<![CDATA[Extreme rainfall requires monitoring due to the potential increased risk of hydrometeorological disasters. The BMKG defines extreme rainfall as rainfall exceeding 150 mm per day; however, rainfall characteristics vary by region. This study aims to apply the Isolation Forest algorithm to detect daily rainfall anomalies and compare the results with the BMKG extreme rainfall threshold. Data from Juanda, Kediri, and Malang stations—totaling 1,006 observations—were used. The model used daily rainfall and month of observation as features, with a contamination value of 0.05. Results showed that the model detected 17 anomalies in Juanda, 17 in Kediri, and 18 in Malang. The model also identified one BMKG-defined extreme event in Juanda (150.8 mm) as an anomaly. Comparative evaluation showed a 100% recall rate for BMKG-defined extreme events; however, this figure requires cautious interpretation, as the dataset contains only one BMKG-defined extreme event. These findings suggest that Isolation Forest can serve as a tool to identify local rainfall anomalies without replacing BMKG operational thresholds.]]>
Copyrights © 2026
