随着信息化时代的快速发展，用户行为数据也随之增长。这些数据不仅记录了用户的消费习惯和偏好，还包含着提升电商业务价值的关键信息。然而，在实际运用中，许多商家未能有效利用这些宝贵的数据资源，导致他们在市场竞争中变的被动。因此，深入分析用户行为，尤其是重复购买行为，对电商平台的精准营销和提升用户复购率具有极其重要的意义。本文基于天猫平台公开的用户购物数据，进行了深入的数据探索与模型分析。首先，在数据探索与分析方面，深入研究了数据集中各字段的分布情况和用户行为特征，尤其关注了双十一大促活动前后用户行为的变化，并提出了处理缺失值问题的策略。其次，针对传统电商网络分析忽略商家影响的问题，引入了商家社区网络，并通过Leiden社区发现算法对商家社区进行划分，探讨了商家社区特征对用户重复购买行为的影响。最后，在特征工程方面，从用户与商家两个角度构建了新特征，并分析了影响用户复购行为的因素，为模型构建提供了全面的特征数据集。进一步，本文采用逻辑回归等基础模型作为基准，进一步引入了随机森林、XGBoost、LightGBM三种集成学习模型进行复购预测，并对比分析不同模型在复购预测问题上的性能。对比模型结果发现集成学习模型在预测准确性、稳定性和泛化能力方面表现出色。此外，通过特征重要性分析了影响用户复购的关键因素，其中商家社区相关特征的重要性显著，进一步验证了从商家角度分析复购行为的合理性。这些关键因素不仅有助于电商平台更好地理解用户需求，还为精准营销和用户关系管理提供了有力支持。综合而言，本文为电商行业提供了深入洞察和实用价值，为未来的数据驱动决策和用户行为预测提供了重要参考。通过对不同模型的性能比较和关键因素分析，本文为电商平台提供了具体可行的策略和方法，以优化营销策略、提升用户体验，并最终实现业务增长和用户忠诚度的提升。

With the rapid advancement of e-commerce platforms and social networks, user behavior data is growing rapidly. These datasets not only capture users’ consumption patterns and preferences but also hold vital insights that can significantly augment the competitiveness of e-commerce enterprises. Despite their potential, many businesses struggle to harness these invaluable data resources effectively, often relegating themselves to a passive stance in fiercely competitive markets. Consequently, conducting thorough analyses of user behavior, particularly repeat purchase patterns, emerges as a critical imperative for enabling targeted marketing strategies and bolstering customer retention rates on e-commerce platforms.This paper undertakes an in-depth examination of data and model analysis utilizing publicly accessible user shopping data from the T-mall platform. Initially, within the realm of data exploration and analysis, the distribution of various dataset fields and user behavior traits is meticulously investigated. Special emphasis is placed on scrutinizing shifts in user behavior pre- and post-promotion during the Double 11 shopping festival, alongside proposing a methodology for handling missing data. Subsequently, in the realm of social network introduction, recognizing the oversight in traditional e- commerce network analysis regarding the impact of merchant networks, this paper introduces the concept of merchant community networks. Leveraging the Leiden community detection algorithm, the paper partitions merchant communities and explores the influence of merchant community attributes on user repeat purchase behavior. Lastly, concerning feature engineering, novel features are crafted from diverse perspectives. This includes the derivation of user attributes, merchant characteristics, and user-merchant interaction features, informed by an analysis of factors influencing repeat purchase behavior. These efforts culminate in the creation of a comprehensive feature dataset tailored for model development.Furthermore, this paper utilizes logistic regression and other fundamental models as benchmarks and then introduces ensemble learning models such as random forest, XGBoost, and LightGBM for predicting repeat purchases. It compares and analyzes the performance of these different models in predicting repeat purchases. The paper concludes that ensemble learning models exhibit outstanding performance in terms of prediction accuracy, stability, and generalization capabilities. Additionally, through feature importance analysis, the paper reveals key factors influencing user repeat purchases, highlighting the significant importance of merchant community-related features. This finding further validates the rationale behind analyzing repeat purchase behavior from a merchant perspective.These identified key factors not only assist e-commerce platforms in better understanding user needs but also provide robust support for precise marketing and customer relationship management. In conclusion, this paper offers in-depth insights and practical value for the e-commerce industry, providing essential references for future data-driven decision-making and user behavior prediction. Through performance comparisons of different models and analysis of key factors, the paper offers specific and feasible strategies and methods for e-commerce enterprises to optimize marketing strategies, enhance user experience, and ultimately achieve business growth and enhanced customer loyalty.