基于ReliefF特征加权的人工蜂群聚类算法  

Artificial Bee Colony Clustering Algorithm Based on ReliefF Feature Weighted

作　　者：龙金辉[1] 叶阳东[2] 

机构地区：[1]河南机电职业学院,郑州451191 [2]郑州大学信息工程学院,郑州450052

出　　处：《火力与指挥控制》2018年第1期52-56,61,共6页Fire Control & Command Control

基　　金：国家自然科学基金(61170223);国家创新基金资助项目(12C26244104350)

摘　　要：针对模糊聚类算法对最优聚类中心的搜索能力偏弱,以及没有度量样本数据各维特征属性对聚类结果贡献度的问题,提出了一种ReliefF特征加权的人工蜂群聚类算法。算法利用人工蜂群算法搜索最优聚类中心,使用ReliefF算法度量数据各维特征权重,分析了各维特征对聚类的不同影响,减弱冗余特征对聚类的干扰,增强有效特征对聚类的贡献,对加权变换后的数据样本进行聚类以提高聚类效果。在UCI数据集上的对比实验结果表明,此算法具有很好的综合性能。In view of the fuzzy clustering algorithm to the optimal clustering center search ability weak and the lack of measured sample data of clustering results contribution of each dimension of feature attribute, An artificial bee colony clustering algorithm based on ReliefF feature weighting is proposed. Algorithm uses artificial bee colony algorithm to search the optimal cluster center, then use the ReliefF algorithm to measure the weight of each dimension of the data, analyzing the impact of different characteristics of each dimension of the cluster, reducing redundancy features for clustering interference, enhancing an effective feature clustering contribution, clustering for the data sample of weighted transform to improve the effect of clustering. Contrast experiment results on UCI data sets show that this algorithm has better comprehensive performance.

关 键 词：人工蜂群算法 模糊聚类算法 RELIEFF 特征加权 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]