Improved k-means algorithm to quickly locate optimum initial clustering number K

K-means algorithm is often used as a clustering algorithm, but it is vulnerable to the impact of the clustering number k. To eliminate the effect, a method seeking optimum initial clustering number k rapidly is put forward for the k-means algorithm. This method is accomplished by subtractive clustering to determine the optimal initial clustering k. The experiments to the data inside the public database UCI and TE data show that the improved k-means algorithm can eliminate the sensitivity to the initial cluster number k. The clustering speed and precision are improved.