Description
: We proposed a machine learning-based forecasting web portal named Novel Coronavirus (COVID-19) for Bangladesh that not only provide real-time evidence, trend analysis of this epidemic but also predict how many individuals will be affected in upcoming days respectively. Therefore, the regular growth factor of confirmed cases and deaths are also explored to assess the risk of COVID-19 state. The daily fatalities records of Bangladesh are gathered from open source GitHub repository of John Hopkin’s University. In this work, various machine learning-based regression models such as linear regression, polynomial regression, support vector regression, multi-layer perception, holt-linear, holt-winter, ARIMA and prophet model are used to investigate and forecast impending circumstances. From the first detection of COVID-19 patients in Bangladesh, the COVID-19 patients has not detected until April 14, 2020. After that, the number of cases has been amplified rapidly throughout the country. The facts before this out breaking state prevent to fit existing data with regression models. Hence, we trained regression models with the records of the last couple of days (e.g., 25 days) to generate more appropriate results that can fit with existing day-to-day records and shows fewer residuals. This date-wise dataset has been split into training and validation sets. This training set is fitted with the regression models and the root means square error (RMSE) is manipulated in view of the validation set. The regression model which is shown the best results (which shows the less RMSE) is available at our web portal. This instant forecasting model can estimate fatalities of forthcoming days where the optimum days can be fixed as usual (e.g., 7, 15, 30 days). To reduce residuals and fit the results of machine learning models with regular cases, the fine-tuning has been employed to the parameters of these models. We also integrated a machine learning-based chatbot which may help the user to access all information related to the COVID-19 outbreaking. When the user asked any query about COVID-19 (Mental Health, Relief Work, Hotline Information, etc.)
