Algorithms For Customer Segmentation And Community Discovery In O2O Marketing

O2O, namely Online to Offline, which is an integrated service model of online and offline business, has become a new trend of Internet derivatives in recent years. Also, community is a new representation in 020 development, which refers to that business enterprise provide products or services for community residents through the integration of online and offline resources. Different from the traditional e-commerce suppliers, the 020 community places more emphasis on offline products and services, so it has an obvious localized feature. Current 020 marketing activities often use price advantage to attract offline users, blindly pursuing large number of new users, but in lack of customer relationship management and maintain.In this article, we focus on two specific issues:how to divide customers with different purchasing abilities in 020 marketing; And how to develop effective strategies to discover the communities in 020. The popularity of the Internet results in accumulated transaction data. One significant feature of the data is that every transaction is associated with a logistics address. In this paper, we predict customer income levels by matching open data from the Internet with the customer’s logistics address, so as to infer the purchasing ability of the customer. For community discovery, we put forward a algorithm based on LDA (Latent Dirichlet Allocation) topic model with fusion of three community influential factors. They are the characteristics of its own neighborhood(individual), the values of its nearby estates (peer), and the prosperity of the affiliated latent business area (zone). We draw an analogy between community and topic, and use the LDA model to discover community.We jointly use open house data of Shanghai and e-commerce transaction data for customer segmentation experiment. To validate, we compare the result with the classic RFM (Recency, Frequency, Monetary) model through correlation analysis. For community discovery, we implement our algorithm and complete real case analysis.