The Effect of Social Affinity and Predictive Horizon on Churn Prediction Using Diffusion Modeling

The social influence of people on their peers in the selection of products and services is frequently modeled as a diffusion process. Recently, such processes have been successfully applied as a tool for the prediction of customer turnover, or churn, in mobile communication carriers. Such prediction is most accurate when the appropriate social ties are used, and is primarily useful when it provides a long forecast horizon, so as to enable the service provider to take mitigating actions. In this work we investigate several measures of social affinity and compare their performance for churn prediction. We demonstrate that these measures capture different calling patterns and show that combining these measures can significantly improve the accuracy of the prediction. We study the predictive horizon of diffusion processes and show that it deteriorates significantly as the horizon increases. Our results from two large mobile phone carriers show how the usefulness of diffusion processes can be enhanced for churn prediction and provide insights to their limitations.

By: Dorit Baras, Amir Ronen, Elad Yom-Tov

Published in: H-0317 in 2012

H-0317.pdf

Questions about this service can be mailed to reports@us.ibm.com .