Jurnal Politeknik Negeri Batam (PoliBatam)
Not a member yet
3001 research outputs found
Sort by
Regional Clustering in Sumatera Based on Walfare Indicators Using Fuzzy C-Means
No full textWelfare refers to a condition in which individuals have sufficient means to meet both physical and spiritual needs. In Indonesia, welfare is a national goal, yet Sumatra experiences the highest development disparity, contributing to unequal welfare distribution across regions. This study aims to cluster regions in Sumatra based on welfare indicators using the Fuzzy C-Means (FCM) method, analyze cluster characteristics, and provide policy recommendations for decision-makers. FCM is used because it accommodates uncertainty and allows each data point to belong to more than one cluster, making it suitable for welfare analysis. Cluster validity was tested using Partition Coefficient Index (PCI) and Silhouette Coefficient, both indicating that the optimal number of clusters is two. The results show that Cluster 1 consists of 62 regions with relatively higher welfare conditions, while Cluster 2 includes 92 regions with lower welfare characteristics. One notable member of Cluster 2 is Ogan Komering Ulu, with a high membership degree of 0.869. Recommended policies include improving access to clean water and healthcare, enhancing education, strengthening local economies, and delivering targeted social assistance to underdeveloped areas. For Cluster 1, sustainable development efforts should be maintained.Welfare refers to a condition in which individuals have sufficient means to meet both physical and spiritual needs. In Indonesia, welfare is a national goal, yet Sumatra experiences the highest development disparity, contributing to unequal welfare distribution across regions. This study aims to cluster regions in Sumatra based on welfare indicators using the Fuzzy C-Means (FCM) method, analyze cluster characteristics, and provide policy recommendations for decision-makers. FCM is used because it accommodates uncertainty and allows each data point to belong to more than one cluster, making it suitable for welfare analysis. Cluster validity was tested using Partition Coefficient Index (PCI) and Silhouette Coefficient, both indicating that the optimal number of clusters is two. The results show that Cluster 1 consists of 62 regions with relatively higher welfare conditions, while Cluster 2 includes 92 regions with lower welfare characteristics. One notable member of Cluster 2 is Ogan Komering Ulu, with a high membership degree of 0.869. Recommended policies include improving access to clean water and healthcare, enhancing education, strengthening local economies, and delivering targeted social assistance to underdeveloped areas. For Cluster 1, sustainable development efforts should be maintained
Lightweight BiLSTM-Attention Model Using GloVe for Multi-Class Mental Health Classification on Reddit
No full textMental health issues such as depression, stress, anxiety, and personality disorders are increasingly prevalent, particularly within online communities. This study proposes a lightweight and efficient multi-class classification framework to identify five mental health conditions using Reddit user-generated posts. While previous studies predominantly rely on conventional CNNs or standard machine learning techniques for binary classification, our work introduces a novel Bidirectional Long Short-Term Memory (BiLSTM) model integrated with an attention mechanism. The architecture is further enhanced by synonym-based data augmentation using the WordNet lexical database, which improves semantic diversity and enhances model robustness, particularly for underrepresented classes. Unlike prior works that focus narrowly on binary classification or employ transformer-based models with high computational demands, our model offers a lightweight, high-performance architecture optimized for multi-class detection and real-world deployment. Experimental results demonstrate that the proposed model achieves a peak validation accuracy of 95.02%, along with precision 95.08%, recall 95.02%, and F1-scores of 95.03%. These findings support the advancement of efficient AI-driven diagnostic systems in mental health analytics and lay the groundwork for future integration into mobile or resource-constrained platforms
Analysis of Factors Affecting the Delay in Completion of Student Final Projects Using the C5.0 Decision Tree Algorithm
No full textDelays in completing final projects are a common problem faced by students and can lead to delayed graduation, increased study load, and reduced readiness to enter the workforce. This study uses a quantitative predictive approach to analyze the factors influencing delays in completing student final projects by applying the C5.0 Decision Tree classification algorithm. Data were collected through a Likert-scale questionnaire from 204 students of the Faculty of Engineering, University of 17 August 1945 Surabaya, who graduated between 2019 and 2021. The analyzed factors include time management, student motivation, campus policies, faculty support, family support, surrounding environment, and academic skills. The C5.0 algorithm was selected for its higher accuracy and efficiency compared to earlier methods such as C4.5 and CART. The results show that the Surrounding Environment factor is the most dominant, followed by Student Motivation, Time Management, and Family Support. Evaluation of the model yielded excellent classification performance, achieving an accuracy of 95.31%, precision of 96.77%, recall of 93.75%, and an F1-score of 95.24%. These results indicate that the model effectively classifies students at risk of delay with strong predictive reliability. The findings provide insights for universities to develop targeted strategies to enhance student motivation, improve time management, and create a more supportive academic environment. In conclusion, the C5.0 algorithm demonstrates a strong capability to identify dominant delay factors and supports data-driven decision-making in academic management.Delays in completing final projects are a common problem faced by students and can lead to delayed graduation, increased study load, and reduced readiness to enter the workforce. This study uses a quantitative predictive approach to analyze the factors influencing delays in completing student final projects by applying the C5.0 Decision Tree classification algorithm. Data were collected through a Likert-scale questionnaire from 204 students of the Faculty of Engineering, University of 17 August 1945 Surabaya, who graduated between 2019 and 2021. The analyzed factors include time management, student motivation, campus policies, faculty support, family support, surrounding environment, and academic skills. The C5.0 algorithm was selected for its higher accuracy and efficiency compared to earlier methods such as C4.5 and CART. The results show that the Surrounding Environment factor is the most dominant, followed by Student Motivation, Time Management, and Family Support. Evaluation of the model yielded excellent classification performance, achieving an accuracy of 95.31%, precision of 96.77%, recall of 93.75%, and an F1-score of 95.24%. These results indicate that the model effectively classifies students at risk of delay with strong predictive reliability. The findings provide insights for universities to develop targeted strategies to enhance student motivation, improve time management, and create a more supportive academic environment. In conclusion, the C5.0 algorithm demonstrates a strong capability to identify dominant delay factors and supports data-driven decision-making in academic management
Comparison of Light Gradient Boosting Machine, eXtreme Gradient Boosting, and CatBoost with Balancing and Hyperparameter Tuning for Hypertension Risk Prediction on Clinical Dataset
No full textHypertension is a long-lasting condition that is highly prevalent and significantly contributes to cardiovascular issues, making early identification a crucial preventive action. This research evaluates the efficacy of three boosting algorithms, eXtreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LGBM), and CatBoost in forecasting hypertension risk. A publicly accessible dataset consisting of 4,363 samples was employed, followed by data preprocessing, feature selection through a voting method that integrates Boruta, Recursive Feature Elimination (RFE), and SelectKBest, as well as addressing class imbalance using the Synthetic Minority Over-sampling Technique (SMOTE) and ADASYN (Adaptive Synthetic Sampling Approach). The models were additionally fine-tuned through hyperparameter optimization using GridSearchCV and Repeated Stratified K-Fold Cross Validation. The evaluation results demonstrate that all three algorithms exhibited strong predictive capabilities, with CatBoost leading the way, achieving an accuracy of 0.992, precision of 0.992, recall of 0.992, F1-score of 0.992, and ROC-AUC of 0.9987. Analyzing the confusion matrix further validated that CatBoost had the lowest number of misclassifications when compared to XGBoost and LGBM. Additionally, the use of SHapley Additive exPlanations (SHAP) for model interpretability highlighted that the key factors influencing the prediction of hypertension risk are blood pressure, body mass index (BMI), overall physical activity, waist circumference, triglyceride levels, age, and LDL cholesterol levels, aligning with established medical knowledge. To facilitate real-world use, the top-performing model was implemented into a user-friendly website interface, allowing users to predict their hypertension risk interactively. These findings illustrate that boosting algorithms, especially CatBoost, offer an accurate, dependable, and interpretable machine learning method for creating hypertension risk prediction systems.Hypertension is a long-lasting condition that is highly prevalent and significantly contributes to cardiovascular issues, making early identification a crucial preventive action. This research evaluates the efficacy of three boosting algorithms, eXtreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LGBM), and CatBoost in forecasting hypertension risk. A publicly accessible dataset consisting of 4,363 samples was employed, followed by data preprocessing, feature selection through a voting method that integrates Boruta, Recursive Feature Elimination (RFE), and SelectKBest, as well as addressing class imbalance using the Synthetic Minority Over-sampling Technique (SMOTE) and ADASYN (Adaptive Synthetic Sampling Approach). The models were additionally fine-tuned through hyperparameter optimization using GridSearchCV and Repeated Stratified K-Fold Cross Validation. The evaluation results demonstrate that all three algorithms exhibited strong predictive capabilities, with CatBoost leading the way, achieving an accuracy of 0.992, precision of 0.992, recall of 0.992, F1-score of 0.992, and ROC-AUC of 0.9987. Analyzing the confusion matrix further validated that CatBoost had the lowest number of misclassifications when compared to XGBoost and LGBM. Additionally, the use of SHapley Additive exPlanations (SHAP) for model interpretability highlighted that the key factors influencing the prediction of hypertension risk are blood pressure, body mass index (BMI), overall physical activity, waist circumference, triglyceride levels, age, and LDL cholesterol levels, aligning with established medical knowledge. To facilitate real-world use, the top-performing model was implemented into a user-friendly website interface, allowing users to predict their hypertension risk interactively. These findings illustrate that boosting algorithms, especially CatBoost, offer an accurate, dependable, and interpretable machine learning method for creating hypertension risk prediction systems
Optimizing Decision Tree and Random Forest with Grid Search and SMOTE for Malware Classification on IoT Network Traffic
No full textThe rapid growth of the Internet of Things (IoT) has increased the risk of malware attacks, posing serious threats especially to micro, small, and medium enterprises (MSMEs) that often lack sufficient cybersecurity resources. This study aims to optimize Decision Tree (DT) and Random Forest (RF) classifiers using Grid Search, while addressing the class imbalance problem through the Synthetic Minority Oversampling Technique (SMOTE). The Security Attacks Malware IoT Networks dataset with five classes (Benign, Malware, DDoS, Brute Force, Scanning) was used and divided into training and testing sets with stratified 80:20 split. Experimental results show that DT achieved 67.3% accuracy with a macro F1-score of 42.9%, while RF achieved 70.7% accuracy but a very low macro F1-score of 21.4%, indicating bias toward the majority class despite balancing. Boosting methods provided stronger baselines, with XGBoost reaching 87.0% accuracy and 66.7% F1-score, while LightGBM achieved 85.6% accuracy and 64.4% F1-score. ROC curves and confusion matrices confirmed that boosting methods were more balanced in recognizing minority classes. In terms of efficiency, DT required the shortest training time (8 seconds), while LightGBM provided the best trade-off between accuracy and computational cost (26 seconds). Paired t-tests further confirmed that performance differences between DT and RF were not significant, while boosting methods significantly outperformed RF. Overall, optimizing DT and RF with Grid Search and SMOTE enhances their performance, but boosting methods remain more robust for malware detection in IoT traffic. These findings provide practical insights for MSMEs in balancing accuracy and efficiency when deploying intrusion detection systems.The rapid growth of the Internet of Things (IoT) has increased the risk of malware attacks, posing serious threats especially to micro, small, and medium enterprises (MSMEs) that often lack sufficient cybersecurity resources. This study aims to optimize Decision Tree (DT) and Random Forest (RF) classifiers using Grid Search, while addressing the class imbalance problem through the Synthetic Minority Oversampling Technique (SMOTE). The Security Attacks Malware IoT Networks dataset with five classes (Benign, Malware, DDoS, Brute Force, Scanning) was used and divided into training and testing sets with stratified 80:20 split. Experimental results show that DT achieved 67.3% accuracy with a macro F1-score of 42.9%, while RF achieved 70.7% accuracy but a very low macro F1-score of 21.4%, indicating bias toward the majority class despite balancing. Boosting methods provided stronger baselines, with XGBoost reaching 87.0% accuracy and 66.7% F1-score, while LightGBM achieved 85.6% accuracy and 64.4% F1-score. ROC curves and confusion matrices confirmed that boosting methods were more balanced in recognizing minority classes. In terms of efficiency, DT required the shortest training time (8 seconds), while LightGBM provided the best trade-off between accuracy and computational cost (26 seconds). Paired t-tests further confirmed that performance differences between DT and RF were not significant, while boosting methods significantly outperformed RF. Overall, optimizing DT and RF with Grid Search and SMOTE enhances their performance, but boosting methods remain more robust for malware detection in IoT traffic. These findings provide practical insights for MSMEs in balancing accuracy and efficiency when deploying intrusion detection systems
IoT-based Soil Nutrient Monitoring and Control Using Fuzzy Logic and Multi-Modal Sensor Integration
No full textThe decline in soil quality due to inappropriate agricultural practices has become one of the main factors contributing to reduced agricultural productivity. The primary focus of this research is on monitoring and controlling soil nutrient quality, particularly in clay soil used for chili cultivation. This study aims to develop an Internet of Things (IoT)-based monitoring system integrated with multi-modal sensors and fuzzy logic algorithms. The system is designed to support precision agriculture by enabling automated decision-making based on real-time environmental data. The research uses an experimental approach, involving the design of a system based on the ESP32 microcontroller, sensor data processing using the Mamdani fuzzy algorithm, and integration with the Blynk platform for remote monitoring and control. The system responds to changes in environmental conditions to determine optimal timing for irrigation and liquid nutrient application adaptively. The test results show that the system achieved a classification accuracy of 84% and an average F1-score of 88.5%, indicating its effectiveness in handling continuous and uncertain sensor data. Evaluation of the fuzzy logic performance revealed a 75.8% success rate in irrigation control and 99.8% accuracy in nutrient delivery, demonstrating the system’s ability to respond accurately and efficiently to actual soil and environmental conditions. With its stable, adaptive, and resource-efficient performance, this system has the potential to become a practical solution for automating irrigation and fertilization processes in support of technology-driven and sustainable agriculture
Integration of Multi-Modal Sensors in Aquaponic Farming for IoT-Ready Based on ESP32 and Raspberry Pi Hybrid Platform
No full textThis study aims to design and implement a smart agriculture system that integrates multi-modal sensors with an aquaponic farming platform, utilizing Raspberry Pi and ESP32 microcontrollers. The integration approach adopts embedded system-based control to connect and coordinate all multi-modal sensor components within the smart aquaponic environment. The primary function of the multi-modal sensors is to acquire comprehensive environmental and operational data from the aquaponic system through instrumentation-based measurement techniques. These data are intended to be further integrated with Internet of Things (IoT) technology using ESP32 and Raspberry Pi as control units. In this study, the integration of the Raspberry Pi and ESP32 platforms demonstrates superior performance compared to a single platform, as it combines a microcontroller capable of reading analog sensor data and transmitting it to the Raspberry Pi, which subsequently functions as the central data processing unit. Experimental results confirm that all multi-modal sensor devices operate reliably when interfaced with the ESP32 and Raspberry Pi, producing accurate data streams that can be utilized in future implementations of IoT-based control systems
Integrating IndoBERTweet and GRU for Opinion Classification on X Towards Public Transportation in Jakarta
No full textJakarta, the capital of Indonesia, faces persistent challenges with its public transportation system due to rapid urbanization, increased use of private vehicles, and poor service quality. While social media platforms such as X (formerly Twitter) offer valuable insights into public opinion, their unstructured nature complicates analysis. This study uses deep learning models to categorize user sentiments into six labels that cover positive and negative aspects of comfort, safety, and punctuality. The results show that IndoBERTweet achieved the highest performance, with 95.43% accuracy and a macro F1-score of 0.9545. It also required the shortest training time, at six minutes and 30 seconds. IndoBERTweet+GRU followed closely behind with an accuracy of 94.62% and a macro F1-score of 0.9460 in six minutes and 50 seconds. This shows that adding a GRU layer provides competitive results, but does not surpass the baseline model. Error analysis revealed that, while the models performed well with explicit sentiments, the models struggled with implicit expressions, such as sarcasm and mixed opinions. These results demonstrate the potential of sentiment analysis in real-time monitoring systems, which could help policymakers identify urgent issues and support data-driven improvements in Jakarta’s urban transportation services
Comparison of Support Vector Machine and Random Forest Algorithms in Sentiment Analysis of the JMO Mobile Application
No full textJMO Mobile is a digital service application that enables the public to access employment-related information and benefits. User reviews serve as a valuable resource for evaluating service quality, yet systematic sentiment analysis on this application remains limited. This study aims to classify the sentiment of user reviews and compare the performance of Support Vector Machine (SVM) and Random Forest (RF) algorithms. A total of 41,673 reviews were collected through web scraping, then preprocessed through text cleaning, tokenization, stopword removal, stemming, and feature extraction using TF-IDF. The reviews were categorized into positive, negative, and neutral sentiments, and divided into training and testing datasets with an 80:20 ratio. The choice of SVM and RF was based on their proven effectiveness in text classification tasks, with SVM excelling in handling high-dimensional data and RF recognized for its stability in producing reliable results. Model evaluation was conducted using accuracy as the primary metric. The findings indicate that Random Forest achieved an accuracy of 86.15 percent, slightly outperforming SVM at 86.06 percent. While SVM showed superior performance in identifying positive sentiment, Random Forest demonstrated greater consistency across classifications. Overall, Random Forest is considered more suitable for sentiment analysis of public service application reviews. This study contributes an automated approach to understanding user perceptions and offers a reference for selecting classification algorithms in similar cases
Anemia Classification with Clinical Feature Engineering and SHAP Interpretation
No full textAnemia is a global health issue that has a significant impact on quality of life and productivity. Early and accurate detection is essential to prevent more serious complications. This study aims to develop an anemia classification model based on machine learning technology using the XGBoost algorithm, as well as compare its performance with Logistic Regression and Random Forest methods. The dataset used in this study was obtained from the Kaggle platform, consisting of 1,421 samples and six clinical attributes, namely Gender, Hemoglobin (HGB), Mean Corpuscular Hemoglobin (MCH), Mean Corpuscular Hemoglobin Concentration (MCHC), Mean Corpuscular Volume (MCV), Result. During the feature engineering process, the derived feature of the hemoglobin-to-MCV ratio (Hb/MCV) was added, which is medically relevant in distinguishing types of anemia. Evaluation results showed that XGBoost and Random Forest achieved an accuracy rate and F1-Score of 100%, while Logistic Regression achieved a rate of 98.9%. XGBoost was selected as the primary model due to its efficient computational capabilities and support for interpretation using SHAP (SHapley Additive exPlanations). SHAP visualization revealed that the Hb/MCV ratio and hemoglobin were the most influential features in classification. This model has the potential to be used as a decision support system for automated anemia screening and can be further integrated into clinical systems