A microeconomic data mining problem: customer-oriented catalog segmentation

摘要: The microeconomic framework for data mining [7] assumes that an enterprise chooses a decision maximizing the overall utility over all customers where contribution of customer is function available on customer. In Catalog Segmentation, wants to design k product catalogs size r maximize number catalog products purchased. However, there are many applications customer, once attracted enterprise, would purchase more beyond ones contained in catalog. Therefore, this paper, we investigate alternative problem formulation, call Customer-Oriented measured by have at least specified minimum interest t catalogs. We formally introduce Segmentation and discuss its complexity. Then two different paradigms efficient, approximate algorithms problem, greedy (deterministic) randomized algorithms. Since may be trapped local optimum crucially depend reasonable initial solution, explore combination these paradigms. Our experimental evaluation synthetic real demonstrates new yield significantly higher compared classical