COVID-19 mortality and nutrition through predictive modeling and optimization based on grid search

Le résumé fourni par la source

Since 2019, humanity has been suffering from the negative impact of COVID-19, and the virus did not stop in its usual state but began to pivot to become more harmful until it reached its form now, which is the omicron variant. Therefore, in an attempt to reduce the risk of the virus, which has caused nearly 6 million deaths to this day, it is serious to focus on one of the most important causes of disease resistance, which is nutrition. It has been proven recently that death rates dangerously depend on what enters the human stomach from fat, protein, or even healthy vegetables. This study aims to investigate a relationship between what people eat and the Covid-19 death rate. The study applies five machine learning (ML) models as follows: gradient boosting regressor (GBR), random forest (RF), lasso regression, decision tree (DT), and Bayesian ridge (BR). The study utilizes an available Covid-19 nutrition dataset which consists of 4 attributes as follows: fat percentage, caloric consumption (kcal), food supply amount (kg), and protein levels of various dietary categories for the experiment. The experiment shows the GBR model without optimization obtained optimal results during comparison with other models. The GBR model achieved a mean squared error (MSE) of 0.1512, a mean absolute error (MAE) of 0.2262, mean absolute percentage error (MAPE) of 0.1351, and r 2 value of 0.963. The settings of the GBR model were refined using grid search (GS) hyperparameter optimization to find an optimal solution. This work employs evaluation strategies such as R 2 , MAE, MAPE and MSE to find the best-fitted model. The results displayed that the GS-GBR can enhance the performance of the original classifier compared with others from 96.3 to 99.4%. GS-optimized GBR predicts COVID-19 mortality rates better than other models, suggesting improvement in nutrition-related disease resistance predictions.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.