Research On Bayesian Network Structure Learning

Bias network(Bsyesian network)is a description of graphical model dependencies between random variables,intelligent optimization is widely applied to uncertain problems.Bias network learning is divided into learning and structure learning and parameters,this paper mainly studies the structure of network learning.Bias Bias is on the network of the relationship between variables,the inherent attribute the results of the study to ensure the feasibility of the structure,also provides a basic idea for structure learning.Network structure learning Bias goal is to find a best fit variable relationship between network structure relations between the data table.The attribute dependency of rough set is a good representation of the dependencies among the two variables.The method is proved to be effective and accurate in comparison with some existing methods for building Bayesian networks.It is an effective method to decompose the complex network into a number of small networks for more complex networks with multiple attributes.In this paper,we propose a method to decompose Bayesian networks based on the concept of joint tree and Bayesian network.This method can effectively deal with complex Bayesian networks,and can solve the problem of information retention in the decomposition process.All the small networks obtained by this method are consistent with the original complex network,and there is no new dependency in the decomposition process.Therefore,the decomposition method is effective.First of all,this paper introduces the background and significance of the Bias network,and briefly introduces the research status of the network,as well as the basic knowledge of the Bias network.Then,several methods of Constructing Bayesian network structure are introduced in detail.In this paper,we study the properties of attribute dependency and propose a new method for constructing Bayesian networks.Finally,the significance of the decomposition of Bayesian networks is studied,and a method to decompose Bayesian networks is proposed.