Algorithm Research And Application Of Bayesian Network

Bayesian network is a kind of probabilistic network, which is a graphical networkbased on Probabilistic inference. The basis of this network is Bayesian formula. Bayesiannetwork is a mathematical model based on Probabilistic reasoning that is to obtainadditional probabilistic information of others through the information of some variables.In order to solve the problem of uncertainty and incompleteness, Bayesian network isproposed, which has great advantage in troubleshooting those fault caused by theuncertainty and relevance in a complex equipment. It has a wide range of applications inmedical diagnosis, statistical decision, expert system, prediction and other fields. In aword, the Bayesian network theory is an important research field in the ArtificialIntelligence, which has a high research value.This paper firstly introduces the related theory of Bayesian network, including thebasic knowledge, the concept of Bayesian network and the its practical application andinfluence in various fields. And the classic learning algorithms for Bayesian network arediscussed.Secondly, the K2structure learning algorithm is analyzed, and a preliminaryimprovement scheme is proposed for prior node-order and scoring criteria. Then under thecondition of small sets of data Bayesian network parameters learning are discussed, andthe preliminary solution proposed was validated on the Classic Bayesian networks.Thirdly, Classic junction tree inference algorithm is introduced in detail, thoroughdeep analysis of it, and it is pointed out that time consumption problems in the process ofits transforming structure and message propagation. Aiming at these deficiencies, asimplifying-structure algorithm and a effective junction tree algorithm are proposed in thispaper, and through simulation experiment in Classic Alarm network, it was demonstratedthat the validity and feasibility of the two algorithms presented.Finally, in order to verify the effect of the inference algorithms improved, the mainparameters involved in the process of cement calcination are analyzed briefly and a faultdiagnosis scheme based on Bayesian network is presented. Based on the actual datacollected and integrating expert knowledge, a fault diagnosis model of the cement burning system is built using classic K2method and MLE method. Then Classic junction treealgorithm and the improved junction tree algorithm were utilized in some fault diagnosis,and diagnosis effect was analyzed.