|
|
 |
| |
|
|
|
Artificial Intelligence models versus a machine learning aided model for severity prediction of COVID-19: A Population-Based Study |
|
|
|
PP: 131-139 |
|
|
|
doi:10.18576/amis/190111
|
|
|
|
Author(s) |
|
|
|
Eman Khorsheed,
Hawra Alshowaikh,
|
|
|
|
Abstract |
|
|
| The innovative potentials of the Artificial Neural Network (ANN) and the Random Forest artificial intelligence technologies along with a machine learning aided logistic regression technique were utilized to predict COVID-19 severity. The three models were compared based on sensitivity, specificity, and overall model accuracy. Data of 205 patients were analyzed in this study. 70.6% of the observations were used for model training and 29.4% for model validation. In each approach, ten potential risk factors for severity symptoms were considered and evaluated. At the training phase, the overall accuracy of the ANN, logistic regression, and Random Forest models were 84.7%, 83.3%, and 81.4%, respectively. However, the ANN model outperformed the other two models during the validation process and scored an overall accuracy of 85.0% versus 71.7%. This superior performance underscores the robustness of the ANN approach in the context of predictive accuracy for COVID-19 severity.
|
|
|
|
|
|
|
|
