Combined structural equation modelling- A rtificial neural networks model for predicting customer loyalty

Abstract

Customer loyalty becomes considerations by service providers to maintain for reducing the churn rate. Many studies propose factors that are significantly influencing customer loyalty, and apply them for predicting it. Based on mathematical models, loyalty prediction methods are developed, and it involves new approaches including machine learning. This research aim is predicting customer loyalty using the combination of structural equation model (SEM) and artificial neural networks (ANN). The methodology starts by applying SEM for selecting statistically significant factors affect the loyalty. The linear SEM model ensures this relationship by fulfilling statistical hypothesis and fulfilled assumptions. Once selected factors are found, they are treated as inputs for ANN modelling. ANN is selected because of its ability in nonlinear modelling to enhance its prediction. ANN then learns the relationship between those inputs and the loyalty in real time as any additional observation recorded in. Based on trained ANN, prediction of customer loyalty based on input factors could be done. A case study was conducted at a Hotel by asking 130 customers. SEM inputs includes tangibles, facility, and staff attitudes, while loyalty scores become output. Combination of SEM-ANN has successfully predicted the customer loyalty and brought up chances for improvement strategies.