Abstract Views :198 |
PDF Views:3
Authors
Affiliations
1 Manonmaniam Sundaranar University, Tirunelveli, IN
Source
Data Mining and Knowledge Engineering, Vol 6, No 7 (2014), Pagination: 273-281
Abstract
In this paper a novel approach is proposed for the exploration of datacubes in large multidimensional schema. The attributes are categorized as nominal and numerical attributes. The numeric attributes are ranked based on Principal Component Analysis (PCA) method; the nominal attributes are ranked based on Information Gain. Here, multidimensional scaling is applied to find the semantic relationships among the values for each nominal attribute. The informative datacubes are constructed based on the highly ranked facts and dimensions. Importance measure is calculated to find the interestingness measures among the association rules. The interested rules are classified based on IRCA algorithm. The proposed algorithm categorizes the interested rules into highly interested, medium interested and low interested rules. The proposed approach selects a subset of informative dimensions and facts from an initial set of dataset. The results are tested on real world datasets taken from the Bank Loan dataset. The proposed method provides accurate prediction and consumes less computing time when compared with the existing method.
Keywords
Association Rules, Datacubes, Data Mining, Multidimensional Schema, Information Gain, Principal Component Analysis (PCA).
