Ensembling of Feature Selection Methods for HIGH DIMENSIONAL DATASET

The feature selection is an important preprocessing step in data mining that helps in increasing the performance of the model. The aim of feature selection is to choose a subset of features with high information and to eliminate the irrelevant features with less or no predictive information. Many researches had been done to improve the performance of a single feature ranking methods, but not so much in the area of combinations of feature ranking methods. In this paper, we propose an ensemble method for feature selection, in which multiple feature selection methods are combined to yield more robust and stable results. We have done our experiment on the four feature ranking methods: ttest, entropy, Wilcoxon and Bhattacharyya methods. Through the experiment, we have shown that the combination of multiple feature ranking methods can outperform the single feature ranking method only if each individual feature selection method has high performance. The dataset we have used for our work is High Dimensional resolution Ovarian Cancer Dataset. We have used MATLAB tool in our work.