Reproduce “Multi-task CNN Behavioral Embedding Model For Transaction Fraud Detection” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Bo Qu, Zhurong Wang, Minghao Gu, Daisuke Yagi, Yang Zhao, Yinan Shan, Frank Zahradnik
Researchers at eBay developed a Multi-task CNN Behavioral Embedding Model designed for transaction fraud detection, which analyzes sequential user behavior with multi-range convolutional kernels and positional encoding. This model achieves competitive fraud detection performance against Transformer-based approaches while offering improved computational efficiency for large-scale e-commerce platforms.
Key insights
- User behavior sequences on e-commerce platforms contain distinctive patterns that can effectively signal fraudulent intent.
- Convolutional Neural Networks (CNNs) can efficiently capture complex sequential patterns in user behavior data, including local patterns of varying lengths, offering a scalable alternative to RNNs and Transformers.
- A multi-task learning framework with dynamic loss weighting improves the model's generalization capabilities by simultaneously predicting various fraud-related tasks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.