基于孤立森林和决策树的服务器故障检测  

Server Failure Detection Based on Isolation Forest and Decision Tree

作　　者：李文铖 戴新发[1] 徐新羽 LI Wencheng;DAI Xinfa;XU Xinyu(Wuhan Digital Engineering Institute,Wuhan 430205)

机构地区：[1]武汉数字工程研究所,武汉430205

出　　处：《计算机与数字工程》2024年第11期3386-3389,3397,共5页Computer & Digital Engineering

摘　　要：服务器是现代互联网的重要基础设施,其稳定运行关系到各类应用系统的可靠性,对其进行健康管理有着重大意义,而故障检测是服务器健康管理的重要工具。服务器故障检测通过采集服务器状态并对采集数据进行分析,检测出服务器的异常状态。由于服务器采集的数据中,正常数据占绝大部分比例,数据极度不平衡,常见的分类算法难以收敛。因此,研究了一种基于孤立森林和决策树的服务器故障检测方法。依赖孤立森林算法检测出服务器状态数据中的异常点,然后对检测出的异常点通过决策树算法进行分类,得到故障类型。实现服务器的故障检测。Server is an important infrastructure of modern Internet.Its stable operation is related to the reliability of various ap-plication systems,and its health management is of great significance.Fault detection is an important tool for server health manage-ment.The server fault detection detects the abnormal state of the server by collecting the server state and analyzing the collected da-ta.Because normal data accounts for most of the data collected by the server,the data is extremely unbalanced,and common classi-fication algorithms are difficult to converge.Therefore,a server failure detection method based on isolation forest and decision tree is studied.Relying on the isolated forest algorithm,this paper detects abnormal points in the server status data,and then classifies the detected abnormal points through the decision tree algorithm to obtain the fault type.Server failure detection is implemented.

关 键 词：孤立森林 决策树 故障检测 

分 类 号：TP368.5[自动化与计算机技术—计算机系统结构]