基于储备池计算的传染病预测研究与优化

Research and optimization of infectious disease prediction based on reservoir computing

以新冠疫情传播为例,利用储备池计算网络建立传染病预测模型。针对算法本身与测试数据的特点,采用一种独特的同步时间差预测方式:使用多数地区的疫情发展数据来预测少数地区的发展趋势。同时使用标准SIR传染病模型生成的高维数据对多种层次的储备池计算网络在预测时间线上的误差进行分析,并以减小误差为目的提出一种时变权重储备池计算网络。结果显示时变权重算法模型相比较于单一类型的储备池可以有效地提高预测精度,在测试数据集上可以将误差降低45.6%。

Taking the spread of the COVID-19 as an example,the prediction model of infectious diseases is established by using the classical reservoir computing.Based on the characteristics of the algorithm and test data,a unique synchronous time difference prediction method is adopted:using epidemic development data from most regions to predict the development trend of a few regions.Simultaneously,high-dimensional data generated by the standard SIR infectious disease model is used to analyze the errors of multiple levels of reservoir computing networks on the prediction timeline,and the time-varying weight reservoir computing network is proposed to reduce errors.The results show that the time-varying weight algorithm model can effectively improve prediction accuracy compared to a single type of reservoir computing,and can reduce the error by 45.6%on the test dataset.