A Comparative Study on Wrapper and Filter Methods for Feature Selection in Data Mining

In recent years, the size of databases has increased enormously. This has led the world to grow interest in the development of tools that can extract knowledge automatically from data. Here, data mining proves to be the useful tool to discover the knowledge from huge data repositories. This requires proper classification methods and algorithms. In a particular dataset, we require minimal set of features that can lead us to the proper classification of instances of the dataset. Here, the need of the good feature selection algorithm comes to existence. In this paper, we perform a comparative study of the two most popular methods used for the feature selection: Wrapper and Filtering method. A number of algorithms exist that fall under these two categories. In this paper relative merits of the respective algorithms and their performance on a number of datasets is analyzed and depicted graphically also.