BEHAVIOR-BASED ANOMALY DETECTION IN ONLINE ATTENDANCE SYSTEMS USING A GENETIC ALGORITHM–OPTIMIZED ISOLATION FOREST

Wahyudi Ariannor (Department of Information System, STMIK Banjarbaru, Indonesia)
Budi Susarianto (Department of Informatics Engineering, STMIK Banjarbaru, Indonesia)
Siti Fathimah (Department of Information System, STMIK Banjarbaru, Indonesia)
Siti Abidah (Department of Information System, STMIK Banjarbaru, Indonesia)

AL-ULUM: JURNAL SAINS DAN TEKNOLOGI
Vol 12 No 2 (2026)

Digital attendance systems are vulnerable to anomalous behaviors that may reduce attendance integrity and reliability. This study proposes an anomaly detection approach using Isolation Forest optimized with a Genetic Algorithm to improve anomaly detection performance in digital attendance environments. The proposed model utilizes behavioral features, including location, device consistency, network connectivity, and working duration. Experimental evaluation using Stratified 5-Fold Cross Validation demonstrates that the GA Hybrid approach significantly improves model performance, increasing the F1-score from 0.2934 to 0.6004 and recall from 0.1791 to 0.4345 while maintaining high precision. Feature analysis further reveals that location-based attributes are the most influential indicators in detecting anomalous attendance behavior. The results indicate that the proposed optimization approach can improve anomaly detection effectiveness while maintaining stable model performance.