论文部分内容阅读
针对限制动态 Bayesian网络方法应用的 Markov假设和转移概率时不变假设 ,研究了如何利用部分观测信息建立宏观经济系统的 Markov模型以及如何建立转移概率具有时变特性的宏观经济系统模型。对不满足 Markov假设的演化过程 ,通过在模型中添加隐藏变量建立 Markov模型 ,并对 EM- EA算法进行扩展 ,使之用于带隐藏变量的动态 Bayesian网络的学习。对不满足时不变性的转移概率 ,应用多项式拟合方法直接从数据构造时变转移概率模型。理论分析表明了论文方法的正确性和可行性
Aiming at the Markov assumptions and the invariance invariance when the dynamic Bayesian network is applied, this paper studies how to use some observational information to build the Markov model of macroeconomic system and how to set up the macroeconomic system model with time-varying transition probability. For the evolutionary process that does not satisfy the Markov hypothesis, the Markov model is built by adding hidden variables in the model and the EM-EA algorithm is extended for the learning of dynamic Bayesian networks with hidden variables. Applying the polynomial fitting method to transfer the probability model directly from the time-varying data structure to the transition probability of invariance not satisfied. Theoretical analysis shows that the essay method is correct and feasible