Bayesian Networks and Structural Equation Modelling to develop service quality models: Metro of Seville case study

Abstract

Service Quality (SQ) in Public Transport (PT) has been a crucial aspect to improve for years because of its strong influence on user satisfaction and its capacity to attract new passengers. Different techniques have been applied for analysing SQ and Structural Equation Modelling (SEM) is one of the most widely used due to its ability to address different kinds of variables and to model a whole phenomenon occurring at one time. Nevertheless, its confirmative nature requires previous knowledge, a hurdle that can be overcome by applying Bayesian Networks (BN) as a technique that learns directly from data without pre-assumptions. The aim of this paper is to apply a novel methodological approach in the field of SQ, based on a two-step process, which combines the techniques of BN and SEM, to model SQ in the Metropolitan Light Rail Transit (LRT) Service of Seville (Spain). In other words, in this paper, the proposed methodological approach has been applied to extract and confirm, directly from data and without necessity of assumptions, the possible relationships among the LRT service characteristics and how they are related with passengers’ overall SQ perception. For this purpose, firstly, a BN was automatically learnt from the data and allowed to establish relationships between SQ dimensions describing the service. SEM was then used to check the SQ model and the relationships between the dimensions extracted from the BN. The model fit parameters of SEM and its consistency with the real life expected scenario supported and validated the SQ model designed in this study. Furthermore, the different relationships among dimensions extracted from BN were analysed and support the usefulness and potential of this methodological process that could lead to the development and confirmation of new theories and models in any field of knowledge based on data and expert supervision