Information theory vs. correlation based feature ranking methods in application to metallurgical problem solving

  • Authors:

  • Marcin Blachnik;Adam Bukowiec;Mirosław Kordos;Jacek Biesiada

  • Affiliations:

  • Silesian University of Technology, Electrotechnology Department, Katowice, Poland;Silesian University of Technology, Electrotechnology Department, Katowice, Poland;University of Bielsko-Biała, Department of Mathematics and Computer Science, Bielsko-Biała, Poland;Silesian University of Technology, Electrotechnology Department, Katowice, Poland

  • Venue:
  • ICAISC'10 Proceedings of the 10th international conference on Artificial intelligence and soft computing: Part I
  • Year:
  • 2010

Quantified Score

Hi-index 0.00

Visualization

Abstract

Feature selection is a typical stage of building any classification or regression model. There are several approaches to it, however one of the fastest is based on determining the relevance of each feature independently by calculating ranking values. In this paper we provide empirical comparison of four different ranking criteria that belong to two different groups information theory and correlation metrics. The comparison is performed on the empirical datasets obtained while building a model used for predicting mass of chemical compounds necessary to obtain steel of predefined quality.