Design And Implementation Of Rule-Based Bayesian Network Developer

As a method of describing indeterminal things and indeterminal computation, Bayesian network is applied extensively in many fields. Bayesian network is a kind of model that is an Organic combination of Bayesian method and DAG Topological structure and describes the data field and their relationship of dependence. It is a graph _based model come from probability relationship. But how to get Bayesian network Topological structure is a very crucial problem. In order to find a good Bayesian network Topological structure, to learn structure is very imported.The Thesis analyses many kinds of Algorithm about Bayesian network structure learning, and then Setting-up a new Algorithm about structure learning Foundation on hydro-electrical simulation system. The Algorithm adopts Statistics Policy to get effective rules from rule warehouse, discard Weak rules and hold strong rules. The algorithm can get better Bayesian network Topological structure form these rules. If there is loop in the Topological structure, it also can make it by different method such as Statistics, priority and so on. Finally get a better Bayesian network Topological structure that has a good combination of data and Priori knowledge. CPT learning is another important part. In order to get a full Bayesian network developer, the thesis has a meaningful change in CPT learning through tight reasoning.Finally,The Algorithm is applied in structure learning and CPT learning of guzhangzhenduan in shuidianfangzhen system. A better Bayesian network based_ fault diagnosis model is founded. The model plays fully the advantage of Bayesian network in solving the no determinacy problem. The model proves right and high-efficient.