EGLK-Means: An Improved K-Means Clustering Algorithm with Enhanced Global and Local Information

In this paper, an improved K-means clustering algorithm, EGLK-Means, is proposed, which optimizes the clustering results by enhancing global and local information. The traditional K-means clustering algorithm faces challenges in dealing with complex coupled, high-dimensional and unbalanced datasets, especially in the case of improper selection of the initial center of mass, which significantly affects the final clustering results. The EGLK-Means algorithm optimizes the selection of the initial center of mass by introducing the method of calculating the feature weights to enhance the global information. Meanwhile, the algorithm improves the K-means clustering process by binding mutual nearest neighbors to enhance the local information. The experimental results show that the EGLKMeans algorithm outperforms the standard K-means algorithm in clustering on complex coupled, high-dimensional and unbalanced datasets, and the improvement is more significant especially when dealing with unbalanced datasets.