Horizontal Case Representation

  • Authors:

  • Rosina Weber;Sidath Gunawardena;Craig Macdonald

  • Affiliations:

  • The iSchool at Drexel, College of Information Science & Technology, Drexel University,;The iSchool at Drexel, College of Information Science & Technology, Drexel University,;The iSchool at Drexel, College of Information Science & Technology, Drexel University,

  • Venue:
  • ECCBR '08 Proceedings of the 9th European conference on Advances in Case-Based Reasoning
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

We present a new case representation that seeks to make case-based reasoning (CBR) more suited to real world applications. We propose a horizontal representation that is composed of two features, one to represent the problem and one to represent the solution. We also present a similarity metric tailored to our representation. Rather than parametrizing the distance function with weights, it requires one parameter that recommends the cardinality of values for new problems to be solved by the system. Our representation is less restrictive during case acquisition as it does not constrain how non-experts can populate cases and it requires less knowledge engineering effort than the traditional method. We compare our representation to the traditional case representation and show that it is superior when cases are incomplete. Finally, we illustrate the effectiveness of our representation in a real world application, where the demarcation between problem and solution is blurred.