Research on Alarm Log Clustering Based on DBSCAN Algorithm

Abstract

A method of alarm time series preprocessing using DBSCAN and muti-constrained algorithm was proposed. According to the characteristics of original alarm data, the DBSCAN density clustering algorithm is used to divide the original alarm data into multiple alarm events in time dimension and multi-constrain method is used to optimize the parameters of the inputs of DBSCAN. By using the optimal input parameters of DBSCAN and the sliding time window in each time period, the alarm event is extracted from alarm time series. The experimental results show that this method can effectively filter out the noise impact in order to improve the overall quality of actual alarm practice analysis, and use multi-constrain method to effectively improve the overall alarm log analysis.

Keywords

• alarm log analysis
• muti-constrained algorithm
• dbscan clustering
• sliding time window