Quantitative Analysis of Price Elasticity in E-commerce Using Machine Learning: A Case Study of JD.com Transaction-Level Data

With the rapid development of e-commerce, it is increasingly important to understand consumer decisions and make forward-looking product management strategies according to those behaviors. Based on the 2020 MSOM Data-driven Research Challenge, this study uses JD.com's transaction-level product data in March 2018 to compare and explore the differences in the effectiveness of multiple machine learning models in predicting product price elasticity, as well as identifying the core factors that might affect. The models include Linear regression, Ridge regression, Decision Tree, and Random Forest. Meanwhile, the parameter tuning strategy of Random Search is also adopted to optimize the model performance and generalization ability. The final results show that the Ridge regression model shows excellent prediction accuracy and stability, with the lowest Mean Square Error of 0.2685 on the test set. This indicates that the model effectively solves the multicollinearity problem between features and avoids the overfitting problem. In addition, the research identifies the decisive factors affecting commodity price elasticity as the customer's purchase channel, product's functional attributes, and the type of merchant to which the commodity belongs. The conclusions drawn in this study are robust and pragmatic, providing valuable insights for merchants to formulate reasonable pricing strategies to promote consumption.