基于簇特征的球员跑动大规模数据聚类研究  

Clustering Algorithm for Large-scale Data of Running in Soccer Based on Cluster Features

作　　者：尹倩[1] 

机构地区：[1]安徽体育运动职业技术学院,安徽合肥230051

出　　处：《常州工学院学报》2017年第6期35-39,100,共6页Journal of Changzhou Institute of Technology

基　　金：安徽省高校人文社科重点项目(SK2015A659)

摘　　要：传统聚类算法无法实时处理足球运动员跑动产生的动态增量数据,为此提出一种基于簇特征的大规模跑动数据聚类算法。利用k-means算法对初始数据进行聚类,并保留聚类后各簇特征,当跑动过程产生的增量数据到来时,利用表示原始簇信息的簇特征与增量数据进行增量聚类,避免传统算法因需重新聚类而导致耗时过长的问题。针对k-means算法容易产生概念偏移的现象,利用簇特征快速检测,避免聚类结果不一致。实验结果表明,该算法能快速处理动态增量数据,且聚类结果一致。Traditional clustering algorithms cannot deal with the dynamic incremental data of running in soccer in real time.The clustering algorithm for large-scale data of running in soccer is put forward based on cluster features.Firstly,initial clustering is performed by k-means algorithm.Then the cluster features of each cluster are saved.Finally,when incremental data are generated by running in soccer,they are incrementally clustered with those cluster features representing the original cluster information,which solves the problem that the traditional k-means algorithm is slow because of repeated re-clustering.Because k-means algorithm tends to produce concept drifting,cluster features are used to realize rapid detection and avoid inconsistent clustering results.The experiment results show this algorithm can deal with dynamic incremental data quickly with consistent clustering results.

关 键 词：K-MEANS聚类 跑动数据 增量数据 概念偏移 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]