Customer Segmentation: Data-Driven Research on Transaction-Level Data of JD.com

As a premier e-commerce platform in China, JD.com offers significant research potential through its extensive user data, clicks, and transaction records. This study utilizes the JD transaction-level dataset from the 2020 MSOM Data Driven Research Challenge to predict consumer preferences and extract data-driven insights into customer behavior. Following the initial adoption of the k-means clustering methodology, a manual classification-based approach was subsequently proposed to mitigate the issue of boundary blurring. Various machine learning models, including Logistic Regression, Decision Tree, Random Forests, and Stacking, were subsequently employed to enhance the predictive analysis of users’ Stock Keeping Unit (SKU) preferences. The implemented model achieved an accuracy rate of 74%, thereby advancing the comprehension of JD's heterogeneous customer base. This study presents a novel perspective for optimizing user experience on e-commerce platforms and contributes to the sustainable development of the industry.