An Alternative Metric to Support Case Base Maintenance in Case-Based Reasoning Processes

Abstract

Retention is one of the final steps of Case-Based Reasoning approach. Defining the criteria for case retention and maintaining the case base is a complex process. New studies have emerged in the area of nutrition and physical activity recommendation systems following the case-based reasoning cycle. However, the retention phase has only been superficially analysed. Typically, the success of recommendations is evaluated using subjective criteria, such as experts or users’ opinions. Moreover, there are few established processes to prevent uncontrolled growth of a case base. The main innovation of our work lies in the fact that an objective and scientifically approved metric is used for quantifying the success of the generated recommendations. Our research shows that by using well-being indexes, it is possible to define reliable mechanisms for deciding about the inclusion of new cases in the case base, without the need for external intervention. On the other hand, we also analysed the probability with which each case from the case base leads to successful solutions. Thus, solutions that have a low probability can be eliminated. This approach ensures continuous learning of the system without significantly affecting memory, and the cases that contribute most to cardiovascular well-being are kept in the case base.