TY - JOUR ID - 4930 TI - Regression and validation studies of the spread of novel COVID-19 in Iraq using mathematical and dynamic neural networks models: A case of the first six months of 2020 JO - Caspian Journal of Environmental Sciences JA - CJES LA - en SN - 1735-3033 AU - A. Khadom, Anees AU - Khudhair Al-Jiboory, A AU - S. Mahdi, Mustafa AU - B. Mahood, Hameed AD - Department of Chemical Engineering, College of Engineering – University of Diyala – Baquba City 32001, Diyala governorate, Iraq AD - Department of Mechanical Engineering, College of Engineering – University of Diyala – Baquba City 32001, Diyala governorate, Iraq AD - Department of Chemical and Process Engineering, Faculty of Engineering and Physical Sciences, University of Surrey, Guildford GU2 7XH, UK Y1 - 2021 PY - 2021 VL - 19 IS - 3 SP - 431 EP - 440 KW - Mathematical modelling KW - COVID-19 KW - Statistical analysis KW - Confirmed cases KW - Neural networks DO - 10.22124/cjes.2021.4930 N2 - The dramatic spread of COVID-19 has put the entire world at risk. In this work, the spread of COVID-19 in Iraq has been studied. Due to the increase in the number of positive cases and deaths with this disease, huge pressure acts on the economy and world professionals worldwide. Therefore, building mathematical models to predict the growth of this serious disease is extremely useful. It helps to predict the future numbers of cases in Iraq. Therefore, dynamic neural networks and curve fitting techniques have been developed to construct such a model. Data from the World Health Organization (WHO) are used as a source for mathematical model construction. The period between 25, February to 18, June 2020 was used for regression, validation, and model construction of Dynamic Neural Networks (DNN). Nine samples (19 – 27 June 2020) were used for predicting the future infected and death cases. Descriptive statistical studies showed that the standard deviation varies sharply on June as compared with earlier months of 2020. Three mathematical models are proposed. Linear, polynomials (2nd, 3rd, and 4th orders), and exponential models are used to correlate confirmed infected cases (CIC) and confirmed death cases (CDC) that represent the dependent variables as function of time (independent variable). Nonlinear regression based on least-square method is used to estimate the coefficients of models.  Exponential models were the most significant with 0.9964 and 0.9974 correlation coefficients for CIC and CDC, respectively. Validation analysis showed a significant deviation between real and predicted cases using exponential models. However, DNN models showed better response than exponential models. This is evidenced by objective and subjective comparisons. Finally, the CIC and CDC may be increased with time to approach 50000 and 2000 respectively at the end of June 2020. UR - https://cjes.guilan.ac.ir/article_4930.html L1 - https://cjes.guilan.ac.ir/article_4930_f2ccf866d8a24a0d6a872664ddbf6b4f.pdf ER -