An enhanced artificial neural network model using the Harris Hawks optimiser for predicting food liking in the presence of background noise

Abstract

This paper presents an enhanced artificial neural network (ANN) model for predicting the liking of food using the Harris Hawks optimiser (HHO) in the presence of different masking background noise types and levels, relative to the ambient background noise (i.e. no noise conditions). The results showed that the proposed model can predict the relative liking food ratings with higher performance (R2 = 0.70, RMSE = 0.8), as compared to traditional ANNs using feedforward neural networks (FFNNs) (R2 = 0.61, RMSE = 1.1) and statistical mixed models (R2 = 0.42, RMSE = 1.8). This model was used to find the threshold level that gives maximum relative food liking ratings for different types of noise. This threshold level varied between 30 and 35 dBA for three noise types. The liking of food in the presence of background noise depends on acoustic and non-acoustic factors. A feature analysis using the “RReliefF” algorithm was used to investigate the relative importance of these acoustic and non-acoustic factors on the relative food liking using predicted model outcomes. Acoustic factors such as noise type and level had higher importance weights on the relative liking of food in the presence of background noise than non-acoustic factors such as gender, sensitivity, and age. The results presented herein are relevant for future and more targeted noise assessment and mitigation strategies in dining areas.