Predicting Carbon Monoxide Emissions with Multivariate Adaptive Regression Splines (MARS) and Artificial Neural Networks (ANNs)

Emissions from motor vehicles need to be predicted fairly accurately to ensure an appropriate air quality management plan. This research work explores the use of a nonparametric regression algorithm known as the multivariate adaptive regression splines (MARS) in comparison with the artificial neural networks (ANN) for the purpose of best approximation of the relationship between the input and output from datasets recorded from on-board measurement and dynamometer testings. The performance of the models was evaluated by comparing the MARS and ANN predictions to the measured data using several performance indices. The results are evaluated in terms of accuracy, flexibility and computational efficiency. While MARS are more computationally efficient to reach the final model ANN are slightly more accurate. The proposed techniques may be used to assist in a decision-making policy regarding urban air pollution.