Automatic knowledge learning in an expert system for space management and planning

Abstract

A procedure of self-learning and knowledge reducing is described. It concerns the step of knowledge acquisition for an expert system for space management and planning. Section 1 gives a general presentation of the expert system. It underlines the main characteristics of the tool. Section 2 describes the problem of knowledge acquisition and internal representation: it underlines that experts often give different expressions for the same ‘reality’, so that it becomes necessary to define ‘knowledge conflict resolution’. Section 3 explains the logical bases of self-learning and knowledge reducing: we propose: an index of consistency which ‘evaluates’ if two proposed rules are ‘near’ and can be reduced into a unique one, or not; an index of leadership which defines, while knowledge reducing, what is the ‘predominant’ rule; a ‘compromise law’ which makes the system able to operate this reducing. Section 4 briefly presents the developed procedure (Appendix 1 presents the corresponding algorithm).