MBBCTREE Algorithm for telecom customer retention model

For the purpose of improving the predictive accuracy and interpretability of telecom customer retention prediction model, MBBCTree Algorithm, which integrates the advantage of Markov blanket Bayesian networks (MBBC) and decision tree, is proposed in this paper to predict customers´ lifetime. MBBCTree classifier uses a tree structure to split the instance space into sub-spaces and generates one MBBC classifier in each sub-space. Experimental results show that the new algorithm outperforms Naive-Bayes decision tree (NBTree), MBBC and Naive Bayes classifiers in accuracy. And customer retention modeling based on this new algorithm got effective results.