Neural network modeling of consumer satisfaction in mobile commerce: An empirical analysis

Abstract The mobile commerce (m-commerce) industry has rapidly grown in value in recent years, as has the number of m-commerce service providers and interest in it from consumers and academia alike. In order to ensure customer loyalty, providers must determine which factors influence consumer satisfaction in m-commerce. Therefore, the objective of this study is to determine and rank the significant predictors of satisfaction in m-commerce. The paper also develops a procedure for artificial neural network model design and parameter setting in technology acceptance studies. Data was collected from 224 users of m-commerce services. The results presented are based on a combination of structural equation modeling (SEM) and artificial neural network (ANN) analyses. A multi-layer perceptron was used for ANN modeling. The results show that the optimal ANN model has one hidden layer and a sigmoid as an activation function in both layers, while the number of hidden nodes should be determined using a recommended rule-of-thumb. In addition, mobility and trust were found to be the most significant determinants of consumer satisfaction in m-commerce. The results of the study are significant as they have important implications for both academia and companies, due to the fact that some of the factors investigated in the study, such as mobility, have rarely been explored in previous consumer satisfaction studies, but were proved to be very significant. Another important result of the study is the proposal of a detailed procedure of ANN model design and the recommendations made for the selection of ANN model architecture and parameter settings.