Mathematics · Modeling and Simulation
Adaptive Elastic-net Sliced Inverse Regression to Identify Risk Factors Affecting Covid-19 Disease Fatality Rate
Sajedeh Lashgari, Mohsen Mohammadzadeh, Foad Ghaderi
See on the mapAbstract
Abstract Identifying and modeling variables that affect coronavirus disease (Covid-19) infection fatality rate leads to better macro decisions and reduces the spread of the virus. Moreover, identifying these variables leads to better understanding and interpretation of the prediction models and can improves their performance. So far, different methods have been proposed for variable selection. However, they often face challenges such as non-modeling, high computational cost, overfitting, following certain assumptions like specific distribution, specific model, or lack of collinearity. They also do not consider sufficient dimensions. In this article, to overcome these limitations, we propose a method based on Adaptive Elastic-net Sliced Inverse Regression to identify risk factors for the Covid-19 disease in the presence of collinearity between explanatory variables. Considering the penalty of elastic-net and sliced inverse regression, this method leads to sufficient dimension reduction and presentation of a more stable and accurate model and prevents the occurrence of overfitting. We applied the proposed method to simulated data and new real-world Covid-19 disease dataset with 60 explanatory variables that we collected from available sources and prepared after going through pre-processing steps. We observed that the proposed method has significantly improved by reducing the experimental standard error of bootstrapping by 13\% compared to the previous methods for both datasets. So, it is an efficient and stable method for modeling and identifying risk factors for the Covid-19 disease. This method also has high generalizability, prevents overfitting, and can be used for other similar diseases.