动态贝叶斯网络在复杂系统中建模方法的研究

提出了用动态贝叶斯网络(DBN)对复杂系统进行建模的有效方法。基本思路是将扩展后的隐变量引入了DBN的演化过程中来建立马尔可夫模型,并给出了引入扩展后的隐变量的DBN结构学习算法框架。进而,利用贝叶斯概率统计方法对后续时间片的充分统计因子进行了估计,并通过当前已存在的充分统计因子和估计的充分统计因子对基于时间变化的转移概率进行了学习。原理性分析和仿真实验结果也验证了该方法的有效性。

Research on Modeling with Dynamic Bayesian Network in Complex Systems

A new method was proposedfor modeling complex systems with DBNs.Firstly,the extended hidden variables were introduced into the evolutional process to build Markov models and a structure learning algorithm of DBNs was provided in the presence of the extended hidden variables.Secondly,the sufficient statistics of posterior time slices were estimated using Bayesian probability statistical method,and then the time-variant transition probabilities were learned with both current sufficient statistics and estimated sufficient statistics.Finally,the theoretical analysis and simulation results show that the proposed method is valid.