Research And Application Of Probability Graph Model

The probability graph model uses a simple and clear graphic structure to express the relationship between variables,which combines the advantages of graph theory and probability theory.It provides a powerful tool for the solution of uncertain reasoning system and complex system problems,and can carry out quantitative calculation and uncertain reasoning of the system.In recent years,with the in-depth study of uncertain reasoning system and the rapid development of machine learning,the integration of probability graph model with computer vision,statistical inference and data mining is more in-depth,and its application prospect is increasingly inestimable.The main research content of this paper is the learning and application of directed graph model.Firstly,the basic theory of probability graph model is introduced,and then the learning algorithms of hidden Markov model and Bayesian network are discussed emphatically.The corresponding algorithms are given for the three basic problems of applying hidden Markov model in practice.For Bayesian networks,parameter learning method and structure learning method are given according to whether the data set is complete or not.As we all know,hidden Markov model is widely used in weather prediction.Therefore,this paper hopes to extend its application in reality from short-term weather prediction to long-term climate prediction.Under the background of global warming,the atmospheric carbon dioxide concentration,global ocean surface temperature,global land surface temperature and the solar radiation flux are included in the prediction index system through variable rationality analysis.Then hidden Markov model is used to predict global temperature,and Bayesian network is used to re-estimate the prediction results in the next 20 years.It not only improves the accuracy and credibility of the prediction trend,but also provides a new method for climate prediction.