摘要
针对通信网告警中预示重大故障的告警数量少、不适合用传统预测方法的特点,提出了一种基于稀疏贝叶斯的通信告警序列预测方法(PBM),并与支持向量机(SVM)预测方法进行了比较。实验结果表明,PBM方法非常适用于小样本的通信告警预测,其不仅具有SVM的预测性能,而且在样本数目增加时的预测误差率要小于SVM,具有非常好的预测精度。
The alarm which indicates major failure in communication networks has a small number,in this case it is not suitable for traditional forecasting methods.This paper proposed a PBM,and compared it to SVM estimate method.The results show that,PBM is applicable to the small sample of telecommunication alarms predict.It not only has the predict performance of SVM,but also has lower predict errors than SVM when the number of samples increase,even it has a very good prediction accuracy.
出处
《计算机应用研究》
CSCD
北大核心
2010年第4期1427-1429,共3页
Application Research of Computers
基金
国家自然科学基金资助项目(60572091)
关键词
告警序列
决策函数
核函数
稀疏贝叶斯
预测精度
alarm sequences
decision function
kernel function
sparse Bayesian
prediction accuracy
