cover
Contact Name
Mesran
Contact Email
mesran.skom.mkom@gmail.com
Phone
-
Journal Mail Official
jurnal.bits@gmail.com
Editorial Address
-
Location
Kota medan,
Sumatera utara
INDONESIA
Building of Informatics, Technology and Science
ISSN : 26848910     EISSN : 26853310     DOI : -
Core Subject : Science,
Building of Informatics, Technology and Science (BITS) is an open access media in publishing scientific articles that contain the results of research in information technology and computers. Paper that enters this journal will be checked for plagiarism and peer-rewiew first to maintain its quality. This journal is managed by Forum Kerjasama Pendidikan Tinggi (FKPT) published 2 times a year in Juni and Desember. The existence of this journal is expected to develop research and make a real contribution in improving research resources in the field of information technology and computers.
Arjuna Subject : -
Articles 889 Documents
Implementasi Arsitektur MobileNetV2 dengan Metode Transfer Learning untuk Identifikasi Objek Wisata Religi Wijaya, Nabilah Putri; Christioko, Bernadus Very
Building of Informatics, Technology and Science (BITS) Vol 7 No 3 (2025): December 2025
Publisher : Forum Kerjasama Pendidikan Tinggi

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47065/bits.v7i3.8447

Abstract

Semarang has five religious tourism icons that represent pluralism, but their promotion is still conventional and not yet optimal in the digital era. This problem hinders its tourism potential in reaching a wider audience. This study aims to develop an accurate and efficient automatic image identification model as a modern solution to these promotional challenges. The method implemented is deep learning using the MobileNetV2 CNN architecture through a transfer learning approach. MobileNetV2 was chosen because it is superior in computational efficiency on resource-constrained devices compared to other models like EfficientNet. The model was trained and validated using a dataset consisting of a total of 7,500 images comprising five classes of religious tourist attractions, namely Grand Mosque of Central Java, Blenduk Church, Buddhagaya Watugong Temple, Pura Agung Giri Natha Temple, and Sam Poo Kong Temple. The dataset was divided into 70% training data, 15% validation data, and 15% test data. The evaluation results on the test data showed satisfactory performance, where the developed model achieved an overall accuracy of 98%, with a macro average F1-Score of 0.98. This figure indicates high and balanced performance across all classes. Individual testing also proved the model's ability to recognize relevant images with high confidence and reject images outside the class. This success shows that the implementation of MobileNetV2 is effective and can be basic technology for development of innovative digital tourism applications in Semarang.
Perbandingan Random Forest dan XGBoost Untuk Prediksi Penjualan Produk E-Commerce Rumah Madu Hayatunnisa, Destaria; Permata, Permata; Priandika, Adhie Thyo; Gunawan, Rakhmat Dedi
Building of Informatics, Technology and Science (BITS) Vol 7 No 3 (2025): December 2025
Publisher : Forum Kerjasama Pendidikan Tinggi

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47065/bits.v7i3.8491

Abstract

Inventory management is one of the main challenges for small and medium enterprises (SMEs), including Rumah Madu in Bandar Lampung, where honey stock levels are often determined based on estimation rather than precise calculation. This study aims to analyze and compare the performance of the Random Forest and XGBoost algorithms in predicting honey sales to achieve more measurable stock management. The dataset consists of 1,699 honey sales transactions that have undergone cleaning, feature transformation, and standardization processes. The variables used include honey type, unit price, day, month, holiday status, and promotion indicators. Modeling was conducted using a time-series split approach, where historical data served as the training set and recent data as the testing set. The evaluation results show that Random Forest achieved an MAE of 24.35, RMSE of 29.04, and R² of -0.9685, while XGBoost achieved an MAE of 25.50, RMSE of 30.58, and R² of -1.1825. The negative R² values indicate that both models were unable to explain data variation optimally, with performance falling below a simple baseline. Nevertheless, the feature importance analysis revealed that unit price and honey type were the dominant factors influencing sales. This study highlights the need for further model development through parameter optimization and improved data quality to enhance prediction accuracy.
Optimasi Hyperparameter Gaussian Naive Bayes Untuk Prediksi Risiko Stroke Pada Data Tidak Seimbang Nida, Khoirun; Mahenra, Ridwan; Susanto, Erliyan Redi
Building of Informatics, Technology and Science (BITS) Vol 7 No 3 (2025): December 2025
Publisher : Forum Kerjasama Pendidikan Tinggi

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47065/bits.v7i3.8497

Abstract

Stroke is a serious disease with global impact that requires high-accuracy early detection. Significant difficulties in designing machine learning-based predictive models arise due to disproportionate data conditions (imbalanced datasets). This occurs because the number of stroke cases (minority class) is very small compared to non-stroke cases. This imbalanced data situation often causes models to become biased and potentially produce high false negative rates, which is very risky in a clinical setting. This study focuses on improving the sensitivity of the Gaussian Naive Bayes (GNB) model through hyperparameter optimization and classification threshold adjustment. The research process included data preprocessing, stratified dataset division (70% training and 30% testing), feature scaling, var_smoothing parameter optimization using GridSearchCV, and threshold adjustment to maximize the Recall value. The results showed that the standard GNB model only achieved a Recall value of 0.4400. However, after var_smoothing optimization (1.00×10⁻¹⁰) and threshold adjustment to 0.0100, the Recall value increased significantly to 0.8000. This increase was accompanied by a decrease in Accuracy (0.5988) and Precision (0.0909). This improvement was accompanied by a decrease in Accuracy (0.5988) and Precision (0.0909). The high Recall (0.8000) indicates that the model is better for mass screening (early detection phase), although it must be balanced with further diagnostic processes due to low precision. This high Recall value confirms the model's success in minimizing False Negatives, which is a top priority in stroke risk prediction cases.
Analisis Respon Publik Terhadap Tren Penggabungan Foto Gemini AI Menggunakan Naive Bayes Afiani, Nanda; Mahenra, Ridwan
Building of Informatics, Technology and Science (BITS) Vol 7 No 3 (2025): December 2025
Publisher : Forum Kerjasama Pendidikan Tinggi

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47065/bits.v7i3.8504

Abstract

The rapid advancement of Artificial Intelligence (AI) technology has brought numerous innovations to the digital world, one of which is Gemini AI — an application capable of automatically merging photos based on user instructions. This phenomenon has gone viral on the TikTok platform and has sparked diverse public reactions, ranging from admiration for its visual results to concerns about ethical issues and the potential misuse of deepfake technology. This study aims to analyze public sentiment toward the trend of Gemini AI photo merging on TikTok using a sentiment analysis method based on the Naïve Bayes algorithm. Data were collected through a web scraping technique using the Apify platform, resulting in 5,061 user comments. The data processing stages included text preprocessing, TF-IDF transformation, and sentiment classification into three categories: positive, negative, and neutral. The results indicate that neutral sentiment dominates (4,059 comments), followed by positive (745 comments) and negative (257 comments). The dominance of neutral sentiment occurs because most user comments are informative or descriptive, expressing ordinary responses without strong emotional tones, rather than showing indifference to ethical concerns. The Naïve Bayes model demonstrated good performance with an accuracy of 85.72%, precision of 87.84%, recall of 85.72%, and F1-score of 81.95% through 5-fold cross-validation. These findings confirm that the Naïve Bayes algorithm is effective for classifying public opinion toward generative AI technologies. Overall, this study contributes to a deeper understanding of public perception of AI innovations in the creative digital domain and their social implications on social media platforms.
Analisis Sentimen Komentar YouTube terhadap Kenaikan Tunjangan DPR RI menggunakan Naïve Bayes, SVM, dan Random Forest Dani, Jemmi Rama; Parjito, Parjito
Building of Informatics, Technology and Science (BITS) Vol 7 No 3 (2025): December 2025
Publisher : Forum Kerjasama Pendidikan Tinggi

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47065/bits.v7i3.8513

Abstract

The rise of digital technology encourages the public to actively voice their opinions through social media, including in response to political issues such as the policy on increasing the remuneration of the Indonesian House of Representatives (DPR RI). This research aims to analyze public sentiment towards this issue on the YouTube platform using a comparative approach with three Machine Learning algorithms: Naïve Bayes, Support Vector Machine, and Random Forest. The data was acquired from viewer comments via the YouTube Data Application Programming Interface (API), totaling 78,866 lines of comments collected from seven videos discussing the DPR RI controversy. The data collection process utilized the googleapiclient.discovery.build module with API version V3, where the API_Key served as the authentication key to access data from YouTube. The research stages included preprocessing for data cleaning, sentiment labeling based on the InSet Lexicon Based method, and the application of the Synthetic Minority Over-sampling Technique (SMOTE) to address class imbalance in the data. The results show that before SMOTE application, the Support Vector Machine (SVM) model achieved the highest accuracy of 89%, followed by Random Forest at 81%, and Naïve Bayes at 62%. After applying SMOTE, the performance of all three models increased significantly, with SVM obtaining the highest accuracy of 93%, followed by Random Forest at 86%, and Naïve Bayes at 75%. For the positive class, SVM also demonstrated the best performance with a Precision value of 96%, Recall of 95%, and an F1-Score of 95%. Overall, the findings of this study confirm that SVM is superior in maintaining class balance in classification, both before and after SMOTE. The Machine Learning-based sentiment analysis approach is proven capable of providing a comprehensive overview of public opinion on political issues, while also offering important input for policymakers in formulating more transparent and responsive communication strategies.
Evaluasi Komparatif Algoritma Naïve Bayes, KNN, Logistic Regression, SVM, dan Extra Trees untuk Analisis Sentimen Tokopedia Ciputra, Indramawan; Fahmi, Amiq
Building of Informatics, Technology and Science (BITS) Vol 7 No 3 (2025): December 2025
Publisher : Forum Kerjasama Pendidikan Tinggi

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47065/bits.v7i3.8537

Abstract

The rapid evolution of digital technology has catalyzed a shift in consumer behavior, particularly in online shopping activities facilitated by e-commerce platforms such as Tokopedia. User-generated reviews yield large-scale textual data that can be systematically analyzed to uncover consumer sentiment in a factual and structured manner. This study aims to evaluate and compare the performance of five sentiment classification algorithms Naive Bayes, K-Nearest Neighbors (KNN), Logistic Regression, Support Vector Machine (SVM), and Extra Trees Classifier based on user review data from Tokopedia. The analytical workflow begins with web crawling, followed by text preprocessing procedures including tokenization, case folding, and stop-word removal, culminating in sentiment classification using the aforementioned algorithms. Performance evaluation was conducted using four standard metrics accuracy, precision, recall, and F1-score. The results reveal that SVM achieved the highest accuracy at 85%, outperforming KNN and Extra Trees Classifier (84%), Logistic Regression (82%), and Naive Bayes (79%). SVM’s superior performance is attributed to its ability to identify optimal hyperplanes that effectively separate sentiment classes, particularly in high-dimensional feature spaces. These findings offer practical insights for developers of sentiment analysis systems in selecting the most effective algorithm, while reinforcing the strategic application of Natural Language Processing (NLP) techniques within Indonesia’s e-commerce landscape.
Peningkatan Kinerja Model Naïve Bayes untuk Analisis Sentimen Komentar Terkait “Sound Horeg” Menggunakan SMOTE dan Tuning Parameter Kaisalana, Mustafid; Trisnapradika, Gustina Alfa
Building of Informatics, Technology and Science (BITS) Vol 7 No 3 (2025): December 2025
Publisher : Forum Kerjasama Pendidikan Tinggi

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47065/bits.v7i3.8554

Abstract

The phenomenon of “Sound Horeg” on online platforms has sparked diverse public sentiments, making sentiment analysis an essential tool for understanding public opinion. This study aims to classify user sentiments (positive/negative) related to “Sound Horeg” using the Naïve Bayes algorithm. The dataset used in this research exhibits significant class imbalance, with a predominance of negative sentiments. The methodology involves a series of text preprocessing stages, including case folding, tokenizing, normalization, lexicon-based sentiment labeling, stopword removal, stemming, and duplicate removal. The sentiment labeling process utilizes an Indonesian sentiment lexicon compiled from two sources lexicon_positif.csv and lexicon_negatif.csv containing predefined lists of words with positive and negative sentiment scores based on Indonesian public opinion lexicons. Subsequently, text features are extracted using the Term Frequency–Inverse Document Frequency (TF-IDF) method. To address data imbalance, the Synthetic Minority Oversampling Technique (SMOTE) is applied to the training data to balance the number of positive and negative samples. The Naïve Bayes model is then optimized using GridSearchCV to determine the best alpha value. Experimental results show that the unoptimized Naïve Bayes model achieved an accuracy of 73%, but struggled to classify minority classes (positive sentiments) due to data bias. After applying SMOTE and parameter tuning, the model’s performance improved significantly, demonstrating the effectiveness of these techniques in producing a more balanced and robust model. This study concludes that the Naïve Bayes algorithm, when optimized with SMOTE and hyperparameter tuning, is effective for Indonesian-language sentiment analysis, particularly on imbalanced datasets. Future work may include exploring other algorithms and employing broader sentiment lexicons and more complex linguistic features to further enhance model performance.
Analisis Sentimen Pengguna X terhadap Kasus Korupsi Gula Tom Lembong Menggunakan Naïve Bayes, SVM, dan Random Forest Kuncoro, Aneira Vicentiya; Budiman, Fikri; Kurniawan, Defri
Building of Informatics, Technology and Science (BITS) Vol 7 No 3 (2025): December 2025
Publisher : Forum Kerjasama Pendidikan Tinggi

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47065/bits.v7i3.8577

Abstract

The alleged sugar import corruption case involving Tom Lembong has become one of the most widely discussed public issues on social media, generating diverse reactions. This phenomenon illustrates how public opinion on legal issues is often influenced by perceptions of the public figures involved. This study aims to analyze public sentiment regarding the case on the social media platform X (formerly Twitter). The dataset consists of 1,802 tweets collected through a crawling process using the X API with the keyword “Tom Lembong.” The research stages include data cleaning, case folding, text normalization, tokenizing, stopword removal, stemming, sentiment labeling using a lexicon-based approach, and feature extraction with the Term Frequency–Inverse Document Frequency (TF-IDF) method. The prepared dataset was then tested using three classification algorithms: Naïve Bayes, Support Vector Machine (SVM), and Random Forest. The results show that the SVM algorithm achieved the highest accuracy (84%), followed by Random Forest (80%) and Naïve Bayes (76%). Based on the sentiment labeling results, positive sentiment dominated with 61%, while negative sentiment accounted for 39%. Although the analyzed issue concerns an alleged corruption case, the dominance of positive sentiment indicates that public opinion tends to focus on Tom Lembong’s personal image or public track record, which is viewed positively rather than on the substance of the legal allegations. These findings demonstrate the effectiveness of the SVM algorithm in analyzing high-dimensional text and provide insights into how public perception of legal issues can be influenced by image factors and the socio-political context on social media.
Comparative Analysis of Loss Functions for Predicting Autoimmunity from Molecular Descriptors Using Deep Learning Gunawan, Candra; Robet, Robet; Hendri, Hendri
Building of Informatics, Technology and Science (BITS) Vol 7 No 3 (2025): December 2025
Publisher : Forum Kerjasama Pendidikan Tinggi

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47065/bits.v7i3.8581

Abstract

Drug-induced autoimmunity (DIA) presents a complex obstacle in pharmacological safety due to its rare occurrence and unpredictable manifestation, often compounded by class imbalance in clinical datasets. This study investigates the influence of three loss functions, Binary Cross-Entropy (BCE), Focal Loss, and Dice Loss, on the performance of deep learning architectures comprising Multi-Layer Perceptron (MLP), Convolutional Neural Network (CNN), and 2-Layer Neural Network (SimpleNN). Models were trained using numerical molecular descriptors from the publicly available DIA dataset. The architectures were chosen based on their complementary properties: MLP is suitable for high-dimensional tabular descriptor data, CNN was examined to explore whether 1D convolutions can capture localized feature interactions among correlated descriptors, and 2-Layer Neural Network served as a lightweight baseline for comparison. A stratified 5-fold cross-validation strategy was employed to ensure statistical robustness. The results demonstrate that the MLP model, optimized with Focal Loss, consistently delivered the highest classification performance, achieving average scores of 94% accuracy, 93% precision, 95% recall, 94% F1-score, and an AUC of 0.97. In contrast, CNN and SimpleNN architectures yielded less favorable outcomes under the same loss configurations. These findings highlight the importance of aligning loss function choice with model complexity in the context of imbalanced biomedical data. The insights from this work contribute to the development of more reliable computational frameworks for early-phase immunogenicity screening and support the advancement of precision pharmacovigilance strategies.
Analisis Klasterisasi Kualitas Internet Seluler Menggunakan Metode K-Means dan Gaussian Mixture Model Irwansyah, Muhammad Aziiz; Meiriza, Allsela; Lestarini, Dinda
Building of Informatics, Technology and Science (BITS) Vol 7 No 3 (2025): December 2025
Publisher : Forum Kerjasama Pendidikan Tinggi

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47065/bits.v7i3.8615

Abstract

This study utilizes internet network data from Ookla Open Data (Speedtest Global Performance), comprising three main variables: download speed, upload speed, and latency. The aim is to analyze the condition and performance of mobile internet networks across 17 regencies/cities in South Sumatera Province in 2025 and to provide data-driven recommendations for the Department of Communication and Informatics to promote equitable and improved digital infrastructure through a Knowledge Discovery in Databases (KDD) approach. The applied methods include RobustScaler for data normalization, Principal Component Analysis (PCA) for dimensionality reduction, and K-Means and Gaussian Mixture Model (GMM) algorithms for clustering regions based on network characteristics. The analysis shows that both algorithms form three clusters (K=3) with distinct patterns. GMM demonstrates higher stability than K-Means, achieving a Silhouette score of 0.426 and Davies–Bouldin Index of 0.284, compared to K-Means with 0.351 and 0.688, while the lower Calinski–Harabasz score of GMM (9.960) indicates a trade-off between cluster compactness and stability, highlighting its adaptive behavior to data variation. Urban areas such as Palembang and Prabumulih belong to the high-performance cluster, whereas Ogan Komering Ulu Selatan lies in the low-performance cluster (18.87 Mbps; 33 ms), revealing a digital gap of approximately 18 Mbps across regions. These findings emphasize the need for equitable digital infrastructure strategies through fiber-optic expansion, BTS capacity enhancement, and multi-stakeholder collaboration toward Indonesia’s Digital Vision 2045.