A Heterogeneous Multi-Model Integration Approach for Unbalanced Direct Marketing

The crucial factor for enterprises to succeed in direct marketing with the help of Internet technology is to deal with massive unbalanced data and extract useful information. Traditional classification algorithms tend to favor majority categories and ignore minority categories. Therefore, solving the unbalance of direct marketing data becomes the key to precision marketing. Here, we propose a heterogeneous multi-model integration approach from the data pre-processing level, feature level and classification algorithm level to improve the prediction effect of unbalanced direct marketing. Using multiple evaluation indicators such as Weighted_Fl and AUC, the experimental evaluation on two publicly-available data set shows that the proposed heterogeneous multi-model integration approach achieves the best recognition performance. This approach is effectively used to explore whether users will respond to promotional activities, so as to help enterprises carry out targeted marketing plans and save redundant expenses.