TabNet Based Prediction Model for ICU admission in Covid-19 patients

Abstract

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is the responsible virus for coronavirus disease 2019 (COVID-19). It was reported the first time in Wuhan (China) by late December 2019. The COVID-19 pandemic has become a global health risk due to the urgent need for an Intensive Care Unit (ICU) that exceeded its capacity. To cope with this exponential spread the fast adoption of Artificial Intelligence (AI) tools and advanced technology is crucial. For this reason, many research works in AI are conducted. In the current paper, we intend to report AI applications and solutions based on machine learning, deep learning, and data mining algorithms for detecting, predicting, and diagnosing COVID-19. Furthermore, this study aims to develop a new deep learning-based method capable of predicting whether a COVID-19 patient requires admission to an intensive care unit using clinical tabular data from Kaggle. This model will contribute to the optimization of ICU resources. The experimental results showed that combining Synthetic Minority Oversampling Technique (SMOTE) and TabNet classifier improved the prediction performance and surpassed the state-of-the-art models: MLP, RF, LR, and KNN.

Ähnliche Arbeiten