Toward the Evaluation of Case Base Maintenance Policies Under the Belief Function Theory

摘要

The life cycle of Case-Based Reasoning (CBR) systems implies the maintenance of their knowledge containers for reasons of efficiency and competence. However, two main issues occur. First, knowledge within such systems is full of uncertainty and imprecision since they involve real-world experiences. Second, it is not obvious to choose from the wealth of maintenance policies, available in the literature, the most adequate one to preserve the competence towards problems' solving. In fact, this competence is so difficult to be actually estimated due to the diversity of influencing factors within CBR systems. For that reasons, we propose, in this work, an entire evaluating process that allows to assess Case Base Maintenance (CBM) policies using information coming from both a statistical measure and a competence model under the belief function theory.