods, achieving an increase in recognition accuracy of 5.22% for the CK+ and 3.07% for the Oulu-CASIA dataset
Design of an Arduino Based Automatic Sealing Machine with DS18B20 Sensor for Smart Temperature Control
Full text linkSealing quality is a critical concern for Micro, Small, and Medium Enterprises (MSMEs) that rely on conventional machines lacking temperature control mechanisms. These systems often result in overheating or poor bonding, especially when applied to thin plastic materials such as polyethylene (PE), polypropylene (PP), and oriented polypropylene (OPP), each with varying melting points. This research aims to design an Arduino Uno-based automatic horizontal sealing machine integrated with a DS18B20 temperature sensor to provide smart temperature control during the sealing process. The proposed system employs a threshold-based ON-OFF control algorithm with a hysteresis margin of ±2.5°C, and displays real-time thermal feedback on a 16x2 LCD. The experimental methodology includes temperature deviation analysis and quality scoring of sealing results across PE, PP, and OPP films. Results show that the manual system deviated up to 15.4°C from the target temperature, leading to inconsistent outcomes. In contrast, the Arduino-based system maintained thermal stability within ±5°C and achieved a significant increase in sealing quality score from 60 to 92. These improvements indicate enhanced operational reliability, safety, and sealing consistency. The system provides a low-cost, scalable solution for MSMEs and can be upgraded to include PID control, IoT integration, or adaptive thermal profiling. This work demonstrates that embedded microcontroller-based automation is feasible and effective for small-scale packaging application