A Novel Effective Distance Measure and a Relevant Algorithm for Optimizing the Initial Cluster Centroids of K-means

The traditional K-means algorithm is very sensitive to the selection of the initial clustering point and the calculation of the distance measure, which is likely to result in the convergence of only partly optimal solutions. An improved k-means algorithm is proposed to solve the problem of unbalanced clustering effect caused by the fact that the first initial clustering centre falls in the non-dense region of the boundary in the initial clustering centre optimisation process. An improved k-means algorithm for initial clustering centres is proposed, namely, the optimal matching algorithm for K-means clustering, and related experimental analysis of the algorithm is carried out. The improved algorithm first selects the initial points of the traditional K-means clustering algorithm and analyses the clustering results. Then, the initial clustering centre selection and distance determination were tested and the clustering effect was evaluated by introducing the contour coefficient. Experiments on both artificial data sets and UCI data sets show that the algorithm can achieve better clustering results. The experimental results indicate that the improved algorithm has a much higher clustering quality than the traditional K-means algorithm and other improved algorithms.