Research on selection of initial center points based on improved K-means algorithm

Traditional K-Means clustering algorithm is very sensitive to the initial center point, the selection of the different initial center points will bring about different clustering results, and clustering performance is greatly affected by the initial center point. After the analysis of the characteristics of the initial center point, the selection of the initial point of the K texts as different categories in the text collection makes the sum of the k texts similarity be smallest. In the paper, the selection of the initial center point based on improved K-means algorithm is proposed. Experimental results show that the method effectively reduces the the clustering algorithm iteration process and improves the clustering performance.