The Improve Of Reduction And Combination Of Hybird Uncertainly Bayesian Net Reasoning Model And Prediction Of Colleges’ Input-output

At the present stage,the shortage ofeducation resources is still one of the important challenges facing the development of higher education in China.For each college and university, how to make limited resources play a biggest role to get more output is an important problem.The research about universities’ input and output helps reflect the existing problems of resources’ utilization and guide them to improve the management of resources and the quality of education.The uncertainty of input and output data exists widely,bayesian network is a powerful tool to solve the problem of uncertainty knowledge reasoning.But to build the bayesian network for application about input and output prediction research is facing the following difficulties:firtsly, it involved a large number of variables, it is well known that the bayesian network method for multivariable network learning is a np-hard problem; secondly, the variables it involved include both discrete random variables and continuous random variables, in tradition, continuous random variables are dealt with variables discretization to learn networkwhich faces problems of edge sharpening.To solve these difficulties, this paper puts forward subtraction andcombination of hybrid uncertainly bayesian multi-subnet algorithm used tolearn network which used for colleges’ input and output prediction research The main research content of this paper is as follows:Firstly,for the problem of edge sharpening, continuous random variableswill be treated with fuzzy processing.To learn the hybrid bayesian network whichcontains both real random variables and variablesmixed fuzzy and random property,this paper gives its structure learning, parameter learning and network reasoning method;Secondly,use the reduction and combination algorithm of bayesian network learning to sovle the difficult of building multivariate network.This paperimproves the existing reduction and combination algorithm and provesits advantage overscale of networks, model training time and reasoning accuracy by experimental analysis;Lastly,using the improvedreduction and combination ofbayesian multi-subnetlearning and hybrid uncertainlybayesian multi-subnetlearning algorithmsto construct college’ network and hybrid networkfor input and output forecast reasoning.Compare the two networks’ reasoning accuracy, then find that the former is superior to the latter.