Multi-campaign Assignment Problem and Optimizing Lagrange Multipliers

Customer relationship management is crucial in acquiring and maintaining royal customers. To maximize revenue and customer satisfaction, companies try to provide personalized services for customers. A representative effort is one-to-one marketing. The fast development of Internet and mobile communication boosts up the market of one-to-one marketing. A personalized campaign targets the most attractive customers with respect to the subject of the campaign. So it is important to expect customer preferences for campaigns. Collaborative Filtering (CF) and various data mining techniques are used to expect customer preferences for campaigns. Especially, since CF is fast and simple, it is widely used for personalization in e-commerce. There have been a number of customer-preference estimation methods based on CF. As personalized campaigns are frequently performed, several campaigns often happen to run simultaneously. It is often the case that an attractive customer for a specific campaign tends to be attractive for other campaigns. If we perform separate campaigns without considering this problem, some customers may be bombarded by a considerable number of campaigns. We call this overlapped recommendation problem. The larger the number of recommendations for a customer, the lower the customer interest for campaigns. In the long run, the customer response for campaigns drops. It lowers the marketing efficiency as well as customer satisfaction. Unfortunately, traditional methods only focused on the effectiveness of a single campaign and did not consider the problem with respect to the overlapped recommendations. In this paper, we define the multi-campaign assignment problem (MCAP) considering the overlapped recommendation problem and propose a number of methods for the issue including a genetic approach. We also verify the effectiveness of the proposed methods with field data. Let N be the number of customers and K be the number of campaigns. The MCAP is to find customer-campaign assignments that maximizes the effects of campaigns. The main difference with independent campaigns lies in that the customer response for campaigns is influenced by overlapped recommendations. More detailed description is omitted by space limit. In case of overlapped campaign recommendations, the customer response rate drops as the number of recommendations grows. We introduce the response suppression function for