Jurnal Politeknik Negeri Batam (PoliBatam)
Not a member yet
3001 research outputs found
Sort by
Harmonic Analysis and Performance of a VFD-Controlled Single-Phase Motor
No full textThe application of Variable Frequency Drives (VFDs) in single-phase induction motors is increasingly common, although it may introduce harmonic distortion that affects power quality. This study analyzes harmonic characteristics and motor performance under variable frequency operation. Measurements of voltage and current harmonics were carried out using a Fluke 434 Power Quality Analyzer at VFD frequencies ranging from 10 to 50 Hz. The results indicate that voltage harmonic distortion (THD-V) decreased from 5,0% to 3,2% on L1 and from 5,8% to 4,1% on the neutral conductor, remaining within IEEE 519-2014 limits. In contrast, current distortion (TDD) reached 71,8% at 50 Hz but decreased to 23% at 20 Hz due to network impedance effects. Dominant harmonics were observed at H3–H13, with triplen harmonics prevailing in the neutral line. Overall, VFD-based frequency control improves motor efficiency and voltage quality but requires additional filtering to limit current distortion within standards
Penciptaan Video Klip Melayu Menggunakan Teknik Camera Follow Untuk Visualisasi Museum Raja Ali Haji
No full textMusic videos are an audiovisual art that enhances multisensory experiences while acting as a medium of cultural information and communication, such as Malay culture. In this creation, cinematographic techniques are employed, particularly the camera follow technique, to create an exploratory and immersive experience. This approach is implemented in the Malay music video “Nurlela” composed by Asbon Madjid and Bing Slamet, arranged in progressive Malay pop. The study aims to determine the effectiveness of creating a Malay music video using the camera follow technique to visualize the Raja Ali Haji Museum in presenting its historical collections and visitor regulations. The production process applies the Multimedia Development Life Cycle (MDLC), consisting of six stages: concept, design, material collecting, assembly, testing, and distribution. The testing stage uses the EPIC Model, which includes four dimensions: empathy, persuasion, impact, and communication.Video klip merupakan seni audio visual yang berperan dalam meningkatkan pengalaman multisensori pengunjung sekaligus menjadi sarana informasi dan komunikasi budaya, seperti budaya Melayu. Dalam penciptaannya, penggunaan teknik sinematografi, salah satunya teknik camera follow untuk menciptakan kesan eksploratif dan imersif. Pendekatan ini diterapkan dalam video klip Melayu pada lagu “Nurlela” ciptaan Asbon Madjid & Bing Slamet yang diaransemen dalam pop progresif Melayu. Penelitian ini bertujuan untuk mengetahui efektivitas penciptaan video klip Melayu menggunakan teknik camera follow untuk memvisualisasikan Museum Raja Ali Haji dalam menampilkan koleksi bersejarah dan tata tertib Museum Raja Ali Haji. Proses penciptaan menggunakan Multimedia Development Life (MDLC) yang terdiri dari enam tahapan utama, yaitu concept, design, material collecting, assembly, testing dan distribution. Pada bagian testing menggunakan metode EPIC Model yang mencakup empat dimensi, yaitu empathy, persuasion, impact dan communication
Z-Score Based Initialization for K-Medoids Clustering: Application on QSAR Toxicity Data
No full textThe efficiency of clustering algorithms significantly depends on the initialization quality, especially in unsupervised learning applied to complex datasets. This study introduces an enhanced K-Medoids clustering approach using Z-Score-based medoid initialization to improve convergence speed and cluster validity. The method was evaluated using the QSAR Fish Toxicity dataset, consisting of 908 instances and seven numerical features. Initial medoids were selected based on standardized Z-Score values, resulting in a substantial reduction in convergence time from an average of 6 iterations to just 2. Clustering performance was assessed using three internal validation metrics: Davies-Bouldin Index (DBI), Silhouette Coefficient (SC), and Calinski-Harabasz Index (CHI). The DBI score decreased from 1.7328 to 0.8768, indicating improved cluster compactness and separation. In parallel, the SC increased from 0.327 to 0.619, and the CHI rose from 214.75 to 562.43, confirming more coherent and well-separated clusters. These results demonstrate that Z-Score-based initialization significantly boosts the robustness of K-Medoids, offering a simple yet effective strategy for unsupervised partitioning, particularly in toxicological and biochemical data analysis
e
No full textAsthma prediction demands architectures capable of capturing multifactorial interactions among demographic, clinical, and environmental determinants. This study establishes Random Forest (RF) as the optimal solution through rigorous comparison with Logistic Regression (LR) and Support Vector Machines (SVM) on a 10,000-patient cohort. RF achieved performance: 99.55% accuracy, 100% precision, 98.19% recall, and exceptional stability (σ=0.0019 CV) surpassing SVM by 6.86% recall, preventing 167 missed diagnoses per 10,000 cases. Hereditary factors dominated feature importance (Gini=0.20), generating 18.7% greater node purity reduction than BMI, while the paradoxical "No Allergies" signal (3.726) revealed non-atopic phenotypes. Critically, sparse linear correlations (94% |r|<0.02) contrasted with RF’s capture of nonlinear thresholds like sedentarism (2.243) > smoking impact. Clinical implementation requires: (1) threshold calibration (θ=0.3) achieving >99% recall, (2) monthly false-negative audits mitigating 24.33% prevalence skew, and (3) dimensionality reduction eliminating 3.256 features. RF’s capacity to resolve hereditary-environmental interactions establishes a new paradigm for asthma risk stratification
Application of Feature Selection and Comparative Analysis of Machine Learning Models for Rainfall Prediction in Jakarta
No full textAccurate rainfall prediction plays a vital role in reducing disaster risks and supporting public preparedness, particularly in Jakarta where dense population and frequent floods cause serious economic and social impacts. In this study, weather data from the Kemayoran Meteorological Station covering 2004–2023 were analyzed to build rainfall prediction models using machine learning. Three classification algorithms were compared: Logistic Regression, Decision Tree, and Random Forest, selected to represent linear, non-linear, and ensemble approaches. Feature selection was applied using Recursive Feature Elimination (RFE) to identify the most relevant predictors. The models were evaluated using 5-fold cross-validation with metrics including Accuracy, Precision, Recall, F1 Score, ROC AUC, and Cohen’s Kappa. The results indicate that Random Forest achieved the best overall performance with Accuracy of 0.7622, Precision around 0.70, Recall up to 0.63, F1 Score about 0.65, ROC AUC ranging from 0.8044 to 0.8171, and Cohen’s Kappa near 0.48. Logistic Regression also performed competitively with Accuracy of 0.7648, ROC AUC of 0.829, and Kappa of 0.49, while Decision Tree showed lower results with Accuracy of 0.6890 and ROC AUC of 0.6636. The RFE process successfully reduced 18 meteorological attributes to 5 influential features, mainly temperature and relative humidity, which were dominant in distinguishing rainfall events. These findings demonstrate that both Random Forest and Logistic Regression outperform Decision Tree, and Random Forest with RFE can be recommended as the most robust model for rainfall prediction in Jakarta.Accurate rainfall prediction plays a vital role in reducing disaster risks and supporting public preparedness, particularly in Jakarta where dense population and frequent floods cause serious economic and social impacts. In this study, weather data from the Kemayoran Meteorological Station covering 2004–2023 were analyzed to build rainfall prediction models using machine learning. Three classification algorithms were compared: Logistic Regression, Decision Tree, and Random Forest, selected to represent linear, non-linear, and ensemble approaches. Feature selection was applied using Recursive Feature Elimination (RFE) to identify the most relevant predictors. The models were evaluated using 5-fold cross-validation with metrics including Accuracy, Precision, Recall, F1 Score, ROC AUC, and Cohen’s Kappa. The results indicate that Random Forest achieved the best overall performance with Accuracy of 0.7622, Precision around 0.70, Recall up to 0.63, F1 Score about 0.65, ROC AUC ranging from 0.8044 to 0.8171, and Cohen’s Kappa near 0.48. Logistic Regression also performed competitively with Accuracy of 0.7648, ROC AUC of 0.829, and Kappa of 0.49, while Decision Tree showed lower results with Accuracy of 0.6890 and ROC AUC of 0.6636. The RFE process successfully reduced 18 meteorological attributes to 5 influential features, mainly temperature and relative humidity, which were dominant in distinguishing rainfall events. These findings demonstrate that both Random Forest and Logistic Regression outperform Decision Tree, and Random Forest with RFE can be recommended as the most robust model for rainfall prediction in Jakarta
Frontend Implementation on EngVenture Application at IntSys Research Lab
No full textIn today\u27s digital era, the use of mobile applications for English learning is increasingly popular as an alternative to self-study. However, many available applications still lack the ability to provide an interactive, adaptive, and enjoyable learning experience, and do not provide integrated proficiency measurement features such as the TOEFL test. This research focuses on the frontend implementation of the EngVenture application, an English learning platform developed at IntSys Research Lab using the Rapid Application Development (RAD) method. This application is designed to address these issues by integrating gamification elements and a TOEFL-like practice test system to increase engagement and measure user progress. Data were collected through literature studies and questionnaires distributed to 100 respondents from various educational levels. The results showed that 82% of respondents needed a fun learning medium, 92% wanted a TOEFL test feature, and 88% were interested in the gamification feature. The application was developed using Flutter and Dart, with a responsive UI/UX design and real-time feedback features. System testing was conducted using two methods: black-box User Acceptance Testing (UAT) to assess functionality, and a System Usability Scale (SUS) to measure the application\u27s usability. Test results showed that all features functioned well, with an average SUS score of 84.25, which falls into the Acceptable (Grade B+, Excellent) category. These results demonstrate that EngVenture meets user needs in terms of functionality and usability, and has the potential to become an interactive and effective English language learning tool
Comparative Analysis of 1D CNN Architectures for Guitar Chord Recognition from Static Hand Landmarks
No full textVision-based guitar chord recognition offers a promising alternative to traditional audio-driven methods, particularly for silent practice, classroom environments, and interactive learning applications. While existing research predominantly relies on full-frame image analysis using 2D convolutional networks, the use of structured hand landmarks remains underexplored despite their advantages in robustness and computational efficiency. This study presents a comprehensive comparative analysis of three one-dimensional convolutional neural network architectures—CNN-1D, ResNet-1D, and Inception-1D—for classifying seven guitar chord types using 63-dimensional static hand-landmark vectors extracted via MediaPipe Hands. The methodology encompasses extensive dataset preprocessing, targeted landmark augmentation, Bayesian hyperparameter optimization, and stratified 5-fold cross-validation. Results show that CNN-1D achieves the highest mean accuracy (97.61%), outperforming both ResNet-1D and Inception-1D, with statistical tests confirming significant improvements over ResNet-1D. Robustness experiments further demonstrate that CNN-1D maintains superior resilience under Gaussian noise, landmark occlusion, and geometric scaling. Additionally, CNN-1D provides the fastest inference and most stable computational performance, making it highly suitable for real-time or mobile deployment. These findings highlight that, for structured and low-dimensional landmark data, simpler convolutional architectures outperform deeper or multi-branch designs, offering an efficient and reliable solution for vision-based guitar chord recognition
Optimization of Application Deployment Architecture in Container Orchestration
Get PDFContainer orchestration has become a widely adopted standard for application deployment among medium to large-scale organizations. Docker Swarm is one of the popular container orchestration tools due to its relatively simple configuration. However, if the Docker Swarm cluster architecture is not properly designed, the goal of container orchestration, which is availability, cannot be achieved optimally. Challenges such as centralized traffic on a single node and service dependency on a single node are critical issues that need to be addressed. This study proposes solutions through an experimental approach involving the design, implementation, testing, and evaluation of a Docker Swarm cluster architecture to address these challenges. The results of this study demonstrate that the proposed architecture successfully resolves these issues. Traffic can be distributed more evenly across all nodes. When only one node is available, 5 out of 10 requests can be handled with a response latency of 197.4 ms. With two nodes available, the number of requests handled increases to 7 out of 10, with a response latency of 534.86 ms. The greater the number of available nodes, the more requests can be successfully processed. Services also become more flexible, and capable of running on any node, while offering additional benefits such as dual load balancing through DNS-based load balancing and the default load balancing provided by Docker Swarm\u27s routing mesh. However, limitations such as the need for more complex adjustments and configurations should be considered, especially when implementing this architecture in on-premise environments, to ensure the best adoption and results
Implementation of SVM Algorithm to Predict Song Popularity based on Sentiment Analysis of Lyrics
Get PDFIndependent musicians face significant challenges in enhancing the visibility and appeal of their work amid intense competition on music streaming platforms. Although numerous studies have been conducted to analyze and predict song popularity, most of them focus on English-language songs. This creates a research gap for Indonesian-language songs, particularly in the context of predicting popularity based on lyrics. The dataset used includes 652 Indonesian songs from 2017 to 2024. The research methodology includes data pre-processing, feature extraction using TF-IDF, handling data imbalance with SMOTE, implementing SVM, and model optimization. The results show an improvement in model accuracy from 84% to 89% after parameter optimization using GridSearchCV. In the model evaluation with 5-fold cross-validation, an average accuracy of 86.19% with a standard deviation of 0.90% was obtained. Precision, Recall, and F1-score metrics for the Less Popular class are 0.98, 0.85, and 0.91; for the Moderately Popular class, 0.79, 0.95, and 0.86; and for the Very Popular class, 0.92, 0.86, and 0.89. The implementation of the model in a Streamlit application allows for the prediction of song popularity based on lyrics, providing valuable insights for musicians in choosing word choices that can potentially increase the popularity of their songs.Independent musicians face significant challenges in enhancing the visibility and appeal of their work amid intense competition on music streaming platforms. Although numerous studies have been conducted to analyze and predict song popularity, most of them focus on English-language songs. This creates a research gap for Indonesian-language songs, particularly in the context of predicting popularity based on lyrics. The dataset used includes 652 Indonesian songs from 2017 to 2024. The research methodology includes data pre-processing, feature extraction using TF-IDF, handling data imbalance with SMOTE, implementing SVM, and model optimization. The results show an improvement in model accuracy from 84% to 89% after parameter optimization using GridSearchCV. In the model evaluation with 5-fold cross-validation, an average accuracy of 86.19% with a standard deviation of 0.90% was obtained. Precision, Recall, and F1-score metrics for the Less Popular class are 0.98, 0.85, and 0.91; for the Moderately Popular class, 0.79, 0.95, and 0.86; and for the Very Popular class, 0.92, 0.86, and 0.89. The implementation of the model in a Streamlit application allows for the prediction of song popularity based on lyrics, providing valuable insights for musicians in choosing word choices that can potentially increase the popularity of their songs
Long Short-Term Memory as a Rainfall Forecasting Model for Bogor City in 2025-2026
Get PDFIndonesia is a country with a tropical climate that has unique and changing weather patterns. Accurate rainfall prediction can help local governments, farmers, and the broader community plan activities that depend on rainfall patterns. This research aims to develop a rainfall prediction model for Bogor City using past rainfall data in Bogor City, which is known as an area with high rainfall levels and dynamic rainfall patterns. The analysis utilizes rainfall data recorded by the JAXA satellite from January 1, 2014, to December 31, 2024. The prediction method implemented in this research is the long short-term memory (LSTM). The LSTM modelling process evaluates various models by comparing RMSE, MAE, and correlation values through expanding window cross-validation, selecting the model with the lowest average RMSE and MAE with the highest correlation as the optimal choice. The best-performing model was achieved with 25 epochs and a batch size of 1, resulting in an average RMSE of 56.3340, MAE of 35.5223, and correlation of 0.3209. This best-performing model is then employed to predict rainfall for the next two years. The results show significant daily variations in the predicted rainfall but can capture existing seasonal patterns