电子商务在互联网普及和技术创新的推动下迅速发展，伴随而来的是海量的数据生成。数据挖掘技术被广泛运用于用户数据和商品数据的分析，以提升用户的交易水平并打造更完善的电子商务生态系统。然而，海量的用户信息和用户行为的动态性给用户细分和其行为刻画带来了更高的要求。本文选取TFA模型作为用户分层方法，并使用基于密度的K-means聚类算法和经过预剪枝的EClats关联挖掘算法作为用户基本信息和行为数据的分析方法，可以帮助商家在用户信息与行为数据挖掘的关键流程里实现动态信息的有效捕捉，提高挖掘速度和内存效率。结果发现，价格敏感型人群更倾向于不同商品之间的关联购物从而成交更多的子单量，而高消费时间少的人群更倾向于商品的独立购买，不太容易受关联商品的影响。商家可以结合用户分层和关联规则对特定人群进行商品有效推荐，以增加商品的成单量和成交金额。

Driven by the popularity of the Internet and technological innovation, e-commerce has developed rapidly, accompanied by massive data generation. Data mining is widely used in the analysis of user and commodity data to achieve personalized recommendation, improve user's transaction level and build a more complete e-commerce ecosystem. However, the massive user information and dynamic user behavior has brought higher requirements for user segmentation and behavior characterization. In this paper, the TFA model is selected as the user stratification method, and the density-based K-means algorithm and the pre-pruned EClats algorithm are used as the analysis method of users' basic information and behavior data, which can help merchants realize the effective capture of dynamic information in the key process of user information and behavior data mining and improve efficiency. The results show that price-sensitive people are more inclined to buy related goods among different products and thus deal with more sub-orders, while people with high consumption level and less time are more inclined to buy goods independently and are less susceptible to the influence of related goods. Merchants can effectively recommend products to specific groups by combining user stratification and association rules to increase the number of orders and transaction amount of goods.