基于改进SVM的电力企业信息系统异常检测方案的优化

近年来,电力企业信息系统的网元数量不断增多,数据处理规模不断扩大,数据类型日益复杂,导致系统的异常事件频发。论文提出采用改进SVM算法对传统异常检测方案进行优化,并通过两步交叉验证的方法提高了SVM参数寻优质量,显著强化了该算法的主动分析和预测能力。结合电力企业信息系统异常的特征,构建了同点时间序列模型,并实现了对网络性能异常点的实时化检测。

Optimization of Abnormal Detection Scheme for Power Enterprise Information System Based on Improved SVM

In recent years,because of the growing number of network elements,the expanding scale of data processing and the increasing complexity of data types abnormal events occur frequently in power enterprise information systems. In this paper,the improved SVM algorithm is proposed to optimize the traditional abnormal detection scheme,and the two step cross validation method is used to improve the quality of SVM parameters optimization,which significantly strengthens the active analysis and prediction ability of the algorithm. Combined with the characteristics of power enterprise information system anomaly,the same point time series model is constructed,and real-time detection of network performance outliers is realized.