AI Based Anomaly Detection in Application Performance Monitoring Using Random Forest

Resumen

This study was carried out to investigate the use of artificial intelligence and machine learning in application performance monitoring software systems and to develop anomaly detection methods. In the project, Random Forest algorithm was applied on the performance data obtained from “AdminServer” and “ManagedServer 1” servers, which are JVM processes, and anomalies were successfully detected using dynamic and dynamic thresholds. In the data processing process, data in JSON format were analyzed, dynamic threshold values were calculated with Z-score and data exceeding the determined threshold values were marked as anomalies. Anomaly detection was performed with the Random Forest classification algorithm, and model accuracy and classification performance were evaluated with metrics. In the visualization steps, histogram, KDE and scatter plot techniques were used to present both the data distribution and the detected anomalies in detail. The results obtained aim to provide an effective approach to system management by automating anomaly detection processes.