CBR for CBR:A Case-Based Template Recommender System for Building Case-Based Systems

摘要

Our goal is to support system developers in rapid prototyping of Case-Based Reasoning (CBR) systems through component reuse. In this paper, we propose the idea of templates that can be readily adapted when building a CBR system. We define a case base of templates for case-based recommender systems. We devise a novel case-based template recommender, based on recommender systems research, but using a new idea that we call Retrieval-by-Trying. Our experiments with the system show that similarity based on semantic features is more effective than similarity based on behavioural features, which is in turn more effective than similarity based on structural features.