基于支持向量机免疫集成预测的电信网络性能监控  

Telecom networks performance monitoring based on artificial immune support vector regression

作　　者：郭建华[1] 杨海东[2] 

机构地区：[1]广东技术师范学院计算机科学学院,广东广州510665 [2]广东工业大学机电工程学院,广东广州510090

出　　处：《中南大学学报（自然科学版）》2012年第3期1020-1026,共7页Journal of Central South University:Science and Technology

基　　金：国家自然科学基金资助项目(60973132);广东高校优秀青年创新人才培育项目(LYM09097)

摘　　要：为实现持续有效的电信网络性能监控,提出一种改进的支持向量机预测基线法。利用人工免疫网络优化支持向量机参数、核函数参数、嵌入维数和样本规模等回归分析的自由参数,提出支持向量机免疫集成预测算法。根据电信网络性能的周期性特点构建同点时间序列模型,以预测的置信区间为基线对电信网络性能进行监控,通过对某软交换服务器的CPU负荷进行实验分析。研究结果表明:与经验自由参数相比,支持向量机免疫集成预测算法能取得更加精确的回归模式,其误差平方和减少55.4%,同点时间序列模型能有效克服连续时间序列中存在的异常输入敏感问题,准确发现多个连续的异常点。To continuously monitor telecom network performance, an improved baseline generating method based on support vector machine （SVM） was proposed. An artificial immune support vector regression algorithm （AI-SVR） was presented, which optimized SVM parameters, kernel radius, embedding dimension and sample size by artificial immune network. The same point time series was built according to the cycle oftelecom network performance and the baseline for telecom networks performance monitoring was defined as the predicted confidence interval. Taking the CPU load of a certain softswitch server for analysis, the results show that AI-SVR can obtained a better regression mode than SVR with experienced parameters, the sum of error squares decreases by 55.4%, and the same point time series can overcome the problem that the output is sensitive to the abnormal input when using continuous series, and the monitoring method can find a few continuous anomalies.

关 键 词：电信网络 性能监控 支持向量机 人工免疫系统 时间序列预测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP393[自动化与计算机技术—控制科学与工程]