A non-linear model approach for predicting soundscape perception of study areas

A large majority of the studies use linear regression-based models for soundscape modeling due to their easy applicability. Only a few studies have chosen non-linear structures, such as neural networks. Moreover, students' perceptions of soundscape quality in study areas have yet to be explored. We aimed to predict soundscape perception of study areas by applying neural network models. We also compared our results with models applying linear approaches. Perceptual dimensions were obtained by applying the Principal Component Analysis. In this study, we used the data from a two-phase experiment to find Turkish soundscapes' affective quality attributes. The experiment was conducted in a well-insulated, quiet laboratory room at Bilkent University, and most of the participants were university students and Bilkent University faculty members.