Hybrid GA-ConvLSTM Model for Predicting the Transmission Pattern of COVID-19: A Case Study for Shanghai Cooperation Organisation

COVID-19 has spread very quickly to almost every part of the world, causing many people to experience severe symptoms and lose their lives. In this study, it is aimed to determine the transmission pattern of COVID-19 with deep learning methods so that plans can be made to alleviate the burden on healthcare systems and predict the distribution of the epidemic. For this purpose, the hyper-parameters of the hybrid deep learning model developed using CNN and LSTM models were optimized with genetic algorithm, and a more successful prediction performance was achieved. The GA-ConvLSTM model was tested with XGBoost, SVM, CNN, MLP, LSTM, and ConvLSTM to determine the spread of the epidemic in the member countries of SCO. The study used daily COVID-19 case and death data between 2020/01/03 and 2022/05/31 presented by WHO. Experiments showed that GA-ConvLSTM has over 0.9 R2 in case prediction for all countries. Experiments showed that GA-ConvLSTM has above 0.9 R2 for the majority of countries when it comes to death prediction. In addition, the transmission pattern of COVID-19 among the SCO countries was determined with the chord diagrams created using 5 and 14 days’ incubation periods.