Kombinasi Algoritma K-Means Dan DBSCAN Dalam Identifikasi Anomali Pada Data Log Server

Abstract

Detecting anomalies in server log data is a crucial element of information system management and security. This research seeks to develop a method for identifying anomalies by integrating two well-known clustering algorithms: K-Means and DBSCAN (Density-Based Spatial Clustering of Applications with Noise). K-Means is effective at partitioning data into clusters based on average distances, while DBSCAN excels at detecting anomalies or noise in datasets without a distinct cluster structure. In this study, K-Means is employed for initial clustering of server log data to reveal general patterns and group similar data. The results from K-Means clustering are then examined using DBSCAN to detect anomalies more accurately. Combining these two algorithms aims to enhance anomaly detection accuracy by leveraging the strengths of each approach. The research was performed on a server log dataset encompassing various server activities. The effectiveness of this combined approach was assessed by comparing its anomaly detection performance against the individual K-Means and DBSCAN methods, as well as other anomaly detection techniques. Experimental results indicate that the K-Means and DBSCAN combination successfully improves anomaly detection rates by reducing both false positives and false negatives compared to using each algorithm independently.