Asian Journal of Fundamental and Applied Sciences, e-ISSN: 2716-5957, Vol. 3, No. 2, 1-9, 2022

The k-means algorithm is a conventional unsupervised cluster analysis algorithm, which is fast and easy to implement. Still, the number of clusters needs to be defined, and selecting the centre of mass is uncertain. A K-means algorithm based on the combination of maximum-minimum distance and Between-Within-Proportion (BWP) metrics is proposed to overcome these limitations. The results of simulation experiments on three datasets in the UCI database show that the proposed algorithm outperforms both the conventional K-means algorithm and the maximum-minimum distance-based K-means algorithm in terms of accuracy and clustering effect.