Decision support for real-time telemarketing operations through Bayesian network learning

Many knowledge discovery systems have been developed in diverse areas, but few systems address the use of knowledge in decision problems explicitly. This paper presents a decision support system for real-time telemarketing operations using the information extracted from the Bayesian network learning model. A prototype decision support system was developed for AT&T customer-contact employees to provide a recommendation regarding the promotion of a telephone discount plan. The system integrated a Bayesian network learning model (knowledge discovery process) and decision-making technique (influence diagram) to provide real-time decision support. A Bayesian network learning model was used to predict a probability of the customer's response from the previous promotion/response history. The influence diagram framework was used to integrate the predicted probability with the cost and benefit related to the possible actions. It was demonstrated that decision support by the Bayesian network learning model itself can be misleading. However, by linking the Bayesian network learning model with rigorous decision-making techniques such as influence diagrams, the decision support system developed in this paper was shown to provide an intelligent decision advice.