常见新型冠状病毒肺炎疫情预测方法及其评价

Evaluation of Predictive Models for Novel Coronavirus Pneumonia

自新型冠状病毒肺炎疫情发生以来,一些学者利用疫情公开数据建立预测模型。所用预测方式包括曲线拟合、传染病动力学模型及人工智能算法三大类。传统的曲线拟合预测方式无法考虑传染病特征,预测结果并不可靠。传染病动力学模型是本次疫情预测应用最多的一类,能够考虑传染病的传播速度、传播模式及各种防控措施等因素,但由于考虑的参数不可能全面,且参数可能在疫情不同阶段发生动态变化,因此预测效果往往不佳,但对早期预警、防控决策支持及防控效果评价具有重要应用价值。人工智能方法可以综合考虑不同防控措施以及多种因素的影响,如果考虑得当,预测效果将会有所提高。在综合利用动力学模型优势的基础上,尽可能多地考虑不同影响因素,利用人工智能构建仿真模型,将是一个新的发展趋势。

Since the outbreak of novel coronavirus pneumonia,many scholars used the open data to build prediction models.The most common methods are curve fitting,infectious disease dynamics model and artificial intelligence algorithm.The traditional curve fitting method cannot take the characteristics of infectious diseases into consideration,and the prediction results are not reliable.The dynamic model of infectious disease is the most widely used method in this outbreak,which can take into account the factors such as the transmission rate,transmission mode and various prevention and control measures.Although the consideration and dynamic changes of parameters at different stages of the outbreak would lead to the poor predictions,it still has meaningful applications for early warning,prevention and control decision making support and effect evaluation.The artificial intelligence method can comprehensively consider the influence of different prevention and control strategies and various factors,if properly conducted,forecasting accuracy would be improved.It will be a promising direction to build dynamic simulation system using big data&AI techniques together with the advantages of dynamic model with sufficient fixed and random parameters.