COVID-19 Prediction using Linear Regression and Discrete Wavelet Transform

Corona virus illness (COVID-19), a respiratory condition affecting people in many countries, has swiftly spread to pandemic levels. A respiratory infection (like influenza) is the outcome of the condition, and signs and symptoms include breathing difficulty, cough, cold, and fever. The prediction of new COVID-19 occurrences is not an easy task. This work aims to improve the time-series prediction model based on linear regression by balancing the dataset through oversampling and choosing features using the discrete wavelet transform. Using WHO official data on the COVID-19 pandemic globally, the revised model (ELR) was evaluated and compared with numerous cutting-edge, contemporary algorithms. The ELR model surpassed the current Linear regression model (accuracy=94.7%, MAPE=0.052, MPE=0.486, MAE=0.029, and RMSE=8.383) with accuracy=98.4%, MAPE=0.016, MPE=0.308, MAE=0.016, and RMSE=3.383. ELR model is an effective, quick and easy method to predict COVID-19 trends using open-access data. Our research provides a workable solution for imbalanced Covid-19 data classification with limited sample numbers and high-dimensional characteristics. The suggested method may offer a theoretical foundation for computer-assisted diagnostics and allow for the prediction of any illness that may affect patients during upcoming pandemics.