Prediction Algorithm for Drugstore Consumption Members

It is an important part of precision marketing to identify potential consumer members of specific activities in pharmacies based on the big data formed by members' historical consumption information and participation information. The problem of consumer member prediction is abstracted as the binary classification problem of whether the members will use the discount to consume in the drugstore after the drugstore gives the discount to the members. In order to solve the limitations of the traditional marketing model applied to the drugstore and the low accuracy of predicting consumer members, this paper proposes a multi angle feature expression and prediction model for drugstore members. According to the characteristics of pharmacies, such as the particularity of pharmacies and the association between drugs and members' diseases, this paper puts forward a variety of new features from the three perspectives of members' attributes, dynamics and consumption, and constructs a feature set by combining the new features with traditional consumption features. Through a variety of machine learning algorithms to build pharmacy consumer member prediction model and comparison, the results show that the ensemble learning algorithm based on XGBoost has high accuracy in consumer member prediction.