Attribute Selection for Classificatory Analysis Using a Probabilistic Approach.
Amir Ahmad, Lipika Dey · Indian Conference on Computer Vision, Graphics and Image Processing · 2002
Dimensionality reduction is one of the key data analysis steps. Besides increasing the speed of computation, eliminating insignificant attributes from data enhances the quality of knowledge extracted from the data. In this paper we have proposed an efficient, conditional probability based method for computing the significance of attributes. The algorithm is highly scalable and can simultaneously rank all the attributes. The proposed method can be used to analyze pre-classified data by exploiting the attribute-to-class and class-to-attribute co-relations. The effectiveness of the approach is established through the analysis of various large test data sets. The method can be extended to extract classificatory knowledge from the data.