Enhancing Cases Retrieval and Adaptation in Case Based Reasoning

Abstract

Case-based reasoning (CBR) is a commonly seen methodology of problem-solving in AI. Just like human reasoning, CBR uses prior cases to find out suitable solutions for new problems. Other problem-solving methodologies of in AI must find out the general relationship between problem situations and problem-solving methods, to construct suitable solutions. Unlike the others, CBR pays much attention to the characteristics of each prior case. It can correctly take advantage of the situations and methods in former cases to can handle unexpected situations. Additionally, CBR can achieve human learning behaviors by constantly adding cases, thus raising the accuracy of problem solutions. CBR has been successfully applied to the areas of planning, diagnosis, law and decision making among others. It uses useful prior cases to solve the new problems. CBR must accurately retrieve similar prior cases for getting a good performance. Many researchers have proposed useful technologies to handle this problem. However, performance of retrieving similar cases in large-scale CBR was seldom been discussed. When the number of cases in the case base becomes large, the processing time for retrieving similar cases rapidly increases. The process of retrieving similar cases thus becomes a critical task of CBR.

Throughout this thesis The Novel Case Base Indexing Model based on Power Set Tree has been introduced. This model is used to enhance the performance of indexing and retrieving in Case Based Reasoning (CBR). We designed and built a solution to find the unique combinations to each case in a Case Base, and use these unique combinations to build the Case Base Index. We get over a lot of unbalanced consumption of resources, finally, we have built a better algorithm to balance the resources consumptions and harness them to serve our purposes in finding the unique combinations for large cases that has more than 38 finding. The main strengths of this model that it is applicable for any domain. The Generated Case Base Index can be used for many purposes beyond only being a Case Base Index. After the completion of this thesis, we have successfully built a complete solution to build the Case Base in our format along with reasoning tool to justify any results and a statistics solution to measure the main difference between the original case base and the case base index. The statistics tool can be also used to find the deleted disorders or tuples which have been removed as part of self-cleansing feature added to our model.