马尔可夫模型的流感优势毒株预测分析

摘要：目的　预测流行季节流感的优势毒株，为疫情研判和防控提供依据。方法　根据２００９－２０１３年浙江
省流感监测样本的检测结果，按上下半年对各年流感优势毒株亚型排列，将流感各亚型进行状态划分，并
按时间顺序建立状态变化列表，计算各亚型间的一步状态概率转移矩阵，利用状态转移概率矩阵预测下一
流行季节各流感亚型成为优势毒株概率。结果　浙江省流感监测样本阳性检出率呈上下半年双峰分布，且
每一个阳性检出峰中流感优势毒株类型单一明显，根据一步状态概率转移矩阵，２０１３年下半年流感优势毒
株为Ａ （Ｈ３Ｎ２）、甲型Ｈ１Ｎ１、Ｂ型的概率分别为０、０、１００％。结论　应结合预测结果，流感样病例、分
子实验室、药敏监测等综合判断，同时应根据实际监测结果调整状态概率转移矩阵。
关键词：马尔可夫模型；流感优势毒株；预测
中图分类号：Ｒ１８３．１　　文献标识码：Ａ　　文章编号：１００９ ６６３９ （２０１４）０３ ０１７０ ０３

Prediction for the predominant strain of influenza virus based on the markov model

Objective To predict the predominant strain of influenza virus during epidemic season,and to provide evidence for disease control. Methods The subtypes of annual predominant strains of influenza virus were sorted out according to the laboratory detection results of samples collected from 2009to 2013in Zhejiang,which were then classified in a timely order.The one step state probability shift matrix was calculated for each subtype and was applied to predict the predominant strains for the upcoming epidemic season of influenza. Results The positive rate of samples collected in Zhejiang showed a bimodal distribution in each year, and usually there was only one predominant strain of influenza virus in each peak.The probabilities of the predominant strains in the second half year of 2013were predicted as 0%of H3N2,0%of AH1N1and 100%of B,respectively according to the markov model. Conclusions Markov model can be used to predict the predominant strain of influenza virus during epidemic season.However,a comprehensive judgment should be made based on molecular laboratory result,drug resistant test as well as the adjustment of state probability shift matrix.