On a Method of Rule Generalization for Data-Mining Problems

This paper describes a method for generalizing rules for data mining problems using the ideas of similarity and inductive inference. The game tic-tac-toe is used as a model. To create a training sample, a complete tic-tac-toe game tree is constructed. An algorithm for generalizing rules is described and some results of computer modeling are presented. It is shown that the application of this method allows one to obtain generalized rules for winning strategies. The rules have good explainability and can be easily interpreted by humans.