A Review of Learning-Based E-commerce

Now e-commerce applications based on learning are continuously improving our daily life. These learning methods include graph neural networks, deep learning, machine learning, and reinforcement learning. More and more researchers use these methods to solve many problems in ecommerce, such as shopping, fraud detection, and product recommendation. To help researchers gain progress in the field, we survey the state-of-art work on the topic in this paper. Specifically, we focus on learning-based methods in typical ecommerce tasks: customer service, fraud detection and antifraud, cross-border transactions, and commerce forecast. We brief these methods, compare their pros and cons, and raise some problems for future research.