Title: Data stream outlier detection approach based on frequent pattern mining technique

Authors: Aiman Moyaid Said; P.D.D. Dominic; Ibrahima Faye

Addresses: Department of Computer and Information Science, Universiti Teknologi PETRONAS, 31750, Tronoh, Perak, Malaysia ' Department of Computer and Information Science, Universiti Teknologi PETRONAS, 31750, Tronoh, Perak, Malaysia ' Department of Fundamental and Applied Science, Universiti Teknologi PETRONAS, 31750, Tronoh, Perak, Malaysia

Abstract: The discovery of the rare data points with distinctive characteristics is one of the significant analysis tasks in data mining. This paper concentrates on the detection of outliers in data stream using frequent pattern mining technique. An outlier measurement is presented and an adaptive method for finding outliers in stream of data is introduced. The results of the empirical studies proved that the proposed approach is effective in detecting outliers' data points. The accuracy comparisons confirmed that the proposed approach is as effective as existing static outlier approach and it outperformed the existing dynamic outlier approach. Moreover, the sensitivity of the proposed approach to the change of data distribution was shown to be effective.

Keywords: data mining; data stream outliers; frequent pattern mining; concept drift; static outlier detection; dynamic outlier detection; FPstream; true positive rate; TPR; false positive rate; FPR.

DOI: 10.1504/IJBIS.2015.070892

International Journal of Business Information Systems, 2015 Vol.20 No.1, pp.55 - 70

Published online: 21 Jul 2015 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article