O2O社区营销中客户细分与社区发现算法研究

发布时间：2018-07-22 20:29

【摘要】：020(Online to Offline)指融合了线上和线下的商业服务模式,是近年来互联网衍生出现的新趋势。而社区化是目前020发展的一种新的表现形式,它主要指电商企业通过整合线上线下资源,为社区内的居民提供生活综合类商品及服务,形成商家、中间服务提供者和最终用户的统一平台。与传统电商不同,020社区更强调线下产品和服务,因而具有明显的本地化地域属性。目前020在展开营销活动时,常常先以价格优势吸引线下流量,初期盲目追求新用户的数量,但是在后续的客户关系管理和维系上并没有很好的解决方案。在本文中,我们重点解决两个具体问题：在020营销中,如何对消费能力不同的客户进行划分；以及如何以更有效的策略发现020中的线下社区。互联网的普及使得交易量持续增长,电子商务网站积累了大规模的交易数据。这些数据中有一个显著的特征是每笔交易数据都有一个客户的物流地址。本文利用开放的互联网房产数据,将客户的物流地址与房产数据中的地址进行匹配,通过客户所居住的房屋的价值来对其收入水平做出预判,进而推测客户的价值。对于社区发现问题,我们提出了一种基于LDA(Latent Dirichlet Allocation)主题模型并融合三大社区影响要素的020社区发现算法。这三大要素分别为：楼盘自身所处的地理位置特征、周边对等楼盘构成的影响力及所属商区的潜在商业价值。然后将社区与主题进行类比,使用LDA模型来发现社区,最终得到各楼盘归属于所在社区的概率分布。我们结合上海市开放房价数据和某电商交易数据完成了客户细分实验,为评估算法有效性,我们将实验结果与RFM(Recency,Frequency,Monetary)模型的结果进行比对,通过相关性分析,验证了我们的算法可以有效地对客户进行细分。对于社区发现,我们使用了上海市出租车GPS轨迹数据和上海市分类别的POI数据,实现了我们的算法,证明了社区发现算法的有效性。并且,针对由算法得到的社区发现结果,我们具体分析解释了它在020社区运用中的实际意义。
[Abstract]:020 (online to offline) refers to the integration of online and offline business service models, which is a new trend of Internet derivation in recent years. Community is a new manifestation of the development of 020. It mainly refers to e-commerce enterprises providing comprehensive goods and services for the residents of the community through the integration of online and offline resources, so as to form businesses. A unified platform for intermediate service providers and end users. Different from traditional ecommerce, the 020 community emphasizes offline products and services. At present, 020 often uses price advantage to attract offline flow first and blindly pursue the number of new users at the beginning, but there is no good solution to the subsequent customer relationship management and maintenance. In this paper, we focus on solving two specific problems: how to divide customers with different consumption power in 020 marketing, and how to find offline communities in 020 with more effective strategies. The popularity of the Internet makes the volume of transactions continue to grow, e-commerce sites have accumulated large-scale transaction data. A notable feature of these data is that each transaction has a customer's logistics address. This paper makes use of the open Internet real estate data, matches the customer's logistics address with the address in the real estate data, through the value of the house that the customer lives in, makes the forecast to its income level, and then speculates the customer's value. For the community discovery problem, we propose a community discovery algorithm based on the LDA (Latent Dirichlet Allocation) topic model and integrating the three community impact elements. These three elements are: the geographical location characteristics of the building itself, the influence of the surrounding equivalent property and the potential commercial value of the business district. Then we use LDA model to find the community and get the probability distribution of each building belonging to the community. In order to evaluate the validity of the algorithm, we compared the experimental results with the results of RFM (Reclamation Frequency Monetary) model, and analyzed the correlation between the results of the experiment and the results of RFM (Reclamation Frequency Monetary) model. It is verified that our algorithm can effectively subdivide the customer. For community discovery, we use Shanghai taxi GPS track data and Shanghai sub-class POI data to realize our algorithm, which proves the effectiveness of community discovery algorithm. Furthermore, in view of the community discovery results obtained from the algorithm, we analyze and explain its practical significance in the application of 020 community.
【学位授予单位】：华东师范大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：F724.6;TP301.6

【相似文献】