Volume : I, Issue : IX, June - 2012

Hybrid Attribute Selection Process for Decision Tree Based Classification Algorithms

Mr. A. Jebamalai Robinson, Mrs. S. C. Punitha, Dr. P. Ranjit Jeba Thangaiah

Abstract :

Feature selection can be defined as techniques available for reducing inputs to a manageable size for easy and fast processing and analysis. It is a mandatory step in all data mining tasks involving large datasets. In the present study, a novel method of selecting significant features is proposed which combines the result of three frequently used traditional feature selection algorithms. The three algorithms are correlation–based method, consistency subset evaluation method and Wrapper Subset evaluation method. The result of feature selection is analyzed using two classifiers, namely, J48 and Random Forest. Experimental results with respect to number of features selected and classification accuracy proved that the proposed method is efficient in selecting features that are most important for classification.

Keywords :

Article: Download PDF   DOI : 10.36106/ijar  

Cite This Article:

Mr. A. Jebamalai Robinson, Mrs. S. C. Punitha, Dr. P. Ranjit Jeba Thangaiah Hybrid Attribute Selection Process for Decision Tree Based Classification Algorithms Indian Journal of Applied Research, Vol.I, Issue.IX June 2012

Number of Downloads : 265

References :