Enhancing recognition of a weak class --- comparative study based on biological population data mining

  • Authors:

  • Henryk Maciejewski;Ewa Walkowicz;Olgierd Unold;Paweł Skrobanek

  • Affiliations:

  • Institute of Computer Engineering, Control and Robotics, Wroclaw University of Technology, Wrocław, Poland;Department of Horse Breeding and Riding, Wroclaw University of Environmental and Life Sciences, Wrocaw, Poland;Institute of Computer Engineering, Control and Robotics, Wroclaw University of Technology, Wrocław, Poland;Institute of Computer Engineering, Control and Robotics, Wroclaw University of Technology, Wrocław, Poland

  • Venue:
  • ICAISC'12 Proceedings of the 11th international conference on Artificial Intelligence and Soft Computing - Volume Part II
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper presents an overview of several methods that can be used to improve recognition of a weak class in binary classification problem. We illustrated this problem in the context of data mining based on a biological population data. We analyze feasibility of several approaches such as boosting, non-symmetric cost of misclassification events, and combining several weak classifiers (metalearning). We show that metalearning seems counter-productive if the goal is to enhance the recognition of a weak class, and that the method of choice would consist in combining boosting with the non-symmetric cost approach.