On K-means

Basic knowledge:

Algorithm Tags: clustering, greedy, partitioning algorithm

Algorithm objective: Make clusters as compact and independent as possible (low coupling, high aggregation)

Evaluation criteria: Sum of squared errors of all objects

Algorithm complexity: O (NKT), n is the number of samples, K is the number of clusters, T is the number of iterations

Algorithm restriction: The mean value of the cluster is defined (the nominal attribute cannot calculate the mean, at which point K-majority can be used)

Algorithm disadvantage: The K value needs to be given, the initial point selection has an effect on the algorithm and is sensitive to noise.

Reference:

Http://www.cnblogs.com/kemaswill/archive/2013/01/26/2877434.html

On K-means