Combining Dynamic K-Means and Binary Search Centroid to Optimize Clustering Results on Home Industry Datasets

This study aims to improve the Dynamic k-means algorithm for determining the optimal number of clusters and iteration. Unfortunately, the dynamic k-means algorithm has limitations in determining the centroid point, whereas the binary search centroid algorithm can determine the cluster's centroid point but has limitations in determining the number of clusters. In the case of home industry datasets, it is proposed to optimize clustering results using a combination of Dynamic k-means algorithm and binary search centroid based on the advantages and disadvantages of each algorithm. 3,466 home industries were used as the sample data. The validity of both algorithms was assessed using the Davies Bouldin Index (DBI) methods. In clustering home industry datasets, the minimum value of DBI Cluster algorithm combination of Dynamic k-means and binary search centroid K-means (DK-means with BSCK-means) is the smallest, which is 0.1845, when compared to only Dynamic K-means algorithm. The final result of home industry clustering using a combination of Dynamic K-means and the Binary Search Centroid algorithm yields an optimal value in the fifth iteration, with the acquisition of the beginner cluster C1 is 3224, the developing cluster C2 is 160, and the advanced cluster C3 is 82.