Abstract

Anomalies are patterns that lack normal behavior. Anomaly detection process can be used to predict changes in input patterns and can prevent different malicious activities. Kernel Principal Component Analysis is an effective, anomaly detection technique. This is an effective technique for non-linear data set. In this, kernel Eigenspace splitting and merging approach is used for predicting the anomalies effectively. Kernel splitting process, can extract smaller KES from larger KES. Here a QR decomposition technique can be used for, KES splitting. This is done to remove the data patterns that no longer fit to the current data distribution.KES splitting is combined with KES update. An adaptive update takes an appropriate sliding window size into consideration and reduces the number of updates required, when a change in the data distribution occurs. Thus an adaptive split-merge KES with QR decomposition, shows a superior performance in predicting anomalies that have a non-linear behavior.

References

• V.Chandola,A.Banerjee,andV.Kumar,“Anomalydetection:A survey,” ACM Comput. Surv., vol. 41, no. 3, pp. 15:1–15:58, 2009.
• V.Barnett and T.Lewis,Outliers in Statistical Data .New York,NY,USA:Wiley,1994,vol.3.
• Jiawei Han,Micheline Kamber and Jian Pei,”Data Mining :Concepts and Techniques”,2nd ed.Morgan Kaufmann,2006.
• P. Hall, D. Marshall, and R. Martin, “Adding and subtracting eigenspaces with eigenvalue decomposition and singular value decomposition,” Image Vis. Comput., vol. 20, no. 13, pp. 1009–1016,2002.
• Kou, Y., Lu, C., & Sinvongwattana, S. Survey of Fraud Detection Techniques Yo-Ping Huang, 749–754. 2004.
• E.Eskin et.al ,A geometric framework for unsupervised anomaly detection ,pp 77-101,2002.
• M. M. Breunig, H.-P. Kriegel, R. T. Ng, and J. Sander, “LOF: Identifying density-based local outliers,” ACM SIGMOD Rec., vol. 29, no. 2, pp. 93–104, 2000.
• H.-P. Kriegel, A. Zimek, and M. Schubert, “Angle-based outlier detection in high-dimensional data,” in Proc. 14th ACM SIGKDD Int. Conf. Knowl. Discovery Data Mining, Aug. 2008, pp. 444–452.
• Wei, Li, et al. "Hot: Hypergraph-based outlier test for categorical data."Advances in Knowledge Discovery and Data Mining. Springer Berlin Heidelberg, 2003. 399-410.
• I. Jolliffe, Principal Component Analysis. Hoboken, NJ, USA: Wiley, 2005.
• Y. Lee, Y. Yeh, and Y. Wang, “Anomaly detection via online oversampling principal component analysis,” IEEE Trans. Knowl. Data Eng., vol. 25, no. 7, pp. 1460–1470, Jul. 2013.
• H.Hoffmann,”Kernel PCA for novelty detection, “Pattern Recognit., vol 40,no. 3,pp.863-874,2007.
• T.-J. Chin and D. Suter, “Incremental kernel principal component analysis,” IEEE Trans. Image Process. vol. 16, no. 6, pp. 1662–1674,Jun. 2007.
• B. Liu, Y. Xiao, P. Yu, Z. Hao, and L. Cao, “An efficient approach for outlier detection with imperfect data labels,” IEEE Trans.Knowl. Data Eng., vol. 26, no. 7, pp. 1602–1616, 2014.
• I. B. Khediri, M. Limam, and C. Weihs, “Variable window adaptive kernel principal component analysis for nonlinear nonstationary process monitoring,” Comput. Ind. Eng., vol. 61, no. 3, pp. 437–446, 2011.
• G. Sharma, S. Chaudhury, and J. Srivastava, “Bag-of-features kernel eigen spaces for classification,” in Proc. 19th Int. Conf. on Pattern Recognit., Dec. 2008, pp. 1–4.
• Colin O’Reilly, Member, IEEE , Alexander Gluhak, and Muhammad Ali Imran, Senior Member, IEEE Adaptive Anomaly Detection with Kernel Eigenspace Splitting and Merging IEEE Transactions on Knowledge and Data Engineering, vol. 27, no. 1, 2015.