Learning Solution Similarity in Preference-Based CBR

This paper is a continuation of our recent work on preferencebased CBR, or Pref-CBR for short. The latter is conceived as a casebased reasoning methodology in which problem solving experience is represented in the form of contextualized preferences, namely preferences for candidate solutions in the context of a target problem to be solved. In our Pref-CBR framework, case-based problem solving is formalized as a preference-guided search process in the space of candidate solutions, which is equipped with a similarity (or, equivalently, a distance) measure. Since the efficacy of Pref-CBR is influenced by the adequacy of this measure, we propose a learning method for adapting solution similarity on the basis of experience gathered by the CBR system in the course of time. More specifically, our method makes use of an underlying probabilistic model and realizes adaptation as Bayesian inference. The effectiveness of this method is illustrated in a case study that deals with the case-based recommendation of red wines.