Connection Network and Optimization of Interest Metric for One-to-One Marketing

With the explosive growth of data in electronic commerce, rule finding becomes a crucial part in marketing. In this paper, we discuss the essential limitations of the existing metrics to quantify the interests of rules, and present the need of optimizing the interest metric. We describe the construction of the connection network that represents the relationships between items and propose a natural marketing model using the network. Although simple interest metrics were used, the connection network model showed stable performance in the experiment with field data. By constructing the network based on the optimized interest metric, the performance of the model was significantly improved.