基于免疫克隆算法的LVQ聚类算法权值优化

Optimization of Weights of LVQ Clustering Algorithm Based on Immune Clonal Algorithm

学习矢量量化(LVQ)聚类算法存在严重的对初值敏感的问题,若初值的选择偏差太大,就不会产生好的聚类效果,致使聚类精准度不够。免疫克隆算法具有很强的群体搜索能力,将免疫克隆算法用于优化LVQ聚类算法的初值,并将改进得到的聚类算法用于对IRIS数据集进行分类。分类结果与标准的LVQ算法的比较表明,改进后的聚类算法在稳定性上有了较大幅度的提高。

A major problem of learning vector quantization（LVQ） clustering algorithm is its sensitivity to the initialization, affecting the clustering precision. This paper fist introduced the theory of LVQ clustering algorithm, then used immune clonal algorithm to optimize the initial values of LVQ, And the paper used IRIS data to test this new method. Comparing the evolving LVQ with the standard LVQ, the experiment results indicate thatthe approach of LVQ based on immune elonal algorithm has obvious stability to initial weights.