Project background

With the rapid development of financial technology, traditional banks are facing increasingly complex fraud methods and risk challenges. The original rule engine risk control system of a large state-owned bank could no longer meet the real-time and accuracy requirements, and it urgently needed to introduce AI technology for intelligent upgrades.

Solution

Zhiyuan Technology has customized an intelligent risk control platform based on deep learning for the bank. The core modules include:

• Real-time transaction monitoring engine: Based on streaming computing architecture, millisecond response to each transaction
• Multi-dimensional feature engineering: Fusion 200+ dimensional features such as user behavior, device fingerprints, geographical location, etc.
• Deep learning model: Using a sequence model of Transformer architecture to capture the timing pattern of transaction behavior
• Interpretability module: Provides traceable explanation reports for each risk decision