Purpose: This study evaluates the performance of distributed lag, autoencoder, and LSTM autoencoder methods in detecting point anomalies in simulated data generated from Jakarta's Air Quality Index (AQI). The evaluation is conducted across several simulation scenarios that represent factors that may influence anomaly detection performance. Methods: Simulation scenarios were constructed by varying two anomaly characteristics: anomaly percentage (0.3%, 0.5%, and 1.0%) and anomaly depth (4.4σ, 4.7σ, and 5.0σ), yielding 90 datasets generated via repeated experiments. Anomaly detection was performed using a forecasting-based approach with a 4σ threshold on prediction errors. Model performance was evaluated using mean absolute percentage error (MAPE) for forecasting accuracy and balanced accuracy for anomaly detection. Result: Increasing anomaly percentage significantly degrades both forecasting and anomaly detection performance across all methods. In contrast, anomaly depth has no significant effect on forecasting accuracy but strongly influences detection performance. Among the evaluated methods, the distributed lag model consistently shows the most robust anomaly detection performance across scenarios, outperforming the autoencoder and LSTM autoencoder, particularly at higher anomaly percentages and depths. Novelty: This study introduces a more rigorous and structured evaluation framework for anomaly detection by integrating three key contributions. First, it provides a unified comparison of distributed lag, autoencoder, and LSTM autoencoder methods within a single experimental setting, a limitation in prior studies. Second, it employs a controlled simulation design that systematically varies the anomaly percentage and depth while preserving key characteristics of empirical AQI data, enabling a more objective assessment of model performance across diverse anomaly conditions. Third, it uses ANOVA and interaction analysis to formally examine the effects of anomaly characteristics on both forecasting and detection performance, moving beyond purely descriptive comparisons commonly used in previous research.

Purpose: Lung cancer remains one of the leading causes of death worldwide, especially among active smokers, yet early detection is still difficult because traditional imaging methods have limited sensitivity for identifying early-stage abnormalities. This study was conducted to address the need for a more accurate computational approach capable of detecting lung cancer at a molecular level using gene expression data. The goal is to build a model that can reliably distinguish cancerous from non-cancerous samples based on genomic features. Methods: This study uses the GSE4115 gene-expression dataset consisting of 187 bronchial epithelial samples and 22,215 gene features. The Moth-Flame Optimization (MFO) algorithm was implemented to select the most informative subset of genes from this high-dimensional dataset. A Support Vector Machine (SVM) classifier was then trained using multiple kernels, with hyperparameter tuning performed to identify the optimal configuration for each kernel. Results: Experimental results show that the Polynomial kernel achieved the highest performance using 286 MFO-selected features, reaching an accuracy of 0.84 and an F1-score of 0.85. These results confirm that combining MFO with SVM improves classification performance compared to using raw gene data without feature selection. Novelty: This study provides the first application of MFO-based feature selection for lung cancer prediction in smokers using the GSE4115 dataset. The findings demonstrate the value of nature-inspired optimization for handling high-dimensional genomic data and offer a promising direction for developing early computational detection methods.

Purpose: The rapid expansion and administrative complexity of community service programs (Kuliah Kerja Nyata/KKN) have made manual group formation increasingly inefficient, inconsistent, and prone to imbalance. This creates an urgent need for an automated, fair, and reliable optimization method capable of handling large-scale grouping constraints. This study aims to evaluate the performance of the Particle Swarm Optimization (PSO) algorithm in generating optimal KKN group formations, focusing on computational efficiency, convergence behavior, and solution quality. Methods: PSO was implemented to form 27 KKN groups using 10 independent runs. Performance metrics included execution time, optimal iteration counts, initial fitness scores, and best final scores. Each run was analyzed to observe convergence patterns and stagnation behaviors. Result: The results indicate that PSO is highly efficient, with very fast execution times and rapid convergence, often reaching optimal solutions in the first iteration. However, performance varied: some groups achieved low optimal scores (95–97), while many stagnated at extremely high scores (100000.0) with no improvement. This shows that PSO’s effectiveness depends heavily on problem characteristics and initialization. Novelty: This study identifies and explains stagnation patterns in PSO when applied to discrete, constraint-heavy academic group formation problems, an area rarely examined in prior research. The analysis provides insight into PSO’s strengths and limitations and highlights the need for improved parameter tuning and initialization strategies. The findings serve as a foundation for developing more robust optimization approaches for fair and efficient KKN group formation.

Purpose: Stroke is one of the leading causes of death and disability in the world that requires a fast and accurate diagnosis system. A major challenge in classifying strokes using deep learning is data imbalances, where the number of stroke patients is much less than that of non-stroke patients. Methods/Study design/approach: This research proposes a Hybrid model approach that combines Inception and Xception architectures, and applies Synthetic Minority Over-sampling Technique (SMOTE) to balance the data distribution. The dataset used consisted of 5,110 entries with 12 stroke risk features, and evaluation was performed using accuracy, precision, recall, and F1-score metrics. Result/Findings: The results show that the Hybrid model provides the best performance with an accuracy of 92.2%, outperforming the Inception (86.28%) and Xception (89.26%) models. In addition, the Hybrid model showed high and balanced precision and recall values, reflecting its reliability in detecting stroke cases. Novelty/Originality/Value: The novelty of this research lies in combining the multi-scale feature extraction power of the Inception architecture and the depthwise separable convolution efficiency of the Xception architecture in a hybrid model. This approach is proven to excel in tabular data-based stroke classification and has the potential to be applied in automated medical diagnosis systems.

Objective: Cyberloafing Behavior (CB) has emerged as a major concern in the workplace. Apart from causing decreased performance, cyberloafing also carries the risk of hacking or data breaches. This study seeks to explore how role stressors influence Cyberloafing Behavior among healthcare professionals. Methods: This study employed a quantitative research design, utilizing accidental sampling to collect data from 109 respondents at Sinar Kasih Hospital, Purwokerto. Data analysis was conducted using Structural Equation Modeling (SEM) with the assistance of SmartPLS software. Result: The results of this study reveal that role stressors, which include Role Ambiguity (RA), Role Conflict (RC), and Role Overload (RO), have a statistically significant and positive impact on Cyberloafing Behavior (CB) among healthcare employees. High stress from unclear roles, conflicting duties, and heavy workloads increases the likelihood of non-work internet use. Addressing these stressors can help reduce counterproductive behavior and improve focus in healthcare environments. Novelty: This study recommends that healthcare organizations provide clear work guidelines to prevent role ambiguity, monitor workloads to reduce stress, and address role conflicts among employees. These strategies can help to reduce role stressor factors to prevent cyberloafing behavior among employees in order to avoid the risks arising from such behavior. The novelty of this study lies in its application of varied research subjects and a distinct methodological approach, setting it apart from previous studies. By focusing on the healthcare sector and employing SEM analysis, it offers new insights into the relationship between role stressors and Cyberloafing Behavior.

Purpose: This study aims to enhance the user experience of Learning Management Systems (LMS) by integrating two established design frameworks: the UX Five Elements and Design Thinking. The research addresses the need for a more structured yet human-centered design process to improve the usability and engagement of LMS platforms in higher education. Methods: The research adopts a design and development approach by combining the UX Five Elements, which offer a systematic structure across five user experience layers, with Design Thinking, which emphasizes empathy and iterative user involvement. This integration forms an Extended Model Design (EMD) used to guide the development of a new LMS interface. The final system was evaluated using usability testing involving students as target users. Result: Evaluation of the LMS prototype using the User Experience Questionnaire (UEQ) showed positive perceptions on all six dimensions, with the highest scores on the Efficiency (1.644) and Attractiveness (1.634) aspects, reflecting a practical and attractive system design. Although the Novelty (1.203) aspect had the lowest score, its value was still above the positive threshold, indicating that the system was functionally good but could still be improved in terms of innovation to strengthen user engagement. Novelty: This study introduces a novel design framework by integrating UX Five Elements with Design Thinking in the context of LMS development. Extended Model Design (EMD) offers a replicable model that balances structure and user empathy, contributing to user-centered e-learning system design.

Purpose: Customer churn is a crucial issue for companies, especially those in the telecommunications sector, as it has a direct impact on revenue and new customer acquisition costs. The purpose of this research is to create a customer churn prediction model through performance comparison between the Logistic Regression algorithm and Ridge Classifier, considering the effect of data balancing. Methods: This study developed a churn classification model by comparing the Logistic Regression and Ridge Classifier algorithms in three scenarios: without data balancing, balancing using SMOTE, and balancing using GAN. The dataset used was Telco Customer Churn from Kaggle. Model evaluation was performed using a confusion matrix with accuracy, precision, recall, and F1-score metrics, with a primary focus on the accuracy metric. Result: The results show that data balancing using SMOTE and GAN does not improve model accuracy. The highest accuracy was achieved by the Ridge Classifier without data balancing, at 82.47%, followed by Logistic Regression at 82.25%. However, the recall and F1-score metrics improved when using SMOTE. The highest recall was achieved by Ridge Classifier at 75.34% and Logistic Regression at 75.07% in the SMOTE 50:50 scenario. The highest F1-score was also achieved by Ridge Classifier at 64.76% and Logistic Regression at 64.68% followed by the SMOTE 50:30 scenario. Meanwhile, the precision metric tends to decrease after data balancing. Novelty: The uniqueness of this study lies in the comparison of the performance of the Ridge Classifier and Logistic Regression in data balancing scenarios using SMOTE and GAN, which has not been widely discussed in previous studies. The main findings show that the highest accuracy is achieved when the Ridge Classifier model uses original data or without applying SMOTE or GAN data balancing. However, data balancing using SMOTE has been proven to significantly improve the recall and F1-score metrics.

Purpose: Diabetes is a chronic disease with a globally rising prevalence. Early detection of individuals at risk is essential to prevent long-term complications. This study aims to develop a diabetes prediction model that not only achieves high classification accuracy but also provides transparent explanations of the factors influencing its predictions. Methods: The study utilized the Pima Indians Diabetes Dataset, which contains clinical data from 768 female patients aged over 21. The methodology included data preprocessing (handling of missing values and feature engineering, such as the creation of Age_BMI and Glucose_BMI features), a 70:30 train-test split, class imbalance handling using the ADASYN technique, model development using the Random Forest algorithm with hyperparameter tuning via GridSearchCV, and model interpretability analysis using SHAP. Result: The proposed model achieved an accuracy of 79.2% and a recall of 85.2% on the test data. SHAP analysis revealed that Glucose, Age_BMI, BMI, and DiabetesPedigreeFunction were the most influential features in predicting diabetes. Furthermore, the SHAP heatmap indicated that individuals aged 30–50 years with obesity were at the highest risk. These findings align with existing medical literature, reinforcing the role of metabolic and age-related factors in diabetes development. Novelty: This study presents an integrative approach combining class balancing (ADASYN), classification (Random Forest), and model interpretability (SHAP) in a unified framework for diabetes prediction. It emphasizes the importance of transparent model interpretation for healthcare professionals, enabling not only predictive outcomes but also actionable insights into risk factors. The findings support future research opportunities, including the integration of lifestyle variables and external validation using real-world clinical data from diverse populations.

Purpose: This research aims to assess the maturity level of governance and service management in the Library Information System (SIPERPUS) at UNISNU Jepara. The objective is to identify strengths and gaps, ensuring better alignment with ITIL V4 best practices. Methods: The research was conducted using a qualitative methodology that incorporated a literature review, user questionnaires, and the creation of a web-based audit application designed specifically for this study. Service practices such as Service Request Management, Service Desk, Change Control, and Service Catalog Management were evaluated using the ITIL V4 framework. Result: The findings indicate that SIPERPUS has independently attained a managed maturity level with well-defined Service Request Management and Service Desk practices. Additional work is required to improve Change Control and Service Catalog Management to increase process integration and make information clearer. The new web-based audit application described in this study demonstrates its capability to systematically evaluate, produce data-based recommendations, and continuously improve digital library services. Novelty: This study introduces a new web-based audit application that conforms to ITIL V4, providing an efficient and reliable alternative to manual audit methodologies. The tool significantly reduces audit time, enhances operational efficiency, and optimizes user satisfaction while improving overall service quality. The study adds to the literature on IT governance audits, specifically in higher education, and presents a scalable approach for institutions striving to meet the standards of Society 5.0.

Purpose: This study aims to analyze the dominant trends in color and composition within aesthetic photography on Instagram and explore their influence on user interaction, specifically likes and comments. Given the growing role of visual aesthetics in digital marketing, understanding these elements is crucial for content creators, brands, and businesses aiming to maximize engagement. Unlike previous studies that focus on general social media engagement, this research integrates technology-driven qualitative analysis using ATLAS.ti, enabling structured coding and thematic identification of visual elements. Methods: A qualitative content analysis was conducted on 591 Instagram posts tagged with #AestheticPhotography and #VisualAesthetic. Data was collected using Instagram scraping (PhantomBuster), extracting both visual (color palettes, composition techniques) and textual (captions, metadata) elements. The ATLAS.ti software was used to analyze recurring visual patterns and color extraction was performed via Google Colab and Python for accuracy. Result: The results show that natural colors (48.18%) and pastel tones (30.90%) are dominant in aesthetic photography, contributing to higher engagement due to their harmonious and calming effect. Composition techniques such as center alignment (40.51%) and the Rule of Thirds (23.27%) significantly correlate with user interaction, as they align with cognitive load theory and visual perception principles. Additionally, short captions (≤10 words) were more effective in enhancing engagement, receiving 8,876 likes and 4,432 comments on average, compared to longer captions. Novelty: This study bridges the gap between visual aesthetics and computational analysis, using ATLAS.ti to systematically examine social media trends. Unlike previous studies that focus solely on quantitative metrics, this research provides qualitative insights into how color and composition influence engagement. The findings offer practical guidance for content creators, designers, and marketers, suggesting that strong visual composition and color harmony can enhance audience engagement.