Prediction of COVID-19 Cases based on Human Behavior using DNN Regressor for Canada

Tripathy, D.; Camorlinga, S. G.

The proposed work utilizes Deep Neural Network (DNN) regression model to predict the total number of cases, new cases, and death cases in Canada. It is evaluated based on human behavior such as isolation, wearing mask outside home, contact with symptomatic person, washing hands, and other behaviors. The dataset is collected for the period of March 9th, 2020 to November 2nd, 2020 for Canada. The proposed methodology uses multiple-input deep neural network regression model with Rectified Linear Unit (Re LU) Function as the activation function, five non-linear dense layers of 64 Unit and a single unit of last layer as output for the curve fitting. The dataset is split into train and test sets with test size 20% and training size 80%. A nonlinear regression model is applied to the normalized data for making accurate predictions. The model performance is evaluated based on Root Mean Square Error (RMSE). Also, the Mean Absolute Error (MAE) is estimated for the model to quantify the error between predicted and true values. The results show that the proposed machine learning (ML) method predicts with high accuracy and can also be a convenient tool in making predictions for other countries. © 2021 IEEE.