Bayesian Network Structure Learning Algorithm Research And Application

Bayesian network (BN) is a model based on probability theory and grap theory, it can be applied to reasoning uncertain knowledge and data analysis in complex domain. In recent years, Bayesian network learning has been successfully applied to many fields, such as Artificial Intelligence、Biological Information Finance、Medical Diagnosis and Prediction, and many classical and efficient algorithm was proposed, how to efficient optimal best network structure from a given dataset has been the concern of many scholars. After analyzing relevant theories of Bayesian networks, this paper focuses on the structure learning of Bayesian networks. The main works can be summarized as follows:(1) Introduce to the theory of Bayesian networks. Including the background, the current research state and development trend of Bayesian networks. The basic principle of Bayesian networks, including Bayesian network structure learning, parameter learning and inference.(2) Introduce to the basic theory of Maximal Information Coefficient and Artificial Bee Colony. An algorithm of Bayesian network structure learning (MIC_ABC) was proposed based on them. The algorithem dig out the dependence between variables at firstly, and initial Bayesian network structure by this dependence and condition independence test, so that to effectively reduce the search space of model, then optimization and iteration the better structure from the initialization network structure based on Artificial Bee Colony algorithm and Bayesian score criterion.(3) Introduce to the Junction tree algorithm of exact inference algorithm in Bayesian networks, used the maximum likelihood estimation to parameter learning and junction tree to inference for Bayesian network based on up structure learning.(4) The aging of population in our country has entered a rapid development period. In critically ill patients, the elderly account for a large proportion and take up many resource, but the treatment effect hava not reach the average should be used for the resource. Therefore, study the possibility of benefit of treatment for the elderly in the intensive care unit (ICU) is very important, this not only to treatment diagnosis but also the necessary to enter the ICU for patient and reasonable allocation. In this paper, an evaluation model of the diagnosis was constructed for elderly critical patients based on Bayesian networks and the clinical dataset provided by Guangzhou MICU.