Although more than a year has passed since the coronavirus outbreak globally, the Covid-19 pandemic conditions still exist in many countries, including Iran. Predicting the number of future patients and deaths can help governments and policymakers make better decisions to enforce disease control restrictions. In this study, we aim to use a combined multilayer perceptron (MLP) neural network and Markov chain (MC) model to predict two indicators of the number of discharged and death cases according to their relationship with the number of hospitalized cases in Bushehr province, Iran. This hybrid model is called MLP-MC.

In this study, 136 data (days) are collected from May 13, 2020, to April 1, 2021, divided into two parts: training and test. The training data are used to train the MLP network, and the trained MLP network is used to predict the test data and the next 40 days. Then the residual errors of actual and predicted values are calculated. In the next step, the MC model is used to classify the errors and predict the values of the indicators according to the probabilities related to the error states and improve the performance of the MLP model in forecasting. Finally, the prediction accuracy of MLP and MLP-MC models are compared using three evaluation metrics: MAD, MSE and RMSE. This comparison showed that the MLP-MC model has slightly higher prediction accuracy than the MLP model.

尽管冠状病毒在全球爆发已过去一年多，但包括伊朗在内的许多国家仍然存在Covid-19大流行的情况。预测未来的患者和死亡人数可以帮助政府和政策制定者做出更好的决策来执行疾病控制限制。在本研究中，我们旨在使用组合的多层感知器（MLP）神经网络和马尔可夫链（MC）模型，根据布什尔省的出院人数和死亡人数两个指标与住院病例数的关系来预测它们，伊朗。这种混合模型称为 MLP-MC。

本研究收集了2020年5月13日至2021年4月1日的136条数据（天），分为训练和测试两部分。训练数据用于训练 MLP 网络，训练后的 MLP 网络用于预测测试数据和未来 40 天。然后计算实际值和预测值的残差。下一步，利用MC模型对误差进行分类，并根据与误差状态相关的概率来预测指标的值，提高MLP模型的预测性能。最后，使用 MAD、MSE 和 RMSE 三个评估指标比较 MLP 和 MLP-MC 模型的预测精度。这一比较表明，MLP-MC模型的预测精度比MLP模型略高。