基于并行深度孤立森林算法的水质异常数据检测  

Design and implementation of water quality early warning system based on isolated forest algorithm

作　　者：王海龙 李东波[1] 吴绍锋[1] Wang Hailong;Li Dongbo;Wu Shaofeng(School of Mechanical Engineering,Nanjing University of Science and Technology,Jiangsu Nanjing,210094,China)

机构地区：[1]南京理工大学机械工程学院,江苏南京210094

出　　处：《机械设计与制造工程》2022年第8期83-88,共6页Machine Design and Manufacturing Engineering

摘　　要：为了减少深度孤立森林算法相较于普通孤立森林算法和其他异常检测算法所增加的时间开销,提出将孤立森林算法并行化来提升深度孤立森林算法在级联森林阶段的训练效率,结果表明,相较于普通的深度孤立森林算法,并行化后的深度孤立森林算法节省了一定的时间开销,能够满足水质异常数据检测所需要的实时性和准确性。In order to ensure the accuracy and reliability of water quality anomaly data detection,a deep isolated forest algorithm is adopted in this paper.The experiments show that compared with the ordinary isolated forest algorithm and other anomaly detection algorithms,the deep isolated forest algorithm has a certain improvement in accuracy,but increases the time cost at the same time.Therefore,in order to reduce the increased time cost of deep isolated forest,this paper proposes to parallelize the isolated forest algorithm to improve the training efficiency of deep isolated forest algorithm in the cascade forest stage.The results show that the parallelized deep isolated forest algorithm saves some time and cost compared with the ordinary deep isolated forest algorithm.Considering comprehensively,the proposed parallel deep isolated forest algorithm can well meet the real-time and accuracy of water quality anomaly data detection.

关 键 词：孤立森林 深度孤立森林 异常检测 水质数据 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论]