Intelligent churn prediction for telecom using GP-AdaBoost learning and PSO undersampling
- 344 Downloads
Nowadays, telecom industry faces fierce competition in satisfying its customers. This competition thus requires an efficient churn prediction system to identify customers who are ready to quit. Such churn customers are then retained through addressing relevant reasons identified by the churn prediction system. Therefore, now the role of churn prediction system is not only restricted to accurately predict churners but also to interpret customer churn behavior. In this paper, searching capabilities of genetic programming (GP) and classification capabilities of AdaBoost are integrated in order to evolve a high-performance churn prediction system having better churn identification abilities. For this, frequently selected features in various GP expressions evaluated through AdaBoost based learning, are marked and analyzed. Moreover, the issue of imbalance present in telecom datasets is also addressed through particle swarm optimization (PSO) based undersampling method, which provides unbiased distribution of training set to GP-AdaBoost based prediction system. Particle swarm optimization based undersampling method in combination with GP-AdaBoost results a churn prediction system (ChP-GPAB), which offers better learning of churners and also identifies underlying factors responsible for churn behavior of customers. Two standard telecom data sets are used for evaluation and comparison of the proposed ChP-GPAB system. The results show that the proposed ChP-GPAB system yields 0.91 AUC and 0.86 AUC on Cell2Cell and Orange datasets, in addition to identifying the reasons of churning.
KeywordsTelecom Churn prediction Genetic programming AdaBoost Ensemble classification Feature identification
This work is supported by the Higher Education Commission of Pakistan (HEC) as per Award No. 20-3408/R&D/HEC/14/233.
- 1.ICT: The World in: ICT Facts and Figures, Geneva (2014)Google Scholar
- 2.Shin, Y.H., David, C.Y., Hsiu, Y.W.: Applying data mining to telecom churn management. Expert Syst. Appl. 37, 3665–3675 (2006)Google Scholar
- 13.Xu, H., Zhang, Z., Zhang, Y.: Churn prediction in telecom using a hybrid two-phase feature selection method. In: Third International Symposium on Intelligent Information Technology Application, 2009. IITA 2009, pp. 576–579 (2009)Google Scholar
- 21.Sorokina, D.: Application of additive groves ensemble with multiple counts feature evaluation to KDD Cup ’09 small data set. In: Presented at the JMLR Workshop and Conference Proceedings, Paris (2009)Google Scholar
- 35.Ning, L., Hua, L., Jie, L., Guangquan, Z.: A customer churn prediction model in telecom industry using boosting. IEEE Trans. Industr. Inform. 10, 1659–1665 (2012)Google Scholar
- 37.Amin, A., Khan, C., Ali, I., Anwar, S.: Customer churn prediction in telecommunication industry: with and without counter-example. In: Gelbukh, A., Espinoza, F.C., Galicia-Haro, S.N. (eds.) Proceedings on Nature-Inspired Computation and Machine Learning: 13th Mexican International Conference on Artificial Intelligence, MICAI 2014, Tuxtla Gutiérrez, Mexico, November 16–22, 2014, Part II, pp. 206–218. Springer International Publishing, Cham (2014)CrossRefGoogle Scholar
- 39.Miller, H., Clarke, S., Lane, S., Lonie, A., Lazaridiz, D., Petrovski, S., et al.: Predicting customer behaviour: the University of Melbourne’s KDD Cup report. In: Presented at the JMLR Workshop and Conference Proceedings, Paris (2009)Google Scholar
- 41.Idris, A., Khan, A., Lee, Y.S.: Genetic programming and adaboosting based churn prediction for telecom. In: 2012 IEEE International Conference on Systems, Man, and Cybernetics (SMC), pp. 1328–1332 (2012)Google Scholar
- 42.Busa-Fekete, R., Kegl, B.: Accelerating AdaBoost using UCB. In: Presented at the JMLR Workshop and Conference Proceedings, Paris (2009)Google Scholar