Aesthetic preference and mental restoration prediction in urban parks: An application of environmental modeling approach

摘要

Urban parks enhance the aesthetic quality of urban area and increase the restorative potential of cities to avoid the negative psycho-physiological impact of living in the built environment. This research aimed to model some aesthetic preference and mental restoration values in urban parks based on landscape natural characteristics to compare the MLP (Multi-Layer Perceptron), RBFNN (Radial Basis Function Neural Network) and SVM (Support Vector Machine) models in urban parks aesthetic and mental restoration potential prediction. Therefore, we recorded 11 landscape characteristics in 200 urban parks. We developed the landscape model to predict aesthetic and mental restoration potential using data mining techniques such as MLP, RBFNN, and SVM. The SVM model was developed as the most accurate model to predict the landscape score of urban parks. SVM the model represents the highest value of R-2 in training (0.9), test (0.83) and all data sets (0.88). According to the sensitivity analysis, trees, water bodies, buildings, flowers, and decorations in park landscapes were prioritized respectively as the most significant inputs influencing the SVM model outputs. The results of SVM, especially its determined accuracy (R-2 = 0.83) in comparison with MLP (R-2 = 0.77), and RBFNN (R-2 = 0.78) test results showed that SVM is the most successful comparative landscape assessment model in aesthetic and mental restoration potential prediction. Using MATLAB software, the SVM model is applicable in urban parks where the characteristics of the newly designed landscape are in the range of studied area. The SVM modeling technique would be applicable for landscape architectures to model the landscape of urban areas. The urban park landscapes with more trees, water bodies, flowers, decorations, and fewer buildings would likely attract citizens' attraction and recover their mental stresses. In practice, the designed graphical user interface is applied by landscape architects to predict the landscape score.