A method for feature selection based on the correlation analysis

Feature selection is one of the important issues in the fields of machine learning and pattern classification. The classification ability of features is analyzed from the point of view of correlation and redundancy. Two types of correlation: C-correlation and F-correlation are presented. The C-correlation is applied to identify the relevant features to the category attribute, while the F-correlation is used to measure the redundancy among features. Finally, the dimension of input features is further reduced with the sequential forward search strategy. Thus a method for feature selection based on the correlation analysis of features is derived. The experimental results show that the proposed algorithm is an effective method for feature selection.