自2019年底以来，新发传染病（COVID-19）以高传播性和强变异性席卷全球，各国迅速根据疫情传播形势采取了相应的防控措施。作为全球面临的一场重大卫生危机，其冲击力影响了经济、政治和卫生等多个领域，尽管全球现处于“后疫情”时代，仍需要对COVID-19的流行和传播趋势进行实时监测，以及评价政府出台的防控措施对疫情传播的影响。因此根据疫情数据构建高精度的传染病预测模型，对政府部门制定科学合理的疫情防控政策具有重要意义。本文研究如下

首先为了实时监测陕西省COVID-19传播现状和政策防控措施的有效性，结合每日新增感染人数，通过贝叶斯框架刻画了2021年12月1日至2023年1月1日期间陕西省两轮疫情的实时再生数的流行趋势，并结合每日综合政策指数建立log-log时变系数回归模型得到政策指数影响效应。结果表明陕西省政策指数影响效应随着时间的推移整体呈上升趋势，即早期由部门主导的一级响应策略和措施效果显著，但随着防疫政策的不断调整，后期政策措施对降低放大效应逐渐减小。

其次针对金豺优化算法易早熟和局部开采能力弱等问题，采用随机收缩指数函数修正金豺对的能量递减机制，引入正弦波动速率调节变异因子，提出一种基于自适应变异和动态能量策略的IGJO算法。选取9个基准函数进行数值实验，结果表明相较于ALO、GWO、WOA和GJO算法，IGJO算法具有较强的搜索能力，能够快速地跳出局部最优。

最后建立了融合ANN、SVR、RNN和LSTM四种机器学习模型的非线性GRNN变权组合预测模型，通过IGJO优化算法搜索获得变权组合模型的最优权重系数，并分别对伊朗和秘鲁的每日新增确诊人数进行实验评估模型性能。结果表明，该模型对伊朗和秘鲁的平均R²分别为0.9990和0.9989，相较于四种单一预测模型的平均RMSE分别降低了85.24%、70.81%、80.99%和78.57%，其预测性能优于等权赋值和熵权赋值组合预测模型，可以为疫情实时监测和相关防控措施的制定提供一定的参考。

Since the end of 2019, emerging infectious diseases (COVID-19) have swept the world with high transmission and strong variability, and countries have quickly made corresponding prevention and control measures according to the epidemic situation. As a major health crisis facing the world, its impact has affected many fields such as economy, politics and health. Although the world is in a "Post-pandemic" era, it is still necessary to monitor the epidemic and transmission trend of the COVID-19 in time, and evaluate the impact of prevention and control measures introduced by the government on epidemic prevention and control. Therefore, building a high-precision infectious disease prediction model based on epidemic data is of great significance for government departments to formulate scientific and reasonable epidemic prevention and control policies. This paper studies the following：

Firstly, in order to monitor the current situation of COVID-19 transmission and the effectiveness of policy prevention and control measures in Shaanxi Province, combined with the number of daily new infections, the Bayesian framework was used to analyze the epidemic trend of real-time reproduction number of two rounds of epidemics in Shaanxi Province from December 1, 2021 to January 1, 2023. Using the real-time reproduction number and the daily comprehensive policy index, the log-log time-varying coefficient regression model is established to obtain the effect of the policy index. The results show that the effect of the policy index of Shaanxi Province has an overall upward trend over time, that is, the primary response strategies and measures led by the department in the early stage have a significant effect, with the adjustment of epidemic prevention policies, the amplification effect of the later policy measures on the reduction decreases.

Secondly, aiming at the problems of premature convergence and weak local exploitation ability of the golden jackal optimization algorithm, the random shrinkage exponential function is used to modify the energy decreasing mechanism of the golden jackal pair, and the sinusoidal fluctuation rate is introduced to adjust the mutation factor. An IGJO algorithm based on adaptive mutation and dynamic energy strategy is proposed. Nine benchmark functions are selected for numerical experiments. The results show that compared with ALO, GWO, WOA and GIO algorithms, the improved algorithm has stronger search ability and can jump out of the local optimal solution quickly.

Finally, a nonlinear GRNN combined prediction model integrating ANN, SVR, RNN and LSTM four machine learning models is established. The optimal weight coefficient of the combined model is obtained by IGJO optimization algorithm search, and the daily number of new confirmed cases in Iran and Peru is experimentally evaluated. The results show that the average R² of the model for Iran and Peru is 0.9990 and 0.9989, respectively. Compared with the average RMSE of the four single prediction models, it is reduced by85.24%, 70.81%, 80.99% and 78.57%, respectively. And its prediction performance is better than the combined prediction model of equal weight assignment and entropy weight assignment, which can provide some reference for the real-time monitoring of epidemic situation and the formulation of relevant prevention and control measures.