Абстрактный

A Comparison of Filter and Wrapper Approaches with Data Mining Techniques for Categorical Variables Selection

Bangsuk Jantawan, , Cheng-Fa Tsai

The purpose of this study is to evaluate the most important features of graduate employability in higher education database, in attempt to measure the employability situation for graduate information of the Maejo University in Thailand. The experiment also applies the features selection methods to increases the overall efficiency of classification model. There are two general attribute selection approaches: the Filter approach and the Wrapper approach. The Filter approach includes 3 methods, including Information Gain, Gain Ratio and Chi-square. The Wrapper approach we used Search method consisting of Genetic Search, Best First search and Greedy Stepwise as random search approach for subset generation, wrapped with different bayesian classifiers namely Naïve bayes, Bayes network with K2 algorithm, Bayes network with TAN algorithm and Bayes network with Hill-climber algorithm. The results illustrate, employing feature subset selection using proposed wrapper approach has enhanced classification accuracy.