Events

Urban Science Research Seminar: Ensemble machine learning model to analyze the association between admissions in Emergency Room and environmental factors

Speaker: Vito Telesca

Location: 370 Jay Street, Room 1201

Date: Wednesday, March 12, 2025

Complete Location: 370 Jay St, RM 1201, Brooklyn, NY 11201 RM. 233 (Co-Lab)

This study proposes an ensemble machine learning model to analyze the association between cardiovascular, respiratory, and total admissions in Emergency Room (ER) and environmental factors, such as air pollution and weather-climatic conditions. The aim is to improve the understanding of interactions between these factors, considering non-linearities and temporal dependencies, overcoming the limitations of traditional methods, and providing new perspectives on the relationships between diseases, pollution, climate, and hospital admissions. The analysis was conducted using ensemble learning techniques, applying three regression models: Random Forest, XGBoost, and Adaboost. Ensemble learning improves predictive power by combining different models to reduce overfitting and variance, offering more robust predictions. The Bayesian optimization technique was applied to improve the accuracy of the predictions. The performance of the three optimized models was evaluated through various metrics, also considering cross-validation the k-fold technique (k=10) to provide a more robust estimate of predictive capabilities. The application of SHAP (SHapley Additive exPlanations) analysis allowed us to identify the most important variables for hospital admissions and related patterns.