A Novel Feature Ranking Criterion for Supervised Interval Valued Feature Selection for Classification

In this paper, a novel feature ranking criterion for selecting interval valued features in supervised environment is introduced. The introduced feature ranking criterion works on uni-variate interval valued data. Each feature is evaluated and associated with a score using the proposed ranking criterion. Subsequently the features are sorted based on their scores. A feature sub-setting is accomplished by considering the top d' features where d' is empirically selected. The introduced feature selection criterion is validated using a suitable symbolic classifier on relatively large dataset of flowers and water dataset. The experimental results show the superiority of the proposed feature ranking criterion stating that it outperforms the state-of-the-art feature selection methods both in-terms of dimension and classification accuracy.