Prediction of COVID-19 using recurrent neural networks

The COVID-19 pandemic has caused a worldwide health problem that has presented a lot of difficulties to the global healthcare sector, governments, and societies since management and control of the pandemic depend heavily on accurate estimates of the disease's spread and severity. This thesis investigates the application of deep learning models, such as Uni-LSTM, Bi-LSTM, and DNNs, for COVID-19 prediction. It gathers and examines actual data from various nations, such as figures for cases, hospitalizations, and fatalities. Deep learning models train and evaluate the data, comparing the results based on metrics like Accuracy, Precision, Recall and F1 score. The study's findings demonstrate that Bi-LSTM outperforms other models in accurately predicting the spread and severity of the disease. In the Bidirectional LSTM algorithm, the accuracy rate was 95.38% and the R-Squared value was 99.81%. This study recommends using a deep learning-based approach to recognize Covid-19 and no-finding occurrences in chest X-ray images. It sheds light on applying LSTM and DNN systems for COVID-19 prediction. It emphasizes the significance of data quality and quantity in achieving accurate and trustworthy predictions. The classification performance of the trained models was evaluated using the above metrics.