Using case-based reasoning approach to the support of ill-structured decisions

Case-based reasoning is the most preferred method for problem solving and decision making in complex and dynamically changing situations. Decision makers usually utilize previous cases to help evaluate and justify decisions. Case-based reasoning solves problems by relating previously solved problems or experiences to a current, unsolved problem in a way that facilitates the search for an acceptable solution. Conceptually, case-based reasoning models can be classified as symbolic models, neural networks, and similarity-computational models. These approaches use syntactic patterns in a database of past cases to solve classificatory decision problems. The similarity-computational approach differs from the symbolic approach and neural networks by direct operation on the past cases without using decision trees, rules, or networks as the intermediate structure for problem solving. In this research we propose a similarity-computational reasoning model, and investigate its feasibility to decision support. A performance comparison among our model, a multi-layer neural network, and a symbolic model is also conducted. We view our model as a complementary technique to the traditional rule-based reasoning approaches for the purpose of supporting ill-structured decisions.