ML Based Intelligent Log Analytics Using BERT

Modern distributed systems generate large volumes of log data that contain critical information about system behavior and failures. Manual log inspection is inefficient and traditional monitoring tools produce excessive alerts without identifying the root cause of issues. This paper proposes an intelligent log analytics framework using the Bidirectional Encoder Representations from Transformers (BERT) model for automated log understanding and anomaly detection.The proposed system converts unstructured logs into structured representations through preprocessing techniques such as regex parsing and tokenization. Contextual embeddings generated by a fine-tuned BERT model are used for anomaly classification and root cause ranking. The system integrates data collection, preprocessing, machine learning analysis, and visualization into a unified pipeline.Experimental evaluation demonstrates improved anomaly detection performance with an accuracy of 96.4%, reducing false positives and enabling faster fault diagnosis. The proposed approach enhances system reliability and minimizes mean time to resolution (MTTR) in large-scale computing environments.

Keywords-Log Analytics, BERT, Anomaly Detection, Root Cause Analysis, Machine Learning.

Nishanth V, Kanchana T, Rajamanickam V, Badrinath C, Sathish V (2026). ML Based Intelligent Log Analytics Using BERT. International Journal of Innovative Research in Multidisciplinary Education & Technology (IJIRMET), 11(3).