Incremental dynamic case-based reasoning-MAS: From static to dynamic cycle

Abstract

In this paper we present our approach in the field of Case-Based Reasoning (CBR). This approach is based on the reuse of previous traces that are similar to the current situation in a dynamic way. Several approaches have been used in this area, but they suffer from some limitations in automated realtime management dynamic parameters (the designer should be defined beforehand the different situations possible).We propose a multi-agent multi-layer architecture based on Incremental Dynamic Case-Based Reasoning (IDCBR) able to study dynamic situations (recognition, prediction, and learning situations). We propose a generic approach able to learn automatically from their experiences in order to acquire the knowledge automatically. Based on the Case-Based Reasoning and multi-agent paradigm, we propose a modification of the static CBR cycle in order to introduce a dynamic process of Case-Based Reasoning based on a dynamic similarity measure able to evaluate in real time the similarity between a dynamic situation (target case) and past experiences stored in the memory (sources case) in order to predict the target case in the future.