Design and Implementation of an Improved K-Means Clustering Algorithm

Aiming at the problems of traditional K-means clustering algorithm, such as local optimal solution and slow speed caused by uncertainty k value randomness initial cluster center selection, this paper proposes an improved KMeans method. The algorithm first uses idea elbow rule based on sum squares errors to obtain appropriate number clusters k, then variance a measure degree dispersion samples, selects data points with smallest distance greater than average samples set. Finally, combined “triangular inequality principle,” unnecessary calculation in iterative process is reduced, operation efficiency improved. results show that tested UCI Compared k-means Canopy-KMeans accuracy speedup ratio are significantly improved, quality