A Feature Selection Method Using Effective Range and Class Separation
4 Pages Posted: 22 Mar 2019
Date Written: March 22, 2019
Abstract
Feature selection is an important activity in data mining that can contribute significant improvement in running time as well as in eliminating redundancy and ambiguity in data. Also, careful consideration is to be made in selecting features as it might cause information loss if applied incorrectly. There are many areas where feature selection mechanisms are applied. It can also be applied to recognize the disease causing genes in microarray gene expression datasets. While many feature selection techniques exists today, there is still scope of further improvement by applying domain specific methods, information and knowledge. In this thesis, we are presenting a modified statistical method for feature selection based on the degree of differentiation among classes achieved by a feature using effective range of classes. After experimenting the proposed algorithm with different microarray datasets and comparing the performance with other algorithms like RELIEFF, Mutual Information, and Effective Range based Gene Selection (ERGS) etc. it can be observed that the proposed algorithm is a promising technique for gene selection. It cannot be certainly said which method is better for gene expression datasets. Further experimentation with more datasets is needed to verify the reliability of the proposed algorithm.
Suggested Citation: Suggested Citation