A Hybrid Feature Selection Approach Based on Statistical and Wrapper Methods

Gene expression datasets contain a lot of gene expressions. It is very important to identify only relevant genes on huge datasets. Thus, feature selection is very important process. In this study, it is suggested a method which obtains both fast and high classification accuracy. For this reason, it is suggested a method which uses together statistical methods and wrapper methods. The experiments are repeated 10 times to obtain reliable results. According to the results obtained, the proposed method obtains only with 15 features 95.14% classification accuracy using support vector machines. The results are compared with existing methods and methods in the literature. The proposed method gives more successful results.