Prediction Models of Polycyclic Aromatic Hydrocarbons in Urban Road Dust via Magnetic Properties, Particle Size, and Urban Environmental Features.

Key Points

• The aim is to develop accurate models for predicting polycyclic aromatic hydrocarbons (PAHs) in urban road dust.
• Utilized linear regression, artificial neural networks (ANN), and random forests (RF) to model PAH accumulation.
• Integrated magnetic properties, particle size, and urban environmental features in predictive analysis.
• Evaluated model performance through correlation coefficients (around 0.18 for linear regression, 0.26 for ANN, and 0.40 for RF).
• Integration of features through machine learning effectively identifies pollution hotspots.
• Model performance indicates complex interactions govern PAH distribution rather than simple predictors.
• The findings enhance understanding of urban pollution mechanisms.