Sabir Zulqurnain, Hashem Atef F, Shams Zill E, Abdelkawy M A
Department of Computer Science and Mathematics, Lebanese American University, Beirut, Lebanon.
Department of Mathematics and Statistics, College of Science, Imam Mohammad Ibn Saud Islamic University (IMSIU), Riyadh, Saudi Arabia.
Comput Methods Biomech Biomed Engin. 2024 May 6:1-17. doi: 10.1080/10255842.2024.2349647.
In this study, a stochastic computing structure is provided for the numerical solutions of the SIRC epidemic delay differential model, i.e. SIRC-EDDM using the dynamics of the COVID-19. The design of the scale conjugate gradient (CG) neural networks (SCGNNs) is presented for the numerical treatment of SIRC-EDDM. The mathematical model is divided into susceptible recovered infected and cross-immune while the numerical performances have been provided into three different cases. The exactitude of the SCGNNs is perceived through the comparison of the accomplished and reference outcomes (Runge-Kutta scheme) and the negligible absolute error (AE) that are performed around 10 to 10 for each case of the SIRC-EDDM. The obtained results have been presented to reduce the mean square error (MSE) using the performances of train, validation, and test data. The neuron analysis is also performed that shows the AE by taking 14 neurons provide more accurateness as compared to 4 numbers of neurons. To check the proficiency of SCGNNs, the comprehensive studies are accessible using the error histograms (EHs) investigations, state transitions (STs) values, MSE performances, regression measures, and correlation.
在本研究中，利用新型冠状病毒肺炎（COVID-19）的动力学特性，为SIRC疫情延迟微分模型（即SIRC-EDDM）的数值解提供了一种随机计算结构。提出了尺度共轭梯度（CG）神经网络（SCGNNs）的设计方法，用于SIRC-EDDM的数值处理。该数学模型分为易感、康复、感染和交叉免疫四类，而数值性能分为三种不同情况。通过比较已完成的结果与参考结果（龙格-库塔格式）以及SIRC-EDDM每种情况下约为10到10的可忽略不计的绝对误差（AE），可以看出SCGNNs的准确性。利用训练、验证和测试数据的性能，给出了所得结果以降低均方误差（MSE）。还进行了神经元分析，结果表明，与4个神经元相比，采用14个神经元时AE更高。为了检验SCGNNs的有效性，可通过误差直方图（EHs）研究、状态转换（STs）值、MSE性能、回归度量和相关性进行全面研究。
