A New Kind of Based on the Graph K-Means Clustering Initial Center Selection Algorithm
K-means clustering algorithm is simple and fast, and has more intuitive geometric meaning, which has been widely applied in pattern recognition, image processing and computer vision. It has obtained satisfactory results. But it need to determine the initial cluster class center before executing the k-means algorithm, and the choice of the initial cluster class center has a direct impact on the final clustering results. A selection algorithm is proposed, which based on figure node most magnanimous to determine the initial cluster class center of K-means clustering algorithm. The method compares with the selection algorithm of other initial cluster class center, which has a simple algorithm idea and low time complexity, and it is significantly better than other clustering arithmetic.