面向电力大数据日志分析平台的异常监测集成预测算法

Ensemble forecasting algorithm for anomaly detection on electric-power big data log analysis platform

随着电力企业网络技术的发展,传统和新生的日志处理系统已不能满足大数据状态下的日志分析要求,为了实现系统日志异常分析的目标,该文提出一种基于时间序列的系统异常数量集成预测算法和面向该算法的评价体系。该算法对多种分类预测算法进行集成,对收集到的日志数据进行分类预测,进而实现了以综合最优的准确度预测系统的异常数量,评价体系很好地支持了该算法的工作,算法增强了日志分析平台的安全性。

In view of that the traditional or the new log processing system can not meet therequirements of the log analysis in the current situation of big data entirely with the development of power enterprise network technology, an algorithm for estimating the number of systems based on time series and the evaluation system are presented to realize the system for the algorithm. The algorithm integrates multiple classification prediction algorithms to classify the collected log data, and then realize the purpose of forecasting the number of anomaly systems with the best accuracy. The evaluation system also supports that the algorithm can increase the security of the log analysis platform.