Comparative Analysis of Filter-Wrapper Approach for Random Forest Performance on Multivariate Data

Feature selection is the process of selecting the superlative feature from the preprocessed datasets. It is also useful in machine learning to improve the speed as well as to improve the classification accuracy. This paper deals with filter and wrapper approach to identify their pros and cons with respect to decision tree based classification algorithm. Filter and wrapper approach with a best first search method and genetic search method is used with a decision tree based random forest algorithm to compare the classification accuracy. Datasets are taken from the UCI machine learning repository to test the accuracy rate. The results obtained are compared with the existing algorithms and are discussed based on the classification accuracy.