Jurnal Sains, Nalar, dan Aplikasi Teknologi Informasi

Auditing ABC University's New Student Registration System Maturity Level Through COBIT 2019

Luthfi Indana — 2026-07-10

The New Student Admissions Information System (PMB) at ABC University still faces several challenges, such as slow document verification, data inconsistencies, and a monolithic structure integrated with other systems (e.g., SIAKAD), which negatively impacts its performance. This study evaluates the performance of this web-based PMB system using the COBIT 2019 framework. A quantitative approach was employed, with data collected through interviews, observations, questionnaires, and document analysis. The study focuses on the DSS (Deliver, Service, Support) domain of COBIT 2019 to assess the system's maturity level. The analysis reveals that the PMB system's maturity level is 3.78 (Level 4), indicating that processes are well-established but require further optimization to reach Level 5 (Optimizing). To achieve this, PUSIM (Center for Information Technology and Multimedia) must develop and periodically review planning documents to ensure optimal performance of the information system services in line with institutional targets.

Integration Sentiment Analysis and K-Means Clustering in Semeru Pine Forest Management

Solimun — 2026-07-10

Forest-based tourism management requires integrated institutional and community involvement to ensure sustainability. Semeru Pine Forest in Sumberputih Village, Malang Regency, has significant tourism potential but has experienced a decline in management quality in recent years. This condition is reflected in rising public and visitor complaints about infrastructure, accessibility, and governance. This study aims to empirically examine patterns in public and visitor perceptions of the management of the Semeru Pine Forest. The research employs sentiment analysis of Google review data to extract key perceptual variables, followed by K-Means clustering. Survey data were collected from 200 respondents using a five-point Likert scale. The suitability of the data structure was assessed using Bartlett’s Sphericity Test prior to clustering. The sentiment analysis identified key variables including organizational management, community participation, and tourism sustainability. The K-Means analysis produced two distinct clusters representing different levels of management performance. The first cluster reflects lower perceptions of management quality, while the second cluster indicates relatively better management performance. These findings provide empirical evidence to support data-driven strategies for improving sustainable forest tourism management.

Evaluation of SDCA, LBFGS, LightGBM and FastTree in ML.NET for Diabetes Prediction

Resi Taufan — 2026-07-12

This study aims to develop a machine learning-based diabetes risk prediction model using the ML.NET framework. The dataset utilized is a balanced-split version of the 2015 BRFSS, consisting of 70,692 respondents and 21 health indicator variables. Two training approaches were applied to analyze model performance: a baseline with default parameters and hyperparameter tuning. The preprocessing stage involved combining variables into feature vectors, Min-Max normalization, and an 80:20 train-test data split. The models were trained using four algorithms: SDCA Logistic Regression, LBFGS Logistic Regression, LightGBM, and FastTree. Evaluation results showed that LightGBM with the hyperparameter tuning approach, delivered the most consistent performance, achieving 75.37% accuracy, 82.86% AUC, 76.22% F1-score, 72.92% precision, and 79.83% recall. Feature analysis confirmed that GenHlth, HighBP, BMI, HighChol, and Age contributed dominantly to diabetes risk, aligning with medical literature regarding metabolic factors. The best-performing LightGBM model was then integrated into a .NET-based prototype application with a Razor Pages web interface. The practical contribution of this research is proof of concept for machine learning integration into e-health systems to support early detection and digital prevention of diabetes complications in the future.

Accuracy–Efficiency Trade-off Analysis of CNN Backbones for Multi-Class Waste Classification

Mochamad Rizal Fauzan — 2026-07-12

Automated waste classification is a critical component of intelligent recycling systems, where model selection must balance predictive performance and computational efficiency. This study benchmarks three representative convolutional neural network (CNN) backbones—ResNet50, EfficientNet-B0, and MobileNetV3-Large—for eight-class waste classification under controlled augmentation and unified optimization protocols. Using a fixed 80/10/10 split (7,747 training, 969 validations, and 960 testing images), all models are evaluated across multiple random seeds to ensure statistical reliability. Performance is assessed using macro-F1, precision, recall, and GPU-based inference latency to characterize the accuracy–efficiency trade-off. EfficientNet-B0 achieves the highest macro-F1 (0.9676 ± 0.0034), while MobileNetV3-Large delivers comparable performance (0.9669 ± 0.0023) with substantially lower latency—approximately six times faster than ResNet50. Augmentation sensitivity analysis further reveals architecture-dependent robustness under occlusion-based perturbation. These results demonstrate that lightweight architectures can achieve near-optimal classification performance with significantly reduced computational cost, providing deployment-oriented guidelines for practical waste sorting systems.

Optimizing Emotion Recognition in Indonesian Product Reviews Using LoRA on Transformer

David Suharjanto — 2026-07-13

Emotion analysis in product reviews plays a crucial role in Indonesia’s e-commerce landscape, yet training advanced transformer models demands high computational resources, posing a significant constraint for small businesses. This research evaluates the Low-Rank Adaptation (LoRA) technique for optimizing emotion recognition in Indonesian product reviews, with a focus on resource efficiency. Indobert-large-p2, Indobert-base-uncased, and Indobertweet-base-uncased models were trained on the PRDECT-ID dataset, comparing the performance of full fine-tuning and LoRA. Results show LoRA provided competitive performance; on the Indobert-large-p2 model, the F1-score reached 67.81%, surpassing full fine-tuning’s 67.74%, despite training only about 1% of the total parameters. LoRA significantly reduced VRAM consumption by up to 17% and accelerated training duration by up to four minutes. Furthermore, the transformer models fine-tuned in this study, using both LoRA and full fine-tuning, consistently outperformed the results of previous research that employed complex architectures like CNN and BiLSTM, with the highest F1-score reaching 69.72% compared to the previous best of 66.13%. LoRA proves to be a practical and efficient solution for limited computational resources, enabling effective emotion analysis without expensive infrastructure.

Cloud Infrastructure Security: Detecting and Analyzing Attacks on Windows Server 2019

Ikhwan Alfath Nurul Fathony — 2026-07-13

Cloud infrastructure security represents a critical challenge in addressing cybersecurity threats, particularly for internet-facing services such as Remote Desktop Protocol (RDP) and SQL Server. This research investigates cloud infrastructure security based on Windows Server 2019 through the development of a proactive and responsive attack detection and analysis framework using the Wazuh platform as Security Information and Event Management (SIEM) integrated with the MITRE ATT&CK framework. The research method employs an experimental approach with continuous monitoring for 30 days of two Windows Server 2019 units running RDP and SQL Server services. Attack simulations were conducted using eight different scenarios including RDP brute force, SQL Server authentication brute force, port scanning, privilege escalation, lateral movement, data exfiltration, persistence mechanisms, and defense evasion. Monitoring results revealed 110,492 total security events, dominated by 109,057 authentication failures (98.7%) and only 171 successful authentications, with the remainder consisting of other activities such as port scanning and process execution. The Wazuh-based detection system with MITRE ATT&CK integration successfully mapped 15 attack techniques, 10 of which were actively observed during the 30-day monitoring period, with a detection rate of 93.2%, false positive rate of 6.8%, and average response time of 2.4 seconds. Compliance analysis showed 87% compliance with PCI DSS, 91% with NIST 800-53, 85% with HIPAA, and 89% with GDPR. The research concludes that the integration of Wazuh SIEM with the MITRE ATT&CK framework is effective in detecting and analyzing cyber attacks on Windows Server 2019, with practical contributions in the form of implementation guidelines for rule-based detection and correlation rules for multi-stage attack detection.

A Systematic Review of Transfer Learning and Data Augmentation in Neural Machine Translation of Low-Resource Languages

Nur Fikri Khuluq — 2026-07-18

Modern Neural Machine Translation (NMT) systems have achieved state-of-the-art, performance, largely due to the availability of large-scale parallel corpora. However, the translation quality of NMT for Low-Resource Languages (LRL) remains limited due to data sparsity. Numerous studies have proposed different strategies to address this challenge. Among the most widely adopted strategies are Transfer Learning (TL) and Data Augmentation (DA) strategies. This research aims to present a systematic review of how these techniques, including Back-Translation (BT), Hybrid Transfer Learning (HTL), and the utilization of self-supervised objectives such as Masked Language Modeling (MLM), Causal Language Modeling (CLM), and Denoising Autoencoder (DAE), affect the quality improvement of NMT for LRL. The findings show that a hybrid combination of TL and DA with a self-supervised objective is the most effective solution for extremely low-resource scenarios, capable of producing the highest translation quality (highest BLEU score) and outperforming baseline models and traditional methods such as Statistical Machine Translation (SMT).

UI/UX Design of Sampah Kita Application Using Design Thinking for Circular Waste Management

Kgs Rahmad Bagas Gadi — 2026-07-18

Waste management remains a major urban and environmental problem in Indonesia. Low waste sorting behavior, limited segregated collection, and weak user motivation reduce the effectiveness of waste management at the source. This study aims to design the Sampah Kita mobile application as a digital platform for scheduled waste pickup, environmental education, eco-challenges, rewards, and contribution monitoring. The research used the Design Thinking method through empathize, define, ideate, prototype, and test stages. User needs were collected through interviews and a small survey involving housewives, students, and sanitation workers. The findings show that users need practical pickup access, simple guidance, reward-based motivation, and transparent information about their environmental contribution. A high-fidelity prototype was developed and evaluated using the System Usability Scale with twenty respondents. The application obtained an average SUS score of 82, which indicates very good usability. The result suggests that Sampah Kita has potential to support sustainable waste management and circular economy practices through user-centered mobile interaction.