|M.Sc Student||Rony Ghebali|
|Subject||Real-Time Prediction of the Probability of Abandonment in|
|Department||Department of Industrial Engineering and Management||Supervisors||Full Professor Mandelbaum Avishai|
|Professor Gorfine-Orgad Malka|
|Full Thesis text|
During the last several decades, call centers have turned out to be a central part of every service enterprize, over a very wide range of fields such as banking, technical support, telecommunications and government. Call centers provide the customer with a direct access to services via an online operator who can answer questions and provide the requested services. While waiting to be served, customers become impatient and some abandon. This research deals with predicting the probability of abandonment, in real time, for a specific customer who is waiting during a repeated call. In other words, we develop dynamic prediction models for the probability that an identified customer will hang up before being served, based on this customer history of past calls and present waiting. We provide two prediction approaches: marginalized and conditional: both can be easily implemented on line. Such predictions can support performance management of call centers. For example, suppose that the predicted probability of abandonment is high according to some criteria; then the relevant customer would be assigned a high priority for being answered. Our prediction approaches are based on multivariate survival analysis settings with frailty models. Specifically, via an extensive simulation study and real data analysis, we assess
the performance of the proposed methods under log-normal and gamma frailty models.