Spatial Prediction of High-Risk Areas for Asthma in Metropolitan Areas: A Machine Learning Approach Applied to Tehran, Iran

Asthma prevalence in large urban areas of developing countries is a significant public health concern, with increased rates driven by various socioeconomic and environmental factors. This study aims to predict asthma risk in Tehran, a major urban center in Iran. Data from 1473 asthma patients, alongside demographic, socioeconomic, air quality, environmental, weather, and healthcare access variables, were analyzed using geographic information systems (GIS) and remote sensing techniques. Three ensemble machine learning algorithms—Random Forest (RF), Gradient Boosting Machine (GBM), and Extreme Gradient Boosting (XGBoost)—were applied to model and predict asthma risk. A Negative Binomial Regression Model (NBRM) identified seven key predictors: population density, unemployment rate, particulate matter (PM2.5 and PM10), nitrogen dioxide (NO2), sulfur dioxide (SO2), neighborhood deprivation index, and road intersection density. Among the algorithms, GBM outperformed the others, with a training RMSE of 0.56 and a test RMSE of 1.07, demonstrating strong generalization. Additionally, GBM achieved the highest R-squared values (0.95 for training and 0.76 for testing) and lower MAE values (0.43 for training and 0.88 for testing). Effective pattern recognition was confirmed by EV values of 0.95 for training and 0.75 for testing, along with a Moran’s I value of 0.17, indicating minimal spatial autocorrelation.