Two-Phase Defect Detection Using Clustering and Classification Methods

Autonomous fault management of network and distributed systems is a challenging research problem attracts many activities. Solving this heavily depends on expertise knowledge supporting tools for monitoring detecting defects automatically. Recent activities have focused machine learning techniques that scrutinize system output data mining abnormal events defects. This paper proposes two-phase defect detection using log messages clustering classification. The approach takes advantage K-means method to obtain random forest detect the relationship existing Several experiments evaluated performance message Hadoop Distributed File System (HDFS) bug report Bug Tracking (BTS). Evaluation results disclosed some remarks with lessons learned.