Datasets on the number of hospitalized, discharged, and death cases caused by Covid-19, and their prediction in Bushehr province, Iran

This dataset indicates the current and future trends of three indicators related to Covid-19 in Bushehr province, Iran. Figure 1 shows the daily trends of the number of hospitalized, discharged, and death cases caused by Covid-19 in Bushehr province between May 13, 2020, and April 1, 2021. As shown in this figure, all three indicators have seasonal and irregular changes. First, the number of discharged and death cases have been predicted according to their relationship with the number of hospitalized cases using the MLP neural network model. The structure of the proposed MLP model is shown in Figure 2. This model has been created using the neural network toolbox in MATLAB 15b software package, which also can generate model scripts. The script of the trained model is shown in supplementary .m File 1. Besides, Figure 3, Figure 4, and Figure 5 indicate the plots of performance, training state, and regression corresponding to the trained MLP neural network respectively. After predicting the test dataset using the MLP model through Simulink/MATLAB, the MC model has been used to improve the performance of the MLP model in prediction. The calculations related to the MC part have been performed in Excel 2010 software. The residual errors between the actual and predicted data by the MLP model for each indicator and the corresponding states are shown in Table 1. Figure 6 shows the forecast results of the next 40 days (April 2, 2021, to May 11, 2021) using the MLP-MC model for the two indicators of the number of discharged and death cases. Also, these values are shown in detail in Table 2.

本数据集展示了伊朗布什尔省与新型冠状病毒肺炎（COVID-19）相关的三项指标的当前及未来趋势。图1呈现了2020年5月13日至2021年4月1日期间，该省新冠住院、出院及死亡病例数的逐日变化态势。如图所示，三项指标均兼具季节性与非规律性波动特征。
首先，研究人员依托出院病例数、死亡病例数与住院病例数的内在关联，采用多层感知器（MLP）神经网络模型对后两项指标进行预测。所提出的MLP模型结构如图2所示，该模型依托MATLAB 15b软件的神经网络工具箱构建，同时支持生成模型脚本，训练完成的模型脚本详见补充.m文件1。
此外，图3、图4、图5分别展示了该训练完成的MLP神经网络的性能曲线、训练状态曲线与回归曲线。在通过Simulink/MATLAB利用MLP模型完成测试集预测后，研究人员进一步采用蒙特卡洛（MC）模型优化MLP模型的预测性能，MC模块的相关计算在Excel 2010软件中完成。
各指标的实际值与MLP模型预测值之间的残差误差及其对应状态详见表1。图6展示了采用MLP-MC模型对出院病例数、死亡病例数这两项指标进行的未来40天（2021年4月2日至2021年5月11日）的预测结果，具体数值详见表2。