A novel method for adaptive determination clusters number based on N-order nearest neighbor

For problem of determining clusters number in K-Means, it was proposed a method for adaptively determining value of K based on N-order nearest-neighbor analysis. By calculation of N-order shortest distance between neighbors of various points, it was analyzed the distribution of current random sample data and get all step-points through numerical calculation to determine final K adaptively in case of lacking priori knowledge. Theoretical analysis and numerical simulation results show the effectiveness and reliability, it always has role of guidance for good results of clustering in K-Means.