BP neural network and rough set theory play an important role in the field of prediction. In view of the present situation of customer churn in logistics industry, this paper combines rough set and BP neural network to forecast customer attrition behavior in logistics industry. Firstly, using rough sets to extract rules from normal and abnormal customers to distinguish customer classes in logistics industry. Discrete processing of information entropy of extracted logistics customer attributes based on rough sets being good at handling discrete data. Finally, according to the strong mobility of logistics customers, Adam algorithm is introduced to build an adaptive BP neural network training model. The model proposed in this paper is more suitable for real-time data processing. The experiment proves that the method is feasible and efficient.