马尔可夫模型在疑似预防接种异常反应报告趋势预测中的应用及R语言实现

Application of Markov modeling in the prediction of adverse events following immunization using the R language

目的应用马尔科夫模型对甘肃省2016年11月和12月疑似预防接种异常反应(Adverse events following immunization,AEFI)报告数进行预测。方法选取2015年1月-2016年10月甘肃省分月AEFI报告数,通过10折交叉验证将其划分为6个状态,通过时间与状态的转移概率矩阵预测2016年11月和12月AEFI报告数。结果通过转移概率矩阵得到甘肃省2016年11月和12月转移概率分别为(0.33,0.33,0.33,0.00,0.00)和(0.00,0.33,0.19,0.19,0.19,0.08),11月和12月预测数分别为663例和717例,预测误差分别为14.67%和-38.68%。结论马尔科夫模型进行AEFI报告趋势预测是可行的,需要收集较长的时间序列数据以提高预测精度。

Objective To predict reported cases of adverse events following immunization（AEFIs） in November and December of 2016 in Gansu province by using Markov modeling. Methods We selected AEFIs reported from January 2015 to October 2016 in Gansu,and divided them into six conditions through a 10-fold cross-validation. AEFI cases in November and December of 2016 were predicted by a time and condition transition probability matrix. Results Through the step transition probability matrix transformation,the respective transition probability matrixes in November and December of 2016 were（0. 33,0. 33,0. 33,0. 00,0. 00） and（0. 00,0. 33,0. 19,0. 19,0. 19,0. 08）. The predicted number of cases in November and December were 663 and 717,with a prediction error of 14. 67% and-38. 68%,respectively. Conclusions Markov modeling is feasible for prediction of AEFI reporting. A longer date time series should be collected to improve predictive accuracy.