[1J3-GS-10d] AI application

2:25 PM - 2:40 PM JST(5:25 AM - 5:40 AM UTC)

[1J3-GS-10d-04] Customer Segmentation Using K-means Hierarchical Cluster Merging

〇Yuji Sato¹, Gen Sakaeda¹, Junji Takano², Daito Yamauchi¹, Atsuyoshi Muta¹ (1. ASKUL Corporation, 2. Accenture Japan Ltd)

Keywords:

EC,Customer Segmentation,K-means Clustering,Hierarchical Clustering,Recommend

In e-commerce, understanding customer demand is essential for marketing strategies, including recommendation systems. Therefore, it is important to develop customer segments by grouping customers based on their purchasing tendencies. To create customer segments, it is necessary to group customers based on their purchasing behavior. However, labels representing customer demand types are rarely directly available. Consequently, clustering approaches based on similarities in purchasing behavior are widely used, such as K-means clustering and hierarchical clustering. However, K-means clustering requires the number of clusters to be specified in advance, and hierarchical clustering faces high computational complexity, particularly when dealing with large-scale data. To overcome the limitations of both methods and to develop customer segments from high-dimensional and large-scale purchasing data, we adopt a hybrid K-means hierarchical clustering approach that hierarchically merges representative vectors constructed by K-means clustering. We confirmed that the proposed method can be executed within a practical time frame, even on data consisting of over one million samples with 400 dimensions, and that flexible merging and splitting of customer segments according to specific marketing objectives can be achieved using the representative vectors and a dendrogram.