Abstract:
The future state not only relies on the prior state,but also is influenced by multi-random factors,so prediction models should represent these dependencies between any two stochastic variables.Dynamic Bayesian network is a powerful tool to solve the problem.A state prediction model based on dynamic Bayesian network is developed.According to the mutual information,the degree of sustainment among nodes is propounded.And an evidence propagation algorithm is proposed to amend the predicted results.Finally,it is introduced an integrative process for forecasting.