Fast Means: Enhancing the K-Means Algorithm by Accelerating its Early Classification Version.

In this paper the Fast-Means algorithm, an enhanced version of the K-Means algorithm; inspired by the Early Classification algorithm; is presented. After analyzing the Early Classification algorithm it was found that the higher centroid displacement at each iteration can be used as mean to define whether an object remains in the same cluster for subsequent iterations, also we found that the reduction strategy can be applied from the beginning to the end of the algorithm execution unlike the Early Classification algorithm where the strategy is applied from the second to the last iteration. Results show that the Fast-Means outperforms the standard version because it reaches a time reduction up to 99.02% with only a quality reduction of 7.62% for the Transaction dataset. Additionally, Fast-Means compared to Early Classification reached an improvement in time reduction up to 41.92% with only a difference of 2.73% in quality reduction. [ABSTRACT FROM AUTHOR]