Abstract: Based on the best splitting point, a change mining methods of customer classification is proposed, which conquers the shortcomings of the existing change mining methods. In terms of the features of classification rule, the new definition of change types is proposed, based on which the metric of change and degree of change measures are redefined. In order not to affect the search for the best splitting point of each node in the decision tree, an extended measure of value match of quantitative attribute is designed to analyze the change measures between patterns. Empirical evaluation shows that the methodology is very effective to recognize the change of customer classification from two different datasets.