EVALUATION OF THE NUMBER OF CONSISTENT MULTIATTRIBUTE CLASSIFICATION RULES

Classification rules that are explicitly expressed by even experienced decision-makers are often in conflict with respect to each other The disagreements among conflicting rules have typically to be resolved be fore the rules can be used as tutorial exemplars, in written manuals, or in computerized expert systems. In the latter case, an evaluation o f the number of rules that are in agreement with respect to each other is often needed, for example, to predict memory requirements and resp onse time. Conflict resolution is viewed here as a stochastic process with a single parameter. The statistical distribution of the number of consistent classification rules is derived, as well as a closed-form expression of their expected number. (C) 1997 Elsevier Science Ltd.