Asthma Prediction Tool in Preschool Children: Evidence from Predictive Performance Comparison by Using Innovative Statistical Approach

Abstract

This study aims to develop a non-invasive tool for early asthma diagnosis in preschool children. The tool leverages co-morbidities, environmental factors, and socio-economic determinants. The research compares Traditional Statistical Models, Machine Learning Techniques, and Deep Neural Networks to assess the predictive capabilities of various models against the Leicester tool, which is used as a benchmark. The study’s results, evaluated on accuracy and AUC metrics, indicate that the proposed tool outperforms the Leicester model across all applied models. This research highlights the potential of integrating diverse variables for asthma prediction, suggesting directions for future enhancements and underscoring the need for comprehensive evaluations to validate these findings.