A MATLAB Graphical Tool to Support Knowledge Engineering

Abstract

Bayesian networks give an efficient probabilistic model rep-resentation scheme to encode both expert knowledge and information extracted from database. Although the probabilistic model representa-tion and the storage space problem can be solved using Bayesian net-works, the model evaluation (inference) operation become complicated. The practical solution of this problem is an optimum between the ac-curacy of model and the complexity of the model evaluation task. This trade-off is more easily found if the knowledge engineers could receive an immediate feedback during the developing phase of the network about the computational resources required to evaluate the probabilistic model. This paper presents a Bayesian network modeling software tool focusing on its graphical user interfaces and data structures.