基于不确定性描述的云化Markov链状态预测方法

针对Markov链在预测概率发生跳变时无法有效地衡量样本归属程度的问题，提出一种云化Markov链的状态预测方法，通过云模型描述和处理样本的不确定性。该方法将划分的状态区间视作一种概念，利用云模型对其进行云化表示，据此计算样本对各概念的确定度，得到概念之间的概率转移矩阵，从而实现带有随机特性的状态预测。概念转移概率作为关键随机变量，对其进行了核密度估计。最后以多次随机实验的概率和提取代表性转移概率分别给出了仿真实验结果，表明该不确定性描述的预测方法在解决Markov链预测概率跳变现象的同时，可通过确定度的分配有效地表述样本的归属程度，具有较好的实用性。

Cloud transforming method of Markov chain state prediction based on uncertainty description

To deal with ownership degree of samples in Markov chain effectively facing with a skip of the predicted probability, a cloud transforming method of Markov chain state prediction is proposed. Samples’ uncertainty is described and processed by using the cloud model. Regarded as a kind of concept, the partitioned state intervals are expressed based on the cloud model, and further the certainty of each objective to all concepts is computed. Then to realize stochastic state prediction, the concept transfer matrix is calculated. The kernel density estimation of concept transfer probability is obtained considering its significance. Finally simulation results are given in form of probability of repeated tests and extracted representative transfer probability, and it shows that the uncertain method can both avoid a skip of the Markov chain predicted probability and measure ownership degree of samples effectively, and is more practical as well.