From fuzzy association rule mining to effective classification framework

  • Authors:

  • Osama Alhawsawi;Mayad AL-Saidi;Michael Phi;Tamer N. Jarada;Mohammad Khabbaz;Negar Koockakzadeh;Keivan Kianmehr;Reda Alhajj;Jon Rokne

  • Affiliations:

  • Computer Science Dept, University of Calgary, Calgary, Alberta, Canada;Computer Science Dept, University of Calgary, Calgary, Alberta, Canada;Computer Science Dept, University of Calgary, Calgary, Alberta, Canada;Computer Science Dept, University of Calgary, Calgary, Alberta, Canada;Computer Science Dept, University of British Columbia, Vancouver, BC, Canada;Computer Science Dept, University of Calgary, Calgary, Alberta, Canada;Dept of Elect. & Comp. Eng., University of Western Ontario, London, ON, Canada;Computer Science Dept, University of Calgary, Calgary, Alberta, Canada and Department of Computer Science, Global University, Beirut, Lebanon;Computer Science Dept, University of Calgary, Calgary, Alberta, Canada

  • Venue:
  • IDEAL'11 Proceedings of the 12th international conference on Intelligent data engineering and automated learning
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

Given a set of known classes, classification is a two steps process which uses part of the data to build a model capable of determining the class of new objects not used in the training phase. The accuracy of the classifier is one of the main criteria to judge its usefulness. However, most of the existing classification approaches decide on a single class for a given object. We argue that fuzzy classification is more attractive because it is closer to the real case where it is hard to identify a unique one class per object. To tackle this problem, we developed a framework which produces fuzzy association rules and uses them to build the classifier model. There are two important factors to consider: the method to create fuzzy association rules must be accurate, and the method to build a classifier must be accurate as well. In this paper, we will describe a method to perform fuzzy association rule mining and classification and we will test our results based on numerous factors including accuracy, varying levels of support and confidence.