The Research On Customer-Oriented Catalog Segmentation Problem

In the era of Internet,with the data growing explosively,extracting valuable information from the huge data becomes a pressing need.To meet this demand,data mining technology emerges and attracts more and more scholars' attention.Under this background,some theories of data mining have been put forward.The microeconomic view of data mining is one of them,which is a rigorous framework based on optimization.Customer-oriented catalog segmentation problem is an important research subject applied in business under this framework,designing appropriate catalogs for the enterprise and sending them to corresponding customers to maximize the number of covered customers.In this thesis,novel catalog segmentation problems and an effective catalog segmentation algorithm are proposed based on the in-depth study of the customer-oriented catalog segmentation problem.In the e-commerce environment,customers' purchase behavior is influenced by many factors.In this thesis,the bi-objective catalog segmentation problem considering freight is given through extending the customer-oriented catalog segmentation problem with profit constraint based on the online consumer psychology and behavior,aiming at maximizing the expected profit of the enterprise and the number of covered customers.In addition,in view of customer value theory,the perception-oriented catalog segmentation problem is presented,improving the click rate and transform rate of products and increasing the long-term benefits of enterprises.In this thesis,an improved co-evolutionary genetic algorithm is put forward,splitting the evolutionary process into two parts: the independent evolution in each sub-population and the co-evolution between sub-populations.The method of optimal strategy preservation and the self-adaptive mutation operator is used to protect the excellent individuals from being destroyed,avoid getting trapped in a local optimum and accelerate the convergence of the algorithm.Furthermore,the co-crossover operator is employed to achieve co-evolution by making the sub-populations exchange information.In comparison with the memetic algorithm and the classical RBPF algorithm based on a real data set,the proposed algorithm performs better in covering customers.A prototype system is also developed based on this co-evolutionary genetic algorithm to visualize the catalog customization.