为提高K-means聚类算法在异常检测中的效果,以聚类分析为主线,针对传统的K_means聚类算法在初始聚类中心点选择随机性和K值预先设定的问题,提出了一种改进的K_means聚类分析算法,算法引入密度参数和距离理论。依据密度理论和最大距离找出k个初始化中心点。并对算法进行仿真实验,实验证明,新的算法具有良好的效果。 In order to improve the effect of K-means clustering algorithm in anomaly detection, taking clustering analysis as the main line, aiming at the problem that the traditional K_means clustering algorithm chooses randomness and K value in the initial clustering center, K_means improved cluster analysis algorithm, the algorithm introduces density parameters and distance theory. According to the density theory and the maximum distance to find k initialization centers. And the algorithm is simulated experimentally, the experiment proves that the new algorithm has good effect.