Bayesian networks applied to customer surveys
Bayesian networks are gaining popularity in a wide range of application areas such as risk management, web data analysis, and management science. Availability of software for analysing Bayesian networks is further expanding their role in decision analysis and decision support systems. This chapter introduces Bayesian networks and their application to customer satisfaction surveys. The chapter begins with a theoretical introduction to Bayesian networks. Then Bayesian networks are applied to the ABC survey data set and to the Eurobarometer transportation survey. A summary section concludes the chapter. The chapter proceeds with an overview of publicly available software programs that can be used to implement Bayesian networks and a section comparing the goals of predicting and explaining with Bayesian Networks.
11.1 Introduction to Bayesian networks
A survey with n questions produces responses that can be considered random variables, X1, …, Xn. Some of these variables, q of them, are responses to general questions such as overall satisfaction, recommendation or repurchasing intention. These are considered target variables. Responses to the other n – q questions can be analysed under the hypothesis that they are positively dependent with target variables. The combinations (Xi, Xj), Xi, X1, …, Xn– q, Xj Xn– q+1, …, Xn, are either positive dependent or independent, for each pair of ...