An outlying paraphrase subspace search algorithm for outlier detection

作者：Lei, Dajiang; Zhu, Qingsheng^*; Yang, Peng; Jin, Yifu; Chen, Jun; Lin, Hai

来源：International Journal of Digital Content Technology and Its Applications, 2011, 5(8): 355-364.

DOI：10.4156/jdcta.vol5.issue8.41

• Analysis system
• Attribute sets
• Outlying reductions
• Power graphs
• Search Algorithms
• Searching algorithms
• Threshold-value

Most outlier detection techniques focus on detecting outliers in data while detected outliers are lack of deepen analysis and interpreting. Thus, this paper is devoted to finding meaningful attribute subspace and analyzing its intentional knowledge. In this paper, we utilize the concept of rough set to propose the definition of outlying partition and reduction by analogy and construct an outlier detection and analysis system. Based on the definition of outlying reduction, we present the definition of outlying paraphrase subspace which can analyze and paraphrase outliers, in addition, we can mine outliers based on the outlying paraphrase subspace rather than on the full dimensional attribute set. Based on the previous definition, we propose an outlying paraphrase subspace search algorithm with power graph theory. The algorithm searches all attribute subspaces created based on power graph theory and prune the attribute subspaces which produce the similarity of outlying partitions that is larger than threshold value given as well as prune attribute subspaces in the next layer produced by power graph. After searching algorithm ends, we acquire the optimal outlying reduction, called as outlying paraphrase subspace. We experimentally show that the proposed algorithm has feasibility and efficiency for searching outlying paraphrase subspace than the previous methods.

• 出版日期2011