Effectiveness evaluation of expert classification methods

A special class of multicriteria classification problems is under consideration in the paper: expert classification problem. It consists in the construction of an expert knowledge base in a classification task. The criterion of expert classification methods efficiency is the minimum number of questions needed for the construction of a complete classification. To arrive at an efficiency estimation system for different methods, a comparison with mathematical algorithms of monotone functions decoding is made. The two best algorithms optimal by Shannon are presented. A procedure of simulating different monotone functions is proposed. A new efficient method of expert classification, CYCLE, is proposed. The results of the simulation demonstrate that the method CYCLE has good evaluations of efficiency for arbitrary monotone functions.