一步预测的SVDDBN缺失数据插补算法

变结构离散动态贝叶斯网络（SVDDBN）处理不确定性问题更具有一般性，为了克服SVDDBN缺失数据会导致推理结果精度变差的问题，提出了一步预测的SVDDBN缺失数据插补算法。根据信息可以沿着网络的时间轴方向向下一个时间片传播的规律，利用“混合”信息在线进行信度更新，可得到滤波值，再通过进一步预测得到下一个时间片缺失数据节点的后验概率作为插补值。仿真结果表明：提出的算法能有效插补缺失数据，提高SVDDBN推理的精确度及可靠性。

One-step Prediction SVDDBN Missing Data Interpolation Algorithm

It is more general to deal with the uncertainty of the Structure-Variable Discrete Dynamic Bayesian Network(SVDBN). In order to overcome the problem that SVDDBN missing data leads to poor accuracy of reasoning results, a one-step prediction SVDDBN missing data interpolation algorithm is proposed. According to the law that the information can propagate along the time axis of the network to the next time slice, the filter value can be obtained by using the "mixed" information to update the reliability online, and then the posterior probability of the missing data node of the next time slice can be obtained as the interpolation value by further prediction. The simulation results show that the proposed algorithm can effectively interpolate missing data and improve the accuracy and reliability of SVDDBN inference.