摘要： K-Means 和 DBSCAN算法初始聚类中心的选择对数据挖掘结果的影响较大。针对上述问题，利用信息熵改进初始聚类中心选择方法，提高数据挖掘效率。将改进的K-Means算法与DBSCAN算法结合应用于入侵检测系统，对一个通用检测记录集进行异常检测测试，实验结果证明了该方法的有效性。

关键词: 入侵检测系统, 数据挖掘, 异常记录, 聚类算法

Abstract: How to select original clustering cores of K-Means and DBSCAN is important to the result of data mining. Aiming at the problem, this paper improves the method of selecting original clustering cores via entropy. It applies improved K-Means and DBSCAN to the intrusion detection system, and does anomaly detection test on a common set of records in the system. Experimental result proves that the method is effective.

Key words: intrusion detection system, data mining, anomaly record, clustering algorithm

中图分类号: 

• TP309.2