Journal of Computer System and Informatics (JoSYC) https://ejurnal.seminar-id.com/index.php/josyc <p>Journal of Computer System and Informatics (JoSYC) is an e-Journal applied research in computer systems. This journal contains research articles and scientific studies. Journal of Computer System and Informatics is issued 4 (four) times a year in&nbsp;<strong>November</strong>(issue 1), <strong>February</strong>(issue 2), <strong>May</strong>(issue 3), and <strong>August</strong>(issue 4).&nbsp; Journal of Computer System and Informatics (JoSYC) ISSN&nbsp;<a href="http://issn.pdii.lipi.go.id/issn.cgi?daftar&amp;1569811449&amp;1&amp;&amp;">2714-7150 (Print)</a>, ISSN&nbsp;<a href="http://issn.pdii.lipi.go.id/issn.cgi?daftar&amp;1569812182&amp;1&amp;&amp;">2714-8912 (Online)</a>,&nbsp; is open to submission from scholars and experts. Journal of Computer System and Informatics (JoSYC) covers the whole spectrum of <strong>Artificial Intelligent</strong>, <strong>Computer Systems</strong>, and&nbsp;<strong>Informatic Techniques</strong>.</p> Forum Kerjasama Pendidikan Tinggi (FKPT) en-US Journal of Computer System and Informatics (JoSYC) 2714-7150 <p>Authors who publish with this journal agree to the following terms:</p> <ol> <li>Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under&nbsp;<a href="http://creativecommons.org/licenses/by/4.0/" rel="license">Creative Commons Attribution 4.0 International License</a>&nbsp;that allows others to share the work with an acknowledgment of the work's authorship and initial publication in this journal.</li> <li>Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgment of its initial publication in this journal.</li> <li>Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work (Refer to&nbsp;<a href="http://opcit.eprints.org/oacitation-biblio.html" rel="license">The Effect of Open Access</a>).</li> </ol> Analisis Sentimen Ulasan Pengguna Halodoc Menggunakan TextCNN pada Dataset Tidak Seimbang Berbasis NLP https://ejurnal.seminar-id.com/index.php/josyc/article/view/9331 <p>This study analyzes user sentiment in Halodoc application reviews using a TextCNN architecture on an Indonesian-language dataset with extreme class imbalance. The dataset consists of 2,000 reviews collected from the Google Play Store and manually labeled into three classes: positive, negative, and neutral. The class distribution comprises 1,800 positive reviews (90%), 175 negative reviews (8.75%), and 25 neutral reviews (1.25%). To reduce bias toward the majority class, the TextCNN model was trained using class weighting and evaluated using accuracy, precision, recall, F1-score, and a confusion matrix. The model achieved an accuracy of 93.75% and a weighted F1-score of 0.94, while the macro F1-score was only 0.57. The recall for the neutral class was 0%, indicating that the model was unable to recognize the minority class effectively. The contributions of this study are threefold: an empirical evaluation of TextCNN on Indonesian Halodoc reviews with extreme class imbalance; an analysis of the effect of class distribution using class-level metrics and error analysis; and an identification of the practical implications of the classification results for monitoring digital healthcare service quality. These findings demonstrate that high accuracy alone is insufficient to represent model performance on imbalanced datasets; therefore, strategies such as oversampling, data augmentation, or hybrid approaches should be considered in future research.</p> Fitriyani Fitriyani Budi Tjahjono ##submission.copyrightStatement## http://creativecommons.org/licenses/by/4.0 2026-08-31 2026-08-31 7 4 150 159 10.47065/josyc.v7i4.9331 Deteksi Cyberbullying pada Komentar Instagram Berbahasa Indonesia Menggunakan Fine-Tuning IndoBERT dengan Analisis Kesalahan Model https://ejurnal.seminar-id.com/index.php/josyc/article/view/9330 <p>Cyberbullying on social media is difficult to detect automatically because comments may contain informal language, sarcasm, implicit body shaming, and meanings that depend on context. This study evaluates the ability of IndoBERT to detect cyberbullying in Indonesian-language Instagram comments and analyzes the model’s classification errors. The dataset consists of 1,050 comments collected from public Instagram posts and manually labeled into two classes: cyberbullying and non-cyberbullying. The IndoBERT-base-uncased model was fine-tuned and evaluated using a confusion matrix, accuracy, precision, recall, and F1-score. The experimental results show an accuracy of 84.76% and an F1-score of 0.8462. For the cyberbullying class, precision reached 0.9294, while recall was 0.7524, indicating that a portion of bullying comments were still missed as false negatives. Error analysis shows that difficult cases were mainly associated with sarcasm, implicit body shaming, informal language, and ambiguous comments. The contributions of this study are an empirical evaluation of IndoBERT on Indonesian Instagram comments, a clear identification of the research gap between conventional machine-learning approaches and the need for contextual Indonesian-language modeling, and an error analysis that provides practical evidence for improving AI-based content moderation. The findings indicate that IndoBERT has promising performance, but larger datasets, conversational context, and pragmatic language modeling are still required.</p> Dewi Irma Afriayanti Budi Tjahjono ##submission.copyrightStatement## http://creativecommons.org/licenses/by/4.0 2026-08-31 2026-08-31 7 4 160 169 10.47065/josyc.v7i4.9330 Network Traffic Classification Using Machine Learning Algorithms Based on Quality of Service Parameters https://ejurnal.seminar-id.com/index.php/josyc/article/view/10932 <p>The growth of multimedia services requires network operators to identify traffic types quickly so that Quality of Service (QoS) policies can be applied precisely, while port-based classification and payload inspection are becoming unreliable because of encryption and dynamic port usage. This study analyzes 20,000 labeled flow records from four application classes, namely File Transfer, Video Streaming, VoIP, and Web Browsing, under three network load levels. Seven machine learning algorithms, namely Naive Bayes, Logistic Regression, K-Nearest Neighbor, Decision Tree, RBF-kernel Support Vector Machine (SVM), Random Forest, and XGBoost, were tested on two feature scenarios. Scenario A uses four QoS parameters (throughput, delay, jitter, and packet loss), while scenario B adds ten flow statistics. Port and protocol attributes were removed to prevent label leakage. Evaluation used a stratified 80:20 split, accuracy, precision, recall, F1-score, Cohen’s Kappa, and 5-fold cross validation. All four QoS parameters differ significantly across classes (Kruskal-Wallis test, p &lt; 0.001). SVM-RBF achieved the highest performance in both scenarios, with 74.3% accuracy (F1 74.9%; Kappa 0.656) in scenario A and 98.0% accuracy (F1 98.0%; Kappa 0.974) in scenario B, while Random Forest and XGBoost followed within 0.2 points in scenario B. Throughput (33.0%) and jitter (30.5%) were the most informative QoS parameters, and VoIP was the easiest class to recognize. QoS parameters alone are sufficient to separate interactive traffic such as VoIP but not to distinguish Video Streaming from Web Browsing and File Transfer. Combining QoS parameters with flow statistics raises accuracy by about 24 percentage points and provides a basis for adaptive traffic classification in QoS management.</p> Hermansyah Hermansyah Al-Khaidar Al-Khaidar Lisa Mulia Al Ikhlas Husni Husni ##submission.copyrightStatement## http://creativecommons.org/licenses/by/4.0 2026-08-31 2026-08-31 7 4 170 179 10.47065/josyc.v7i4.10932 A Phishing Detection Model Based on Hybrid Natural Language Processing and Graph-Based Learning for Digital Threat Identification https://ejurnal.seminar-id.com/index.php/josyc/article/view/10624 <p>Phishing remains one of the most damaging digital threats in Indonesia, yet most detectors rely on a single source of evidence that attackers can easily disguise. This study proposes a hybrid model that combines natural language processing of message text (TF-IDF with logistic regression), graph-based learning on a heterogeneous message–sender–URL-domain graph (label propagation, neighbour label ratios, and graph-aggregated text evidence), and eleven URL and behavioural attributes, fused by XGBoost trained with modality-dropout augmentation (MD-Hybrid). A stratified sample of 7,604 Indonesian-language messages (3,802 phishing, 3,802 legitimate) was split 80:20 and validated with 5-fold cross validation under a realistic adversarial protocol in which the domain reputation score is unavailable and 30% of phishing messages disguise one to three evidence channels. Without camouflage, all models were near-perfect (accuracy 0.998 to 1.000), so clean accuracy could not discriminate between them. Under the adversarial protocol, the NLP, URL, graph, and early-fusion models reached accuracies of 0.896 to 0.901, F1-scores of 0.884 to 0.890, and MCC values of 0.810 to 0.818, whereas MD-Hybrid reached an accuracy of 0.950, an F1-score of 0.947, and an MCC of 0.904 without false positives; its remaining errors were messages whose three channels were all disguised. Modality-dropout fusion prevents shortcut learning and keeps detection reliable while at least one evidence channel remains genuine; validation on additional Indonesian phishing sources is recommended before deployment.</p> Hermansyah Hermansyah Al–Khaidar Al–Khaidar Husni Husni ##submission.copyrightStatement## http://creativecommons.org/licenses/by/4.0 2026-08-31 2026-08-31 7 4 180 189 10.47065/josyc.v7i4.10624