Performance Evaluation of Random Forest and Isolation Forest Algorithms for Detecting Anomalies in SIAKAD Server Log Data

Abstract

The increasing reliance on Academic Information Systems (SIAKAD) in higher education institutions has resulted in a significant growth of server log data, which contains valuable information for monitoring system performance and security; however, manually identifying anomalies in large-scale log data remains inefficient, time-consuming, and prone to human error. Therefore, this study aims to evaluate and compare the performance of Random Forest and Isolation Forest algorithms in detecting anomalies within SIAKAD server logs. This research adopts a machine learning-based approach using a dataset of 5,000 server log records collected from the SIAKAD system, which underwent preprocessing stages including data cleaning, feature extraction, and normalization. Random Forest was implemented as a supervised learning method, while Isolation Forest was applied as an unsupervised anomaly detection technique, with performance evaluated using accuracy, precision, recall, and F1-score metrics. The experimental results show that Random Forest achieved an accuracy of 96.3%, precision of 95.8%, recall of 96.0%, and F1-score of 95.9%, while Isolation Forest achieved an accuracy of 94.1%, precision of 92.7%, recall of 93.5%, and F1-score of 93.1%. These findings indicate that both algorithms are effective in detecting anomalies, with Isolation Forest demonstrating strength in handling unlabeled data and rare events, while Random Forest provides higher performance when labeled data is available; thus, this study highlights the potential of integrating machine learning techniques into log monitoring systems to enhance anomaly detection in academic information systems.

GS Cited Analysis