Inducing Parameters of a Decision Tree for Expert System Shell McESE by Genetic Algorithm

There exist various tools for knowledge representation, modelling, and simulation in Artificial Intelligence. We have designed and built a software tool (expert system shell) called McESE (McMaster Expert System Environment) that processes a set of production (decision) rules of a very general form. Such a production (decision) set can be equivalently symbolized as a decision tree. In real life, even if the logical structure of a production system (decision tree) is provided, the knowledge engineer may be faced with the lack of knowledge of other important parameters of the tree. For instance, in our system McESE, the weights, threshold, and the certainty propagation functions – all of these are a part of the machinery handling the certainty/uncertainty of decisions – have to be designed according to a set of training (representative) events, observations, and examples. One possible way of deriving these parameters is to employ machine learning (ML) or data mining (DM) algorithms. However, ‘traditional’ algorithms in both fields select immediate (usually local) optimal values – in the context of a whole decision set such algorithms select optimal values for each rule without regard to optimal values for the whole knowledge base. Genetic algorithms comprise a long process of evolution of a large population of objects (chromosomes) before selecting (usually global) optimal values, and so giving a ‘chance’ to weaker, worse objects, that nevertheless may prove to be optimal in the context of the whole knowledge base. In this methodology case study, we expect that a set of McESE decision rules (or more precisely, the topology of a decision tree) is given. The paper discusses a simulation of an application of genetic algorithms to generate parameters of the given tree that can be then used in the rule-based expert system shell McESE.