Use of Structural Equation Modelling and Neural Network to Analyse Shared Parking Choice Behaviour

Abstract

The shared parking mode represents a feasible solution to the persistent problem of parking scarcity in urban areas. This paper aims to examine the shared parking choice behaviours using a combination of structural equation modelling (SEM) and neural network, taking into account both the parking location characteristics and the travellers’ characteristics. Data were collected from a commercial district in Nanjing, China, through an online questionnaire survey covering 11 factors affecting shared parking choice. The method involved two steps: firstly, SEM was applied to examine the influence of these factors on shared parking choice. Following this, the seven factors with the strongest correlation to shared parking choice were used to train a neural network model for shared parking prediction. This SEM-informed model was found to outperform a neural network model trained on all eleven factors across precision, recall, accuracy, F1 and AUC metrics. The research concluded that the selected factors significantly influence shared parking choice, reinforcing the hypothesis regarding the importance of parking location and traveller characteristics. These findings provide valuable insights to support the effective implementation and promotion of shared parking.