基于时间序列的局部离群数据挖掘优化算法  

Optimization Algorithm for Time Series-Based Local Outlier Data Mining Optimization Algorithm

作　　者：姚红[1] 梁竹 YAO Hong;LIANG Zhu(Chengdu college of University of Electronic Science and Technology of China,Chengdu Sichuan 611731,China;Chongqing University of Posts and Telecommunications,College of Computer Science and Technology,Chongqing 400065,China)

机构地区：[1]电子科技大学成都学院,四川成都611731 [2]重庆邮电大学计算机科学与技术学院,重庆400065

出　　处：《计算机仿真》2024年第3期514-518,共5页Computer Simulation

基　　金：中国高校计算机教育MOOC联盟项目(B190205);电子科技大学成都学院2021年国腾创投基金项目(GTJG-04);2019年(第二批)百度支持教育部产学合作协同育人项目(2022015PC02470)。

摘　　要：针对数据量较大和数据维度较高导致离群数据挖掘困难的问题,提出基于时间序列的局部离群数据挖掘优化算法。将角度优化的全局嵌入算法和共同核主成分分析法相结合构建AOCKPCA降维算法,对海量高维时间序列降维处理;在蚁群算法中引入K-means算法,提升蚁群算法运算效率,降低不稳定性;将降维后的时间序列输入到优化后算法中,实现局部离群数据挖掘。实验结果表明,采用所提方法挖掘离群数据的准确率较高,误判的离群点个数较少,说明其挖掘效果较好。A local outlier data mining optimization algorithm based on time series is proposed to address the difficulty of outlier data mining caused by large data volumes and high data dimensions.Firstly,the global embedding algorithm based on angle optimization was combined with the common kernel principal component analysis method to form an AOCKPCA dimensionality reduction algorithm for reducing the dimensionality of massive high-dimensional time series.Then,the K-means algorithm was introduced to improve the computational efficiency of the ant colony algorithm and reduce the instability.Finally,the time series after dimension reduction was input into the optimized algorithm to realize the local outlier data mining.The experimental results show that the proposed method has high accuracy in mining outlier data and a smaller number of misjudged outliers,indicating that its mining effect is better.

关 键 词：时间序列 局部离群数据挖掘 数据降维 蚁群算法 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]