Unsupervised Customer Segmentation in E-Commerce via PCA-Assisted K-Means Clustering

Dongjin Xie

Abstract

处理后如下：

Background In the competitive ecommerce landscape generalized marketing strategies often fail to engage diverse consumer bases Understanding latent customer behaviors without predefined labels requires robust unsupervised learning techniquesObjective This study aims to partition a high dimensional dataset of online shoppers into distinct actionable cohorts using K Means clustering We further seek to visualize these high dimensional clusters using Principal Component Analysis PCAMethods We applied the K Means algorithm Lloyds heuristic to a dataset of 2000 customers based on Annual Income and Spending Score To determine the optimal number of clusters (k) we utilized the Elbow Method and Silhouette Analysis PCA was employed to reduce dimensionality and verify cluster separabilityResults The analysis identified (k=5) distinct customer segments including Prudent Savers Target Group and Big Spenders The model achieved a Silhouette Score of 0.55 indicating dense and well separated clustersConclusion The combination of K Means and PCA provides a powerful framework for discovering market niches enabling businesses to tailor hyper personalized retention strategies

World Journal of Young and Excellent Scholars

Unsupervised Customer Segmentation in E-Commerce via PCA-Assisted K-Means Clustering

Abstract

Citation

Share