Zhang Lili, Priestley Jennifer, DeMaio Joseph, Ni Sherry, Tian Xiaoguang
Analytics and Data Science Institute, Kennesaw State University, Kennesaw, Georgia, USA.
Department of Statistics and Analytical Sciences, Kennesaw State University, Kennesaw, Georgia, USA.
Big Data. 2021 Apr;9(2):132-143. doi: 10.1089/big.2020.0044. Epub 2020 Dec 29.
Product affinity segmentation discovers groups of customers with similar purchase preferences for cross-selling opportunities to increase sales and customer loyalty. However, this concept can be challenging to implement efficiently and effectively for actionable strategies. First, the nature of skewed and sparse product-level data in the clustering process results in less meaningful solutions. Second, customer segmentation becomes challenging on massive data sets due to the computational complexity of traditional clustering methods. Third, market basket analysis may suffer from association rules too general to be relevant for important segments. In this article, we propose to partition customers into groups with their product purchase similarity maximized by detecting communities in the customer-product bipartite graph using the Louvain algorithm. Through a case study using data from a large U.S. retailer, we demonstrate that the proposed method generates interpretable clustering results with distinct product purchase patterns. Comprehensive characteristics of customers and products in each cluster can be inferred with statistical significance since they are essentially driven by products purchased by customers. Compared with the conventional RFM (recency, frequency, monetary) model, the proposed approach leads to higher response rates in the recommendation of products to customers in the same cluster. Our analysis provides greater insights into customer purchase behaviors, improves product recommendation effectiveness, and addresses computational complexity in the context of skewed and sparse big data.
产品亲和力细分可发现具有相似购买偏好的客户群体，以创造交叉销售机会，从而提高销售额和客户忠诚度。然而，要将这一概念有效且高效地应用于可操作的策略可能具有挑战性。首先，聚类过程中倾斜且稀疏的产品级数据的性质会导致得出的解决方案意义不大。其次，由于传统聚类方法的计算复杂性，在海量数据集上进行客户细分颇具挑战。第三，购物篮分析可能会遇到关联规则过于笼统，对重要细分群体缺乏相关性的问题。在本文中，我们建议通过使用Louvain算法在客户 - 产品二分图中检测社区，将客户划分为产品购买相似度最大化的群体。通过一项使用美国大型零售商数据的案例研究，我们证明所提出的方法能够生成具有不同产品购买模式的可解释聚类结果。由于每个聚类中的客户和产品的综合特征本质上是由客户购买的产品驱动的，因此可以在统计显著性的基础上进行推断。与传统的RFM（最近一次购买时间、购买频率、购买金额）模型相比，所提出的方法在向同一聚类中的客户推荐产品时能带来更高的响应率。我们的分析为客户购买行为提供了更深入的见解，提高了产品推荐的有效性，并解决了倾斜和稀疏大数据背景下的计算复杂性问题。
