K_中心点聚类算法优化模型的仿真研究.pdfVIP

28 1 2011 1 : 1006- 9348( 2011) 01- 0218- 04 K- 1, 2 3 白 旭 , 靳志军 ( 1. , 050016; 2. , 050016; 3. , 050011) : , , , K - , K - , K- , , , , K - , m , m k, , , K- , , K- , : ; ; ; ; : TP393 : A Research on Optim alM odel of K - M eans C lustering A lgorithm BA IXu 1, 2, JIN Z i- jun3 ( 1. College of P ys ics Science and Inform ation Engineering, H ebeiNorm al University, S ijiaz uang H ebei 050016, C ina; 2. H ebeiAdvanced T in F ilm s Laboratory, S ijiaz uang H ebei 050016, C ina; 3. H ebei Province L ibrary, S ijiaz uang H ebei 050016, C ina) ABSTRACT: C lustering is an mi portant researc direction and t eK - center algorit m w ic is efficient and flexible to andle sm all data sets, is one of t em ost mi portant m et od. B ecause its in itial data center is c osed random ly in K- center algorit m, its main draw back is t at it often gets trapped in local optmi um and t e com puting cost is too ig for large data sets. In order to mi prove t e quality of in trusion detection, an mi proved anom aly detection algo- rit m is proposed in t is paper. Firstly, t e training data set is converted to t e standard un it features metric space, t en t e mi proved algorit m is used to d ivide t e data in order to find t e clustering center. In t e end of t is paper, t e mi proved algorit m is analyzed and com paredw it old algorit m. Expermi ental results s ow t at t e mi proved al- gorit m as good stability an