International Journal of Advances in Data and Information System
Not a member yet
161 research outputs found
Sort by
Implementing GCV and mGCV to Determine Optimal Knot in Spline Regression for East Java Life Expectancy
Full text linkLife Expectancy is a vital indicator for evaluating population’s overall welfare and health status within a specific region. According to data published by Badan Pusat Statistik (BPS) National, East Java Province ranks 10th nationally in terms of life expectancy in 2024, with male life expectancy recorded at 70.39 years and female life expectancy at 74.4 years. This research focuses on examining four key factors that are believed to influence life expectancy in East Java during the 2024 including the Percentage of the Poor Population (X1), the Percentage of Individuals Aged 5 and Above Who Regularly Smoke Tobacco (X2), the Expected Years of Schooling (X3), and the Open Unemployment Rate (X4). To determine the optimal knot points in the nonparametric truncated spline regression model, the study utilizes Generalized Cross-Validation (GCV) and the modified Generalized Cross-Validation (mGCV) methode by minimizing their respective error values. The findings indicate that all four variables significantly impact life expectancy. Among the methods applied, the mGCV approach demonstrates good performance, achieving the lowest error value of 0.100 and a coefficient of determination of 82.91%
Improving Credibility of Digital Evidence Investigation in E-Commerce Fraud Cases using ISO/IEC 27037
Full text linkTikTokShop fraud is an emerging challenge in e-commerce investigations, demanding robust digital forensic approaches. This study tackles the complexities of investigating such fraud within the TikTokShop platform, focusing on the acquisition, preservation, and validation of multifaceted digital evidence, including screenshots, payment records, account data, videos, and communication logs. Adhering to ISO/IEC 27037 for evidence handling, Magnet and Oxygen forensic tools were used for systematic evidence acquisition. The analysis using Oxygen Forensic recovered 100% of relevant artifacts, which is slightly higher compared to Magnet Axiom, which recovered 38.46% of artifacts, although both tools were effective in retrieving critical artifacts such as image metadata, account information, and data transfers. Due to image compression by the TikTokShop application, discrepancies in hash values emerged, requiring supplementary validation. Optical Character Recognition (OCR) and Levenshtein distance algorithms quantified textual similarity within image-based evidence, while the Forensically platform enabled advanced image forensic analyses to detect potential tampering and authenticity. This rigorous, multi-layered forensic framework complements traditional hash verification by providing corroborative content-level authentication. Findings confirm that although hash inconsistencies arise from application-induced compression, integrating OCR, Levenshtein, and forensic image analysis enhances the reliability of digital evidence. The novelty of this research lies in its robust synergy of ISO/IEC 27037-compliant handling with advanced digital content verification, contributing to the advancement of digital forensic practices in complex social commerce fraud scenarios
Comparative Analysis of ARIMA and Fourier Series Methods for Air Temperature Forecasting in Surabaya
Full text linkUrban climate change, particularly rising temperatures and the Urban Heat Island (UHI) phenomenon, poses challenges for cities like Surabaya, Indonesia. This study compares the forecasting performance of ARIMA and ARIMA-Fourier models using daily air temperature data from 2020 to 2024. The analysis involved stationarity testing, model estimation, and evaluation across four forecasting horizons. ARIMA models (especially ARIMA(0,1,1) and ARIMA(1,1,0)) showed reliable short-term forecasts, but were less effective in capturing seasonal patterns. To address this, Fourier terms were integrated into the ARIMA framework. The ARIMA-Fourier model achieved better accuracy and higher R² values in short- and medium-term forecasts, particularly with an oscillation parameter of k = 150. However, its performance declined in long-term predictions due to overfitting risks. Overall, the ARIMA-Fourier model is more adaptive for capturing complex temperature seasonality and can support more accurate urban climate forecasting in Surabaya
Spatiotemporal Clustering of Key Food Commodity Prices Using Multivariate Time Series
Full text linkFood price stabilization remains a critical challenge in economic development planning and food security, particularly in developing countries like Indonesia, which exhibit high spatial and temporal diversity. To develop an efficient and adaptive predictive approach for understanding food commodity price dynamics, this study integrates multivariate time series clustering using a Dynamic Time Warping-based K-Means algorithm with a hybrid forecasting model that combines Vector Error Correction Model with Exogenous Variables and Long Short-Term Memory. The clustering evaluation results indicate reasonably cohesive group structures, with a silhouette score of 0.45 and a Davies-Bouldin Index of 0.67. Each cluster profile reveals significant differences in price trends, volatility, and anomaly patterns. Model validation using the Wilcoxon signed-rank test shows that the differences between cluster-level forecasts and individual-level actual values are generally statistically insignificant. These findings suggest that the proposed integrative approach can accurately capture regional price patterns and serve as a foundation for more data-driven and responsive policymaking in food price stabilization efforts. The 30-period forecasts for rice, eggs, and red onions reflected dynamic variations aligned with spatial characteristics: rice shows relatively stable behavior, eggs exhibit strong seasonal patterns, and red onions display the highest price volatility
Discovering Student Learning Paths: An Educational Process Mining Approach in Moodle
Full text linkE-learning platforms like Moodle are critical to modern education, with their effectiveness deeply reliant on fostering optimal student engagement. A thorough understanding of how students interact with these platforms is therefore essential for enhancing the learning experience. This study aimed to analyze student interaction patterns within Moodle by employing educational process mining techniques. The core objective was to uncover hidden behavioral patterns and gain valuable insights into the underlying learning processes. To achieve this, we utilized both heuristic miner and inductive miner algorithms to analyze Moodle\u27s extensive event log data. The effectiveness of various student activity variants was rigorously assessed through fitness checking. This study presents a novel, integrated analytical approach combining frequent sequential pattern mining, user behavior analysis, feature engineering, and process mining to comprehensively evaluate student learning effects in Moodle. While applying both Heuristic Miner and Inductive Miner algorithms to extensive Moodle event logs, we not only generated precise process models highlighting effective and ineffective student activity sequences but also uncovered unique challenges, such as the Inductive Miner\u27s inability to accurately model the \u27Tugas\u27 (assignment) component\u27s complex activity patterns. These findings offer distinct, actionable insights for refining Moodle course design and delivery, moving beyond general observations to pinpoint specific pedagogical interventions. Ultimately, our work advances the understanding of student behavior and academic performance within the Moodle ecosystem by providing a granular, data-driven methodology for optimization
Web-Based Monitoring System for Automatic Coffee Drying in a Smart Dryer Dome
Full text linkThis study developed a web-based monitoring system integrated into a smart dryer dome for automatic coffee drying. The system utilized the RN-GZWS-RS485 sensor to measure critical drying parameters: temperature, humidity, and light intensity. Data acquisition relied on an ESP32 microcontroller, transmitting real-time measurements to a server using the MQTT protocol, while sensor-actuator interactions operated through the Modbus protocol. Actuator performance adhered to predefined threshold values, maintaining drying temperature within 45–50?°C and relative humidity between 20–40%. Real-time monitoring and system status visualization were implemented via a Laravel-based web interface. Experimental tests demonstrated that 71.76% of temperature readings, 64.71% of humidity readings, and 68.24% of light intensity readings consistently fell within optimal ranges. Low standard deviation values confirmed the system’s effectiveness in maintaining stable drying conditions. Additionally, the integration of solar power facilitated system deployment in remote locations without conventional electricity infrastructure. These findings highlight the system\u27s potential to improve the reliability, accuracy, and efficiency of automatic coffee drying processes
Automated Oil Palm Health Assessment Using Object-Based Deep Learning and High-Resolution UAV Imagery in Indonesia
Full text linkIndonesia, as the world’s largest crude palm oil (CPO) producer, faces challenges in plantation monitoring due to reliance on manual data collection methods that are time-consuming, costly, and prone to human error. This study proposes an automated approach for assessing oil palm tree health using high-resolution UAV imagery (5–10 cm) and object-based deep learning models. We evaluate five state-of-the-art detectors—YOLOv5s, Faster R-CNN, Mask R-CNN, SSD, and RetinaNet—to classify individual trees into four health categories: Healthy, Moderately Healthy, Needs Improvement, and Urgent Condition. Using a dataset of 14,749 labeled trees from Kendawangan, Indonesia, YOLOv5s achieved the highest performance with a precision of 0.784, recall of 0.752, and mAP of 0.764. Our findings demonstrate the potential of AI-driven monitoring to enhance plantation management through rapid, accurate, and cost-effective health assessments—contributing a scalable solution to support precision agriculture and sustainable CPO production
Enhanced Classification of Lombok Pearl Quality Based on Shape and Size Using PSO-Optimized Artificial Neural Network
Full text linkThis study aims to develop an intelligent classification model for pearl quality assessment using an integrated approach combining Gray Level Co-occurrence Matrix (GLCM), Particle Swarm Optimization (PSO), and Artificial Neural Network (ANN). Sixteen texture features were extracted from four directional orientations using GLCM. PSO was employed as a feature selection algorithm to reduce dimensionality and enhance classification performance. Two ANN models were compared: a baseline model using all GLCM features and an optimized model utilizing only PSO-selected features. The models were trained and validated using 10-fold cross-validation. Results showed that the PSO-enhanced ANN achieved an accuracy of 94.72%, outperforming the baseline model which reached only 89.17%. Further evaluations using confusion matrix, Receiver Operating Characteristic (ROC) analysis, and Principal Component Analysis (PCA) confirmed the superior discriminative capability and improved class separability of the optimized model. These findings highlight the effectiveness of combining swarm intelligence with neural networks in texture-based classification tasks, offering a robust and scalable solution for automated quality inspection in the pearl industry and related domains
Geospatial Model Optimization for Mapping Social Vulnerability to Natural Disasters Using Fuzzy Geographically Weighted Clustering and Flower Pollination Algorithm
Full text linkThis study analyzes social vulnerability to natural disasters in Indonesia through a geospatial optimization model integrating Fuzzy Geographically Weighted Clustering (FGWC) with the Flower Pollination Algorithm (FPA). The hybrid FGWC–FPA enhances clustering accuracy by optimizing spatial parameters and addressing the limitations of index-based and non-spatial methods. The model tested two to four clusters, with the optimal configuration producing four distinct vulnerability groups. Cluster 1 (114 districts) exhibits high poverty, weak infrastructure, and low literacy; Cluster 2 (79 districts) reflects demographic pressure and gender-related inequality; Cluster 3 (87 districts) shows low education and poor disaster preparedness; while Cluster 4 (234 districts) represents health- and age-related vulnerability. A comparison with the 2024 Indonesian Disaster Risk Index (IRBI) shows strong spatial consistency, especially in high-risk regions such as Papua, Maluku, and Sulawesi. The FGWC–FPA model provides finer spatial granularity, allowing the identification of region-specific social issues not captured by deterministic index approaches. The findings validate national disaster risk patterns and offer complementary insights for implementing the National Disaster Management Master Plan (RIPB) 2020–2044, supporting regional prioritization, resource allocation, and capacity-building strategies
Implementation of Business Intelligence and Data Mining in Money Changer Transaction Analysis (Case Study of PT. Gemilang Artha Valindo)
Full text linkThis study aimed to implement Business Intelligence (BI) and Data Mining for analyzing currency exchange transactions at PT. Gemilang Artha Valindo to support data-driven decision-making. Transaction data was analyzed using Power BI to generate visualizations, including a pie chart for transaction frequency by currency type, a bar chart for the number of buy and sell transactions per currency, and a line chart for monthly average exchange rate fluctuations. The pie chart indicated that the AUD currency dominated transactions, contributing 51.95% of the total. The bar chart revealed that AUD buy transactions accounted for 63.22% of total AUD transactions, while the line chart showed that GBP and EUR had the highest average exchange rates, reaching Rp20,835 and Rp17,700, respectively. The exchange rate prediction process utilized three algorithms: Linear Regression, K-Nearest Neighbors (KNN), and Random Forest. Their performances were evaluated using Root Mean Squared Error (RMSE). The Random Forest algorithm produced the most accurate predictions with the lowest RMSE value of 134.63, followed by KNN and Linear Regression. These findings highlight the importance of leveraging BI and Data Mining to transform transaction data into valuable insights, enabling more informed business decisions