Bayesian Network Approach to Predict Mobile Churn Motivations: Emphasis on General Bayesian Network, Markov Blanket, and What-If Simulation
As mobile telecommunication service becomes indispensable to our daily life, predicting the reasons of mobile churn seems essential from the perspective of mobile service providers. Previous studies have been focused on mobile churn prediction itself, not churn motivations which can play as a good indicator to forecasting real churn. Therefore, main focus of this study is placed on predicting mobile churn motivations, instead of mobile churn prediction. We propose BN approach to predict mobile churn motivation, adopting three types of BN models such as Naïve BN (NBN), Tree Augmented NBN (TAN), and General BN (GBN). To prove its validity in predicting mobile churn motivations, benchmarking classifiers were adopted and their performance was compared with BN classifiers. Through analyzing the empirical results, we found three advantages of GBN-(1) GBN performance is competitive compared with other benchmarking classifiers, (2) Markov Blanket (MB) variables are considerably small in number and make it handy for decision makers, and (3) what-if simulation is possible, which is not possible in other benchmarking classifiers. Practical implications of empirical results were addressed.
KeywordsMobile Churn Prediction Churn Motivation Bayesian Network Markov Blanket What-If Simulation
Unable to display preview. Download preview PDF.
- 2.Bouckaert, R.: Properties of Bayesian belief network learning algorithms. In: Proc. 10th Annual Conf. Uncertainty Artificial Intelligence (UAI), Seattle, WA, pp. 102–110 (1994)Google Scholar
- 4.Cheng, J., Greiner, R.: Comparing Bayesian network classifiers. In: Proc. of the 15th Conf. on Uncertainty in Artificial Intelligence, pp. 101–107. Morgan Kaufmann Publishers, San Francisco (1999)Google Scholar
- 10.Koller, D., Sahami, M.: Toward optimal feature selection. In: Proc. 13th International Conf. Machine Learning, pp. 284–292 (1996)Google Scholar
- 15.Quinlan, J.R.: C4.5: Programs for Machine Learning. Morgan Kaufmann, San Mateo (1993)Google Scholar
- 16.Sarkar, S., Sriram, R.S.: Bayesian Models for Early Warning of Bank (1998)Google Scholar
- 17.Siber, R.: Combating the churn phenomenon. Telecommunications 31(10), 77–81 (1997)Google Scholar