Attribute subset selection by mixed weighting mean classification method

Abstract

The discovery of knowledge from the huge available data is the highest mount setback in practical pattern classification and knowledge discovery problem. The preprocessing of data plays a major role in knowledge discovery as it consequently improves the accuracy of the classifier. One of the preprocessing techniques, attribute subset selection has major importance as the selection leads to better performance of the classifier and the cost of the classification is sensitive to the choice of attributes that used to construct the classifier. This paper proposes a new attribute subset selection method named as Mixed Weighting Mean Classification (MWM-C) method. It evaluates the weights of the available attributes by using 5 major weighting functions such as information gain, information gain ratio, gini index, correlation, chi squared statistic. The five methods are chosen to bias the results of one another. The proposed method is examined on soybean data set and conferred satisfactory results.