Visualization Of Case Base Properties

Case-based reasoning is a widely employed problem solving technique, which is successfully used in many areas of industry and science for diverse tasks. The number and complexity of CBR systems increase rapidly, and creating tools for development, maintenance and evaluation of these systems becomes very important. Visualization can provide valuable assistance in completing these tasks. In this Master\u92s thesis a tool for simultaneous visualization of two major components of a CBR system, case base and similarity metric, has been developed. It uses problem-solution regularity as the criterion for quality assessment of similarity knowledge. Combining clustering of the case base with calculation of local problem-solution regularity and visual representation allows making accurate judgments about the state of both components of the system. The performance of the system can be periodically monitored, and the source of potential flaws or insufficiencies can be located more precisely. User can interactively analyze a CBR system in different modes and perform changes to the case base. The visualization routines were integrated into a CBR framework making it universally applicable to systems defined in the framework and also providing a uniform interface to these systems. The functionality and usefulness of the visualization tool were demonstrated in tests performed with the implemented route planning domain