Efficient Dimensionality Reduction Approaches for Feature Selection

Feature selection is used to eliminate irrelevant and redundant features, which improves prediction accuracy and reduces the computational overhead in classification. This paper presents comparison of 3 methods namely fast correlation based feature selection (FCBF), Multi thread based FCBF feature selection and decision dependent -decision independent correlation (DDC-DIC). These approaches are concerning the relevance of the features and the pair wise features correlation for redundancy checking in order to improve the prediction accuracy and reduce the computation time. The experimental results are tested in weka tool for C4.5 decision tree construction algorithm, which provide better performance for lung cancer, Tic 2000 Insurance company data and breast cancer data sets.