A Hybrid Feature Selection Approach for Handling a High-Dimensional Data

Abstract

We proposed a Hybrid Feature selection method, the combination of Mutual Information (MI) a filter method, and Recursive Feature Elimination (RFE) a wrapper method. The methodology combines the strengths of both filter and Wrapper method. Performance of the proposed method is measured on three benchmark datasets (Ionosphere, Libras Movement, and Clean) from the UCI Repository. We compared the classification accuracy of the proposed Hybrid method with MI, RFE, Original Features by using random forest classifier. The performance are compared using four classification measures i.e. 1. F1-Score 2. Recall 3. Precession 4. Accuracy. It is evidence from the result analysis that the proposed hybrid method has out performed other methods.