Comparative Analysis of Machine Learning Algorithms with Ensemble Techniques and Forecasting COVID-19 Cases in India

Unanticipated information in December 2019 changed the world around us. A relatively contagious disease unfolded through the SARS-CoV-2 virus that travelled throughout the globe and was declared an epidemic by WHO in March 2020. The need of examining the scenario became the inducement behind this research. The assessment of COVID-19 in India is performed from 1 April 2020 to 20 May 2021 which amassed a total of 415 instances. Further, preprocessing of the dataset is executed with the use of normalization. The experimentation is executed through the use of four ensemble strategies which are bagging, boosting, stacking and voting with four distinct machine learning algorithms linear regression, sequential minimal optimizer for regression, multilayer perceptron and Gaussian process. The splitting of the dataset is completed at 75%, and machine learning algorithms with ensemble techniques are applied. Linear regression with the bagging ensemble method gives satisfactory outcomes with the correlation coefficient of 0.935 and 0.919 for confirmed cases and recovered cases, respectively, and Gaussian process presented the best results for deceased cases. In the case of ensemble strategies, bagging indicates the best correlation coefficient in each case. Therefore, with the help of the three best algorithms, confirmed cases, recovered cases and deceased cases predictions are performed. The paper has potential implementations that can foresee the COVID-19 confirmed cases, recovered cases and deceased cases based on historic data and subsequently structure the plan for the future.