E-commerce Customer Segmentation via Unsupervised Machine Learning

Abstract

Customer segmentation through data mining could help companies conduct customer-oriented marketing and build differentiated strategies targeted at diverse customers. However, there has not been a guideline for systematic implementation of customer segmentation given the raw transaction data. This study focuses on a real-world database from an online transaction platform with the purpose to develop a guideline for customer segmentation for the business. Since the raw data are unlabeled, unsupervised machine learning methods are utilized. This study firstly employs the RFM model to create behavioral features; next, the TF-IDF method is applied to the product descriptions to generate product categories; then, K-means clustering algorithm is used to group customers. After customers are grouped, association rules mining by Apriori Algorithm is used to analyze purchased products. Principle Component Analysis (PCA) and T-Distributed Stochastic Neighbor Embedding (T-sne) methods are utilized to reduce the dimension of data in order to create visualizations. Finally, some concrete recommendations for the business based on the results are provided accordingly.