基于模糊逻辑COOT优化K调和均值的数据聚类算法

A Data Clustering Algorithm Based on Fuzzy COOT K-Harmonic Means

关键词:  模糊逻辑  模糊系统  白骨顶鸡优化算法  K调和均值  聚类  收敛性

Key words:  fuzzy logic  fuzzy system  Coot optimization algorithm (COOT)  K-Harmonic Means (KHM)  clustering  convergence

摘要:

针对K调和均值(K-Harmonic Means，KHM)聚类算法易陷入局部最优的不足，本文结合KHM聚类算法的快速局部开发和白骨顶鸡优化算法(Coot optimization algorithm，COOT)的全局勘探能力，提出一种模糊逻辑COOT优化KHM的数据聚类算法(Fuzzy COOT K-Harmonic Means，FCOOTKHM)。将KHM聚类算法生成的初始聚类解输入白骨顶鸡初始种群结构再进行迭代寻优。同时，为了进一步提升COOT的搜索精度，设计模糊逻辑对COOT的收敛因子和领导者种群占比进行自适应调整，均衡算法的搜索与开发能力。使用聚类调和平均值评估种群个体的适应度，结合智能算法启发式搜索对聚类结果迭代寻优。利用加州大学欧文分校(University of California Irvine，UCI)数据库中的7个数据集对FCOOTKHM的聚类性能进行验证分析。结果表明，FCOOTKHM在准确率、精确度、召回率、F度量、Kappa系数和收敛效率等指标上均表现更好，该算法能够实现更精确的数据聚类。

Abstract:

To solve the problem that the clustering algorithm K-Harmonic Means (KHM) is easy to fall into a local optimum,a clustering algorithm Fuzzy COOT K-Harmonic Means (FCOOTKHM) combining the rapid local development capability of KHM and the global exploration capability of Coot optimization algorithm(COOT) is proposed.The initial clustering solution generated by KHM is input as the initial population structure of COOT,and then iterative optimization is carried out.To further improve the search accuracy of COOT,a fuzzy logic is designed to adaptively adjust the convergence factor and leader population proportion of COOT,which can balance the search and development capabilities of the algorithm.The harmonic mean of clustering is used to evaluate the fitness of individual populations and iteratively search for clustering results.Seven datasets of University of California Irvine (UCI) were used to validate the clustering performance of FCOOTKHM.The results showed that the improved algorithm performed better in terms of accuracy,precision,recall,F-measure,Kappa coefficient and convergence speed,which can enable more accurate data clustering.