Research On Two-Stage Case Retrieval Method Of Case-based Reasoning System

Case-Based Reasoning (CBR), an important problem-solving and learning reasoning technology in the field of artificial intelligence, solves current problems by past experience from experts. With continuous development and increasing application of CBR, the scale of case base is gradually expanding, which improves the capacity of solving problems through increasing the amount of knowledge in the case base. In the meantime, the expanding of the case base slows down case retrieval, which affects the efficiency of the Case-Based Reasoning system. Therefore, how to improve the case retrieval speed to improve the efficiency of Case-Based Reasoning system’s operation has become a hot research issue in the field of CBR.In this paper, two-stage case retrieval was used to improve the speed of case retrieval, which ensures the capacity of the case base reasoning system in solving problems, and improves the efficiency of Case-Based Reasoning system. Firstly, some properties were used to complete the first case retrieval, so as to obtain the initial similar result sets. Secondly, exact matches on the initial similar results sets that were obtained in the first step were matched exactly, so as to obtain the case we need.Firstly, the rough sets theory was used to classify all case attributes and to determine each case’s attribute’s weight, aiming at reducing those unimportant attributes. The reduced attributes was used to complete the case retrieval by means of nearest neighbor approach, thus achieving the initial similar results sets.Secondly, targeting at shortage of traditional case retrieval method, the gray correlation analysis was introduced into the case retrieval. By analyzing the strengths and weaknesses of gray correlation analysis, the Euclidean distance and gray correlation analysis were reasonably integrated, and strategy for selecting Grey Cognate Resolution Coefficient was established. Therefore, the exact match on initial similar results sets was completed, and the most similar source case to the target case was successfully obtained.Through the above-mentioned two steps, the calculated amount of the case retrieval was greatly saved, the running time of Case-Based Reasoning system was reduced, the accuracy of case retrieval was ensured, and the efficiency of the Case-Based Reasoning system was promoted.