Induction of decision making rules for context sensitive systems

Publisher Website
Google Scholar
Abstract
System design and analysis of highly context-setrsitive systems is both a difficult and time consuming problem. An induction program is discussed that greatly mitigates this problem. The Operational Evaluation Modeling (OpEM) Induction program receives, as input, a case file generated by an OpEM discrete event simulation program. Each case consists of a decision fact plus all knowledge base facts available for this decision (i.e., the decision context). The OpEM induction program analyzes this set of cases and produces an optimal set of rules that decides all of these cases correctly. An OpEM directed graph model is presented that describes the complex, context-sensitive parallel processes of a single-track railroad system, and a Pascal simulation of this railroad system is described to demonstrate that effective decision rules can be induced from extracted expert knowiedge obtained from simulation generated cases. A description of the OpEM induction program is provided, and rules generated by it are compared with rules generated by Ross Quinlan's ID3 Induction prograin using the saine set of cases.

This publication has 10 references indexed in Scilit: