Jurnal Politeknik Negeri Batam (PoliBatam)
Not a member yet
3001 research outputs found
Sort by
A Improving House Price Clustering Results with K-means through the Implementation of One-hot Encoding Pre-processing Technique
No full textBasic human needs include a house that serves as a place to live and a shelter from everything. In Indonesia, owning a house is still a challenging aspect due to its high price. Information on house prices is needed for prospective buyers or consumers, so that buyers can adjust their needs and finances, and for producers or sellers it is used as a way to determine the segmentation of targeted market groups. House prices are influenced by several factors including, building area, number of bedrooms, number of bathrooms, location, condition and the presence of a garage. This research aims to improve the quality of house price clustering with K-means and the application of one-hot encoding in the data pre-processing process in representing categorical data. The dataset used has two types of data, namely numeric and categorical. The cluster evaluation is based on the silhouette score matrix and the determination of k is based on the elbow graph. The results showed an increase in the silhouette score value after applying one-hot encoding 0.15 which was previously 0.09, with the number of k = 3. The 0.15 matrix result is relatively low, which is caused by the overlap of house price values in the dataset, but it has been shown that one-hot encoding can represent categorical data well in the data pre-processing process so that the data can be processed with the k-means algorithm.Basic human needs include a house that serves as a place to live and a shelter from everything. In Indonesia, owning a house is still a challenging aspect due to its high price. Information on house prices is needed for prospective buyers or consumers, so that buyers can adjust their needs and finances, and for producers or sellers it is used as a way to determine the segmentation of targeted market groups. House prices are influenced by several factors including, building area, number of bedrooms, number of bathrooms, location, condition and the presence of a garage. This research aims to improve the quality of house price clustering with K-means and the application of one-hot encoding in the data pre-processing process in representing categorical data. The dataset used has two types of data, namely numeric and categorical. The cluster evaluation is based on the silhouette score matrix and the determination of k is based on the elbow graph. The results showed an increase in the silhouette score value after applying one-hot encoding 0.15 which was previously 0.09, with the number of k = 3. The 0.15 matrix result is relatively low, which is caused by the overlap of house price values in the dataset, but it has been shown that one-hot encoding can represent categorical data well in the data pre-processing process so that the data can be processed with the k-means algorithm
Enchancing Enhancing Medical Named Entity Recognition with Ensemble Voting of BERT-Based Models on BC5CDR
No full textThe rapid development in biotechnology and medical research has resulted in a large amount of scientific literature containing critical information about various medical entities. However, the primary challenge in managing this data is the vast volume of unstructured text, which requires Natural Language Processing (NLP) techniques for automatic information extraction. One of the main applications in NLP is Named Entity Recognition (NER), which aims to identify important entities in the text, such as disease names, drugs, and proteins. This study aims to enhance the performance of medical Named Entity Recognition (NER) by applying ensemble Voting to three BERT-based models: BioBERT, TinyBERT, and ClinicalBERT. The results show that the ensemble voting technique provides the best performance in medical entity extraction, with improvements in precision (0.9494), recall (0.9483), and F1-score (0.9488) compared to individual models, especially when handling less common medical entities. This approach is expected to contribute to the development of automated systems for analyzing and searching information in medical literature.The rapid development in biotechnology and medical research has resulted in a large amount of scientific literature containing critical information about various medical entities. However, the primary challenge in managing this data is the vast volume of unstructured text, which requires Natural Language Processing (NLP) techniques for automatic information extraction. One of the main applications in NLP is Named Entity Recognition (NER), which aims to identify important entities in the text, such as disease names, drugs, and proteins. This study aims to enhance the performance of medical Named Entity Recognition (NER) by applying ensemble Voting to three BERT-based models: BioBERT, TinyBERT, and ClinicalBERT. The results show that the ensemble voting technique provides the best performance in medical entity extraction, with improvements in precision (0.9494), recall (0.9483), and F1-score (0.9488) compared to individual models, especially when handling less common medical entities. This approach is expected to contribute to the development of automated systems for analyzing and searching information in medical literature
Enhancing Negative Film Colorization through Systematic CycleGAN Architectural Modifications: A Comprehensive Analysis of Generator and Discriminator Performance
No full textThis research addresses the urgent need for deep learning-based negative film colorization technology through systematic modifications to the CycleGAN architecture. Unlike conventional approaches that focus on colorizing black-and- white images, this study targets the conversion of digitized negative film images, which present unique challenges such as color inversion and detail restoration. The dataset consists of 500 negative images (train A), 500 unpaired color images (train B), as well as 5 negative images and 5 color images for testing purposes. The entire dataset was obtained from personal scanning efforts. 19 architectural modifications were proposed and tested individually, without simultaneously implementing all changes. The primary focus was on developing network structures, without utilizing external evaluation metrics such as SSIM, PSNR, or FID. Modifications included the addition of residual blocks, alterations in filter quantities, activation functions, and inter-layer connections. The Evaluation was conducted qualitatively and based on generator and discriminator loss values. The most optimal modification (Modification 4) demonstrated significant loss reduction (G: 2.39–4.07, F: 2.82– 3.66; D_X: 0.36–0.93, D_Y: 0.15–1.39), yielding more accurate and aesthetically pleasing color images compared to the baseline architecture. The fundamental cycle consistency loss structure was maintained to ensure the unpaired training capability remained intact. This research demonstrates that careful architectural modifications can significantly enhance negative colorization results, while simultaneously creating opportunities for the future development of deep learning-based digital image restoration technologies
REFRAMING SUSTAINABILITY A PERFORMANCE FRAMEWORK FOR PUBLIC EDUCATION AGENCY IN INDONESIA
No full textDespite Indonesia’s commitment as a member of the United Nations to promote sustainable development, its implementation remains uneven across various sectors, particularly in education. This study aims to develop a comprehensive framework for measuring sustainability performance within public service agency (PSA) in the educational sector. Employing a mixed-methods approach, data were collected through interviews, observations, and document analysis. The findings indicate that educational PSAs have not yet internalized sustainability principles in an integrated manner. To address this gap, the study proposes a novel sustainability performance measurement framework that has been empirically tested through a case study of a civil service college. Unlike existing frameworks, this model is grounded in the ECON-ESG dimensions—comprising economic, environmental, social, and governance aspects—and is refined through expert insights and relevant regulatory adjustments. This framework offers a practical and contextualized tool for evaluating sustainability performance in Indonesia’s educational public service institutions.Meskipun Indonesia telah menyatakan komitmennya sebagai anggota Perserikatan Bangsa-Bangsa untuk mendorong pembangunan berkelanjutan, implementasinya masih belum merata di berbagai sektor, khususnya sektor pendidikan. Penelitian ini bertujuan untuk mengembangkan kerangka kerja komprehensif guna mengukur kinerja keberlanjutan pada lembaga pelayanan publik (BLU) di sektor pendidikan. Dengan menggunakan pendekatan metode campuran, data dikumpulkan melalui wawancara, observasi, dan analisis dokumen. Temuan penelitian menunjukkan bahwa BLU di sektor pendidikan belum menginternalisasi prinsip-prinsip keberlanjutan secara terpadu. Untuk menjawab permasalahan tersebut, studi ini menawarkan sebuah kerangka pengukuran kinerja keberlanjutan yang telah diuji secara empiris melalui studi kasus pada sebuah perguruan tinggi kedinasan. Berbeda dengan kerangka yang telah ada, model ini berbasis pada dimensi ECON-ESG—yang mencakup aspek ekonomi, lingkungan, sosial, dan tata kelola—dan disusun melalui penyesuaian terhadap masukan para ahli serta regulasi yang relevan. Kerangka ini memberikan alat evaluasi yang praktis dan kontekstual bagi lembaga pelayanan publik pendidikan di Indonesia dalam mengukur kinerja keberlanjutannya
Implementation of Support Vector Machine for Classifying User Reviews on the Sentuh Tanahku Application
Full text linkUser reviews play a crucial role in the development of digital public service applications, as they reflect user satisfaction and service quality. This study aims to classify user reviews of the Sentuh Tanahku application into two sentiment categories, namely positive and negative, by applying the Support Vector Machine (SVM) algorithm. A total of 13,231 reviews obtained from Kaggle were processed through text preprocessing stages including case folding, tokenizing, stopword removal, and stemming. The TF-IDF technique was employed to convert text data into numerical vectors, followed by classification using SVM with hyperparameter tuning via RandomizedSearchCV. The evaluation results showed that the SVM model achieved an accuracy of 91% on training data and 84% on testing data. To assess its performance, the study compared SVM with baseline algorithms, namely Naïve Bayes and Logistic Regression. The comparison revealed that Logistic Regression and Naïve Bayes outperformed SVM with accuracy scores of 88.84% and 88.68%, respectively. Despite this, SVM remained competitive in maintaining balanced metrics across both classes. These findings highlight that algorithm performance in sentiment classification is highly influenced by the nature of the dataset. This study is expected to contribute as a reference for improving user opinion analysis methods in Indonesian-language public service applications.User reviews play a crucial role in the development of digital public service applications, as they reflect user satisfaction and service quality. This study aims to classify user reviews of the Sentuh Tanahku application into two sentiment categories, namely positive and negative, by applying the Support Vector Machine (SVM) algorithm. A total of 13,231 reviews obtained from Kaggle were processed through text preprocessing stages including case folding, tokenizing, stopword removal, and stemming. The TF-IDF technique was employed to convert text data into numerical vectors, followed by classification using SVM with hyperparameter tuning via RandomizedSearchCV. The evaluation results showed that the SVM model achieved an accuracy of 91% on training data and 84% on testing data. To assess its performance, the study compared SVM with baseline algorithms, namely Naïve Bayes and Logistic Regression. The comparison revealed that Logistic Regression and Naïve Bayes outperformed SVM with accuracy scores of 88.84% and 88.68%, respectively. Despite this, SVM remained competitive in maintaining balanced metrics across both classes. These findings highlight that algorithm performance in sentiment classification is highly influenced by the nature of the dataset. This study is expected to contribute as a reference for improving user opinion analysis methods in Indonesian-language public service applications
Image Classification of Red Dragon Fruit Ripeness Levels Using HSV Color Moments and the K-NN Algorithm
Full text linkAccurately determining the ripeness level of red dragon fruit (Hylocereus polyrhizus) is crucial for ensuring post-harvest quality and distribution efficiency. This study proposes a method for classifying red dragon fruit ripeness using color moment features in the HSV color space combined with the K-Nearest Neighbor (K-NN) algorithm. The dataset consists of 2,881 images of dragon fruit with a resolution of 800×800 pixels, categorized into three classes: ripe (886 images), unripe (1,241 images), and rotten (754 images). All images were captured under natural lighting conditions and underwent pre-processing to enhance color value consistency. Color features were extracted by calculating the mean, standard deviation, and skewness of the Hue, Saturation, and Value channels. The K-NN model was trained and tested on data randomly split in an 80:20 ratio. The testing results showed that the model achieved 100% accuracy in classifying the ripeness levels, demonstrating the effectiveness of this non-destructive method in distinguishing fruit ripeness. This approach holds strong potential to support efficient and consistent decision-making in the agricultural sector
Performance Comparison of Machine Learning Algorithms Using EfficientNetB0 Feature Extraction on Dental Disease Classification
Full text linkOral health conditions such as dental caries, calculus, gingivitis, and ulcers are prevalent globally and require accurate early detection to prevent further complications. Traditional diagnostic methods such as visual inspection and manual radiograph analysis often rely on subjective judgment, leading to inconsistencies, delayed treatment, and limited accessibility, particularly in underserved areas. This study proposes an intelligent classification framework for dental disease detection based on intraoral images. Deep features were extracted using EfficientNetB0, followed by classification through eleven machine learning algorithms, including SVM, XGBoost, and K-Nearest Neighbors. Preprocessing steps included image augmentation, SMOTE for class balancing, and feature normalization. Among all models, SVM achieved the highest accuracy of 92,9%, while XGBoost and LightGBM followed closely at 91.3%. Using K-Fold Cross Validation, KNN algorithm has an increasing value with accuracy of 91,24%. This indicate the KNN algorithm able to tackle generalization problem towards the classification. The results demonstrate that features extracted using CNNs, when classified using machine learning algorithms, can provide a scalable and effective alternative to conventional diagnostic practices. Hence, Machine Learning algorithms provide a promising result towards dental disease classification
Transformer-Based Deep Learning Model for Coffee Bean Classification
Full text linkCoffee is one of the most popular beverage commodities consumed worldwide. The process of selecting high-quality coffee beans plays a vital role in ensuring that the resulting coffee has superior taste and aroma. Over the years, various deep learning models based on Convolutional Neural Networks (CNN) have been developed and utilized to classify coffee bean images with impressive accuracy and performance. However, recent advancements in deep learning have introduced novel transformer-based architectures that show great promise for image classification tasks. By incorporating a self-attention module, transformer models excel at generating global context features within images. This ability demonstrate improved and more consistent performance compared to CNN-based models. This study focuses on training and evaluating transformer-based deep learning models specifically for the classification of coffee bean images. Experimental results demonstrate that transformer models, such as the Vision Transformer (ViT) and Swin Transformer, outperform traditional CNN-based models. Swin Transformer model achieves excellent on the coffee bean image classification task, with 95.13% Accuracy and 90.21% F1-Score, while ViT achieves 94.47% Accuracy and 88.93% F1-Score. It indicates their strong capability in accurately identifying and classifying different types of coffee beans. This suggests that transformer-based approaches could be a better alternative for coffee bean image classification tasks in the future
A Hybrid Data Science Framework for Forecasting Bitcoin Prices using Traditional and AI Models
Full text linkBitcoin, a highly volatile and decentralized digital asset, presents considerable challenges for accurate price forecasting. This study proposes an applied data science framework that compares traditional statistical approaches with modern Artificial Intelligence (AI)-based models to predict Bitcoin’s daily closing price. Using BTC-USD historical data from January 2020 to December 2024, we converted prices into Indonesian Rupiah (IDR) to increase local relevance. Our forecasting horizon is 30 days, based on a 60-day lookback window. We evaluate six models: Linear Regression, ARIMA, and Prophet as traditional techniques, alongside Random Forest, XGBoost, and Long Short-Term Memory (LSTM) networks as AI approaches. All models were trained using lag-based or sequence-based time series features and evaluated using MAE, RMSE, R², MAPE, and SMAPE. Results show that AI models, particularly LSTM and XGBoost, offer better performance in capturing short-term non-linear dynamics compared to traditional models. LSTM provides high accuracy, though with greater computational demand, while XGBoost strikes a balance between speed and precision. Prophet and ARIMA remain effective for quick and interpretable forecasts but struggle with abrupt trend shift common in cryptocurrency markets. In addition to performance metrics, we include a robustness analysis based on median absolute error and outlier detection to assess model stability under extreme variations. Visual analytics—including forecast curves, error distributions, and uncertainty bounds—help interpret and communicate model behavior. This comprehensive evaluation offers practical insights for investors, analysts, and fintech practitioners, and the pipeline can be extended to other volatile assets
The Influence of Audit Quality, Independent Board of Commissioners and Managerial Ownership on Profit Management (Case Study on Companies Included in LQ45 in the Indonesia Sharia Stock Index in 2019 -2023)
Full text linkThe primary goal of this research is to analyse the relationship between profit management and factors such managerial ownership, audit quality, and board of commissioners independence. Secondary sources are the backbone of this research. In particular, the LQ45 index members of the Indonesia sharia stock index for the years 2019"“2023, are the subject of this study. The data for this study came from a population of 180 individuals whose records were scanned by the company. A total of 95 participants participated in the survey, which used a purposive summarising technique. Using quantitative methodologies, this study tests theories about the impact of managerial ownership, audit quality, and an independent board of commissioners on profit management. The first estimate was supported by a sig value of 0.05 and a T value of 1.98638, indicating that the audit quality had a smaller impact on profit management than what was initially thought. As demonstrated by an A T value of -.1695 < 1.66515 and a sig value of 0.093 > 0.05, the independent board of commissioners did not significantly impact profit management. With a sig value of 0.118 > 0.05 and a T value of 1.1557 < 1.66515, partial managerial ownership does not significantly impact profit management. Both the individual and combined effects of audit quality, independent board of commissioners, and managerial ownership on profit management are statistically significant (F value of 3.140 < 2.70 and sig value of 0.029 < 0.05). There were additional variables that made up 93.6% of the total, even though audit quality, independent board of commissioners, and managerial ownership only made up 6.4%. According to the statistics given, profit management is significantly affected by audit quality, the independence of the board of commissioners, and managerial ownership