A Spatial Neural Network Application in Consumer Spatial Behavior Modeling

Spatial modeling of consumer behavior becomes more important with the keen competition in retail market. A spatial neural network-multinomial logit model (SNN-MNL) is appropriate to model spatial behavior, accommodating nonlinear utility function. In this article, we approach a new SNN-MNL model in modeling consumer spatial behavior, in which a variable named relative position is included especially measuring the influence of spatial structure on spatial behavior. The origin of the variable is from research on spatial interaction model. To evaluate the model consumer spatial behavior data is got from a survey. Two kinds of choice models, a neural network model and a multinomial logit model, are adapted here in modeling spatial behavior. The former (SNN-MNL) is found to outperform the later. Furthermore, the accuracy of the SNN-MNL model with our new variable added outperforms slightly the ordinary SNN-MNL model