A case-based reasoning and inductive learning approach for heuristic search

Knowledge-based problem solvers traditionally merge knowledge about a domain with several heuristics in an effort to control novel problem situations intelligently. In this dissertation, we use an inductive learning algorithm to obtain domain concepts. Then we use a Case Based Reasoning (CBR) approach to solve heuristic search problems using the domain concepts obtained from the inductive learning approach. The main idea behind CBR approach is to make use of the old solutions while solving a new problem. In this research we identify some properties of the heuristic search functions so that CBR approach can be used to solve them. We also perform an analysis to compare the results of CBR approach with traditional heuristic search techniques.; The above methodologies are then applied to a real world engineering problem, namely wastewater treatment design. This design can be viewed as obtaining an optimal sequence of treatment plants to remove the contaminants in the wastewater. We formulated the treatment problem as a heuristic search problem and used the inductive learning and case based reasoning approaches to design the treatment sequence. We also developed a neural network methodology to obtain the optimal sequence of treatment trains and compared the results from the heuristic approach and the neural network approach.