期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2020
卷号：98
期号：18
页码：3807-3814
出版社：Journal of Theoretical and Applied
摘要：Clustering technique is one of the most important tools for knowledge discovery, during which the samples are divided into categories whose members are similar to each other. One of the most common and widely-used clustering solutions is partition-based clustering algorithms such as K-Means and K-Medoids which have attracted a lot of attention in the field of customer clustering. However, in these algorithms, the initial cluster centroids are usually randomly selected from the initial samples, making the final result of the clustering undesirable in most cases. In this research, a solution is proposed for the optimal selection of initial cluster centroids in K-Means algorithm. In the proposed method, the initial cluster centroids are selected based on a heuristic method to provide the input for the clustering algorithm. To evaluate the effectiveness of the proposed method, the K-Means, K-Medoids, and improved K-Means algorithms were tested on a real data set obtained from Central Insurance Company in Iran. According to standard evaluation criteria, the proposed method had a greater impact on improving clustering results than the other two methods.
关键词：Partition-Based Clustering;Customer Clustering;Selection of Cluster Centroid;K-Means;K-Medoids
