A Two-Step Algorithm to Estimate Variable Importance for Multi-State Data: An Application to COVID-19

Survival data with a multi-state structure are frequently observed in follow-up studies. An analytic approach based on a multi-state model (MSM) should be used in longitudinal health studies in which a patient experiences a sequence of clinical progression events. One main objective in the MSM framework is variable selection, where attempts are made to identify the risk factors associated with the transition hazard rates or probabilities of disease progression. The usual variable selection methods, including stepwise and penalized methods, do not provide information about the importance of variables. In this context, we present a two-step algorithm to evaluate the importance of variables for multi-state data. Three different machine learning approaches (random forest, gradient boosting, and neural network) as the most widely used methods are considered to estimate the variable importance in order to identify the factors affecting disease progression and rank these factors according to their importance. The performance of our proposed methods is validated by simulation and applied to the COVID-19 data set. The results revealed that the proposed two-stage method has promising performance for estimating variable importance. © 2023 Tech Science Press. All rights reserved.

A Machine Learning Approach for Diagnosing and Identifying Symptoms of COVID-19

Coronavirus 2019 (COVID-19) is a pandemic that hit the world and was responsible for the death of millions and the life disruption of billions of people. One of the most critical challenges faced during the earlier breakthrough of the diseases was identifying symptoms confused with colds, flu, and other common infections. Nevertheless, despite all the effort and research conducted for this purpose, this challenge continues as more strains, variants, and mutations appear. This work presents a solution for this problem based on machine learning classification and variable importance algorithms. A public dataset of 274,957 cases has been classified into typical and COVID-19 cases based on the reported symptoms and other variables. The dataset was used for classifying the reported cases using K-nearest neighbor (KNN), Naïve Bayes, and Decision Trees (DT) algorithms and identifying the significant symptoms that were decisive in classifying the patients using Gini, Information Gain, and Information Gain Ratio algorithms. Naïve Bayes and Decision Trees performed best with a Classification Accuracy (CA) score of 95.2% and 96.3%, respectively. The Naïve Bayes classifier scored an Area Under the Curve (AUC) of 88.75%. In addition, the applied variable importance algorithms identified headache, fever, and sore throat as the most important symptoms. © 2022 IEEE.

Exploring Anxiety of Québec University Community during COVID-19 Pandemic via Machine Learning

Many studies showed that COVID-19 global pandemic had a negative impact on the mental health of post-secondary students over the world. To date, very few studies have been conducted in a university setting, not only with students but also with employees. Moreover, almost all studies were based on classical statistical analysis. In this study, we investigated the level of anxiety felt by the Quebec university community (students and employees) during COVID-19 pandemic. Especially, we focused on the generalized anxiety disorder (GAD-7) score with the help of classical data exploration and predictive machine learning techniques. We observed that the best predictive model of the GAD-7 score was provided by the CatBoost algorithm) reaching a squared Pearson correlation coefficient of r2 = 0.5656. Moreover, we also explored variable importance and interaction effects between variables involved in the predictive model obtained using SHapley Additive exPlanations (SHAP). © 2022 ACM.