Application of An Improved K-means Clustering Algorithm in Intrusion Detection

For the initial clustering center usually choose the randomness of the problem, the pa-per proposes a new initial clustering center selection method. First, the algorithm calculates the Euclidean distance of all data to the origin of the coordinate, and then evenly divide the k class, at last, the average value of each class is calculated, and the k center is selected by this method. And through the experimental comparison of the improved algorithm with the merits of the original algorithm and the improved k-means algorithm has been proposed. The experimental results show that the improved algorithm greatly improves the stability and the computation efficiency of the algorithm.