Prediction of Coronavirus Using Various Machine Learning Algorithms

Abstract

Human life and health are in danger due to the coronavirus pandemic, which eventually reduces production. Artificial intelligence is crucial for predicting coronavirus cases because it allows us to identify patients based on their infection and predict their infection using machine learning models using disaggregated data from coronavirus cases. As machine learning models usually require private information, it is imperative to consider data privacy. So, to prevent this from happening, only a country-based data set is used by us in order not to invade or reveal anyone's private information. In this paper, the data set involves some basic questions and attribute answering to train the data model using logistic regression, decision tree, and SVM. Identifying which elements affect model prediction accuracy and loss, such as the number of characteristics, learning rate, rounds, and data size, is attempted during the model training stage. Each training band's model loss and prediction accuracy were recorded and shown to identify the variables influencing model performance. As a result, it was noted that Logistic Regression gives better prediction accuracy and loss for predicting whether a person is vulnerable to COVID-19. Air contamination is also a significant factor, but we focus on more human symptoms than the outsourced ones. The model's accuracy and prediction loss can barely be affected by changing the number of rounds and learning rate. These hyperparameters can be tuned to optimise the model's performance, but changes in their values are unlikely to affect the outcome significantly. The accuracy, performance, and loss differences between the different machine learning models were compared in the final analysis. It is concluded that the prediction model has better prediction accuracy and diminished loss in LR and SVM models, respectively.

Topics & Concepts