Abstrakt

Integration of Subspace Clustering and Action Detection on Financial Data

Meenu Mathai, Mrs.P.Sumathi

Object, attribute and context information are linked in the dimensional data models. Cluster quality is decided with domain knowledge and parameter setting requirements. CAT Seeker is a centroidbased actionable D subspace clustering framework. CAT Seeker framework is used to find profitable actions. Singular value decomposition, numerical optimization and D frequent itemset mining methods are integrated in CAT Seeker model. CAT Seeker framework is improved with optimal centroid estimation scheme. Intra cluster accuracy factor is used to fetch centroid values. Inter cluster distance is also considered in centroid estimation process. Dimensionality analysis is applied to improve the subspace selection process. Experimental results on financial data show that CATSeeker with optimal centroid significantly outperforms all the competing methods in terms of efficiency, parameter insensitivity, and cluster usefulness.