支持向量聚類（SVC）是在支持向量機的思想上發(fā)展而來一種聚類方法，針對其處理大規(guī)模數(shù)據(jù)集速度緩慢的缺點，提出了一種改進(jìn)的分塊支持向量聚類算法。改進(jìn)的算法分為三個階段：前期的預(yù)處理，中期的分塊算法，后期的改進(jìn)標(biāo)類算法。提出的方法顯著加快了SVC 的速度，在保持原來SVC 算法的優(yōu)點的基礎(chǔ)上，對大規(guī)模數(shù)據(jù)集以及非均勻分布數(shù)據(jù)集等具有良好的效果。將其應(yīng)用到網(wǎng)絡(luò)入侵檢測，實驗結(jié)果表明改進(jìn)的算法行之有效。
關(guān)鍵詞：支持向量聚類；數(shù)據(jù)挖掘； 網(wǎng)絡(luò)安全；入侵檢測

A Improved Algorithm of Chunking Support Vector Clustering and Its Application in Intrusion Detection ZHANG A-pin， XU Bao-guo （School of communication and engineering, Southern Yangtze University, Wuxi 214122,China） Abstract: The Support Vector Clustering (SVC) algorithm is inspired by support vector machines. Aiming at its defection of slowly training of large-scale sets, an improved algorithm of chunking SVC is proposed to speed the algorithm. The improvement contains three stages: firstly pre-processing , secondly chunking algorithm and finally improved cluster labeling method. The proposed method strongly enhanced the SVC. Keeping the advantages of original SVC, it has an excellent performance to deal with large-scale sets as well as unevenly distributed data sets. Applied the algorithm to network intrusion detection, the result shows that it’s acceptable.
Key words: support vector clustering (SVC); data mining; network security; intrusion detection