Having a huge customer base, ebusiness customer churns rate is so high that the profit would be improved a lot, if the customer churns rate could be a little bit lower. However, it's difficult to judge which one is lost for ebusiness firms. Pareto/NBD model is a powerful, yet simple, wide spread model to predict customer activity. The model is based on five well empirical validated hypotheses. At the individual level, however, the prediction accurate of Pareto/NBD model is often poor.Pareto/NBD model takes use of only three sufficient summaries of a customer's transaction history, while there is so many covariates were recorded, along with purchasing information. Probability model and data mining approach each has its applying domain. Probability model is robust; data mining is good at deal with huge dimensions. From the point of probability model based data mining approach, 3 strategy of assembling Probability model and data mining algorithm were proposed.Data mining based on prior estimated by probability modelData mining based on score function of probability modelAdaboosting based probability modelingGuided by the first strategy, a Pareto/NBD + Bayes model was proposed... |