基于自适应果蝇优化算法的K-means聚类

K-means clustering based on adaptive fruit fly optimization algorithm

K-means聚类算法全局搜索能力较低并且选择初始质心的具有盲目性,果蝇算法具有优越的全局搜素能力但寻优方向不稳定,因此对果蝇算法(FOA)进行改进并以此优化K-means。在模型基础上利用密度标准差选择初始果蝇个体,并且构建寻优目标精度高的适应度函数进性寻优,设计一种准确率高且鲁棒能力强的IFOA-K-means算法。利用4种测试函数对IFOA性能进行对比,同种测试函数下IFOA算法的迭代次数少且搜索效率高。利用UCI标准数据集对IFOA-K-means算法进行对比分析,实验结果表明改进后的算法寻优误差小并且聚类精度高。

K-means clustering algorithm has poor global search ability and blindness in choosing the initial centroid. The fruit fly algorithm has superior global search ability but has the instability of the optimization direction. Therefore, fruit fly optimization algorithm(FOA)is improved and to optimize K-means. On the basis of the model, the initial drosophila individuals are selected by using the density standard deviation, and the fitness function with high accuracy of the optimization target is constructed for progressive optimization, and a clustering algorithm is designed with high accuracy and robustness IFOA-K-means algorithm. Four standard test functions are used to compare the performance of the IFOA. The IFOA algorithm under the same test function has fewer iterations and higher search efficiency. UCI standard data sets is used to compare and analyze the IFOA-K-means algorithm. The experimental results show that the improved algorithm has small optimization errors and higher clustering accuracy.