Comparison between multiple linear regressions and artificial neural networks to predict urban sound quality

The purpose of this study was to develop a predictive model of urban sound quality from field survey data using multiple linear regressions and artificial neural networks (ANNs). In order to determine a soundscape pleasantness model, passers-by were asked to assess their environment mainly from an acoustic point of view but also from a global perspective (visual and air quality). Users were asked to evaluate the sound environment firstly as a whole and secondly listening to each perceived sound source. The investigation took place at the "Parc de la Tête d'Or" which is an urban park in the French city of Lyon, in two locations on both sides of the main park access. One hundred and twenty subjects, divided equally between the two locations and the three periods of the day (morning, afternoon and evening), were interviewed. Each one had to evaluate twenty-six subjective variables on a rating scale from 0 to 10. In order to propose a relationship between the soundscape pleasantness and the others twenty-five assessed variables, the collected data have been analysed according two models: multiple linear regressions and predictive method based on artificial neural networks were used and compared. The first method is useful to understand which variables explain the assessment of the soundscape pleasantness, but not the second one which can be considered as a "black box". However ANNs seem to better predict the soundscape pleasantness when a new set of data is tested.