Prediction of particulate matter at street level using artificial neural networks coupling with chaotic particle swarm optimization algorithm

• Identify the influencing factors on variation of particulate matter (PM) at street level.
• Propose a hybrid model to predict PM concentration at urban intersection.
• Clarify deeper dependence of fine particle on traffic/meteorological conditions than coarse particle.

The time series of particulate matter at urban intersection consists of complex linear and nonlinear patterns and are difficult to forecast. Artificial neural networks (ANNs) have been applied to air quality forecasting in urban areas, but they have limited accuracy owing to their potential convergence to a local minimum and over-fitting. Chaotic particle swarm optimization (CPSO) algorithm is chaos-based searching algorithms and can recognize nonlinear patterns. Hence, a novel hybrid model combining ANN and CPSO algorithm is proposed to improve forecast accuracy. The proposed model, together with the ANN model with the traditional algorithms (Levenberg–Marquardt and PSO), is examined with the measured data in spring and winter respectively. The proposed model is found to provide the best results among them, implying that the hybrid model can be an effective tool to improve the particulate matter forecasting accuracy. Additionally, the proposed model is found to perform better for fine particles than for coarse particles. The model is also verified to predict better in winter than in spring. The outputs of these findings demonstrate the potential of the proposed model to be applied to forecast the trends of air pollution in similar meso-to mega-cities.