Develop an unsupervised attention-based LSTM network algorithm for forecasting infectious disease

time series Due to better algorithms, more accessible data, and higher computing power over the past ten years, forecasting has become more popular. It is used in a variety of industries, including as financial time series, weather forecasting, and medical diagnostics. In this study, we provide a model of the mechanism governing attention, which enables attended input to be provided to the model in place of actual input. In order for the model to produce more precise predictions, it seeks to demonstrate a fresh perspective on the data. The experiments were conducted with the (encoder-decoder) LSTM model as well to demonstrate the usefulness and superiority of the suggested strategy. The obtained results demonstrate that, when compared to the (encoder-decoder) LSTM base model, the proposed approach could reduce the mean square error (RMSE=9819.05), relative root mean square error (RRMSE=99.09), and coefficient of determination (R Square=0.96). The obtained results support the suggested approach’s efficacy, superiority, and importance in predicting SARS-CoV-2 infection cases.