K-means Algorithm Based on Improved Krill Herd Algorithm and Calinski-Harabasz Index

Aiming at the problems that Krill Herd (KH) algorithm is easy to fall into local optimum, searchability weak, and k-means easily affected by selection of initial clustering centre, a based on improved KH proposed. The initialized chaos, dynamic clustering, elite leadership random mutation strategies improve introduce optimal number adaptive mechanism, which enhances comprehensive optimization ability algorithm. Six benchmark functions test effectiveness was tested verified with UCI machine learning artificial datasets. verification results showed ensuring faster convergence speed. Compared other algorithms, performance this has been significantly in all aspects.