AN INTELLIGENT MODEL-BASED SYSTEM FOR MULTI-EXPERTS COOPERATION

Abstract

Artificial Intelligence applied to the field of business primarily refers to the handling of problem-solving or decision-making situations. Most of the real world problems involved many knowledge domains, other than simply applying rules and formulas, so multi-experts’ cooperation (cooperative agents) is necessary in problem solving. This research aims to build an intelligent system which exhibits some intelligent behaviours like human experts, such as cooperative problem solving in a complex decision environment. To this end, we first build several expert models from human experts’ knowledge. Each expert model is a specialist at solving certain aspects of the overall problem. We then develop a model-based approach to do multi-experts cooperation, i.e. how multi-expert models cooperate to solve a complex problem. This thesis presents the first step towards a framework of cooperation among multi-experts for problem-solving and decision making. The framework is based on a field study of a financial organisation performing a cooperation intensive task: loan evaluation. Accordingly, the first part of this thesis performs detailed analysis of the loan investigation process by varying market operating efficiency, and builds several quantitative expert models for analyzing commercial loan application. Thes second part of this thesis deals with the proposed framework - a model-based system. It focuses on the action selection mechanism for multi-expert models, the method by which these expert models carry out the cooperation when they work together on a common goal, as well as the organization of these expert models. A simulation of the framework using object oriented programming (Allegro Common Lisp) and blackboard technology for an example application in credit evaluation is performed.