Application Of Improved Bayesian Network Algorithm In Sub Health

As a kind of probability network, bayesian network own obvious advantage in reasoning of uncertain problem. It could analyze and predict relevant problem based on probability reasoning, which could effectively resolve uncertain and incomplete problem. Recently a number of researchers apply bayesian network to medical field and achieve remarkable results, this paper apply improved bayesian network algorithm to sub-health analysising, construct the sub-health status model, and analyze the probability relationship among different factors, predict the occurence of sub-health.First of all, this paper introduce the relevant theory of bayesian network, including the fundamental knowledges of bayesian network, the application in some fields and th eir effect. At the same time introduce detailed the all kinds of bayesian network learning algorithm.Next analysis the disadvantages of traditional K2 algorithm, improve the problem th at node is highly dependent on priror order and the problem of score function, propose a K2 algorithm based on mutual information, effectively solve the problem that node is h ighly dependent on priror order, then apply this new algorithm to classical ASIA networ k, the result show our improvement is obvious.In addition, study furthurly the classical joint tree algorithm, analysis the drawbacks that time consuption is long in information transformation and transmit and of classical joint tree algorithm, this paper propose a high effective joint algorithm and a kind of algorithm with easy structure, and use new algorithm to experiment with Alarm network,which testify the feasible and high efficiency of new algorithm.This paper at last study a common problem------sub-health, firstly introduce the problem, Next according to the collected data and combine relevant expert knowledge, apply improved K2 and MLE algorithm to modelling the diagnose model of prediction and analysis of sub-health, at the same time, make use of classical joint tree algorithm and improved joint tree algorithm to predict and analysis, and describe the experimental result.