Aggregating case-based reasoners in ensembles : an approach in support of explanation

Abstract

Among the reasons for the success Case-Based Reasoning (CBR) has achieved in tackling supervised learning problems, is certainly the capability to give a ranking to any case stored in the database depending on its similarity to the query and the subsequent possibility to retrieve a small set of cases to explain the predicted output. Many areas, like medical domains, electronic commerce applications, diagnosis tasks, recommender systems, greatly benefit from this characteristic of CBR.