Artificial neural network analysis of the relationship between road-traffic noise and air pollutants and urban form indicators

Road-traffic noise and air pollutants have adverse effect to urban dweller's health and life quality. For management of the noise and air pollutants, noise and air pollution maps can be used to provide quantitative information of noise and air pollution exposure levels. In this study, the more efficient method of noise and air pollution mapping was developed statistically. Noise and air pollutants' exposure change by small-scaled alteration of urban planning can be predicted by the method. The relationship between road-traffic noise level and air pollutants and urban forms for roads, buildings and land-use was analyzed by artificial neural network analysis. The selected representative urban form indicators are road-related(traffic volume, speed), building-related(floor space index, ground space index), and land-use-related indicators. The artificial neural network model was optimized by adjusting the number of hidden nodes and layers. The 2/3 of data sets extracted from a region was used for the model development to select the model with the least prediction error. The selected model was applied to the remaining 1/3 of data sets for verification. The result from the artificial neural network model were compared with that from engineering model.