Measuring the structural complexity of OPS5 rule-based programs

Abstract

Complexity metrics, such as McCabe's cyclomatic number of a program control graph and Halstead's number of operator/operand occurrences have been used extensively to measure the structural complexity of procedural programs. However, few suitable complexity metrics have been developed for rule-based programs written in OPS5 and OPS5-like rule-based languages. With the increasingly common use of rule-based languages in knowledge-based systems, this paper describes new complexity metrics to more accurately measure the complexity of OPS5 rule-based programs. The practicality of these metrics is empirically demonstrated by applying them to measure the complexity of a suite of benchmark: OPS5 expert systems.