基于随机森林算法的数据中心运维异常告警方法

Method of data center operation and maintenance abnormal alarm based on random forest algorithm

已有的数据中心运维异常告警方法由于数据分类结果模糊,无法在运维中心筛选异常数据,导致告警效果欠佳。为解决上述问题,文中提出一种基于随机森林算法的数据中心运维异常告警方法。通过滑窗方法构造数据样本,设定控制参数,以确定数据样本总数,集成数据集,实现数据预处理;然后建立包含多个决策方法的数据分类器,分别计算数据的平均值、极差和标准差,判断样本线性函数,确定特征集合,实现异常特征提取;最后,根据随机森林算法训练计算子集,建立数据决策树,通过数据的流动次数和分类结果实现异常告警。实验结果表明,基于随机森林算法的数据中心运维异常告警方法通过数据分类器能够在短时间内处理异常数据,并判断异常特征,对于异常数据的识别精准度高于90%,与其他识别方法相比,具有很好的识别效果。

The existing data center operation and maintenance abnormal alarm methods cannot filter abnormal data in the operation and maintenance center due to the fuzzy data classification results,resulting in poor alarm effect.To slove above all,a method of data center operation and maintenance abnormal alarm based on random forest algorithm is propsoed.The data samples is constructed by means of sliding window method,the control parameters are set to determine the total number of data samples,integrate data sets,and realize data preprocessing.A data classifier containing multiple decision methods is built to calculate the average value,range and standard deviation of the data,judge the sample linear function,determine the feature set,and achieve anomaly feature extraction.The subsets is trained and calculated according to the random forest algorithm,and the data decision tree is established,so as to realize the abnormal alarms by means of the number of data flows and the classification results.The experimental results show that the data center operation and maintenance abnormal alarm method based on the random forest algorithm can process abnormal data in a short time by means of the data classifier,judge the abnormal characteristics,and the recognition accuracy of abnormal data is higher than 90%.In comparison with other identification methods,this method has a very good recognition effect.