[PROJECT NAME] Big Data Intelligent Risk Targeting System
[PROJECT OVERVIEW] Big data and artificial intelligence are key drivers of the new technological revolution, enabling industries to evolve from digital transformation to AI-powered governance. Through the development and application of the “Tianxuan” Intelligent Targeting Model,China Customs has effectively integrated decades of expert experience with AI technology. The system continuously learns policy updates, risk patterns, and import-export behaviors to continuously self-improve, achieving rapid risk identification and precise targeting while significantly expanding risk control coverage and accuracy.
[PROJECT DETAILS]
I. Key Approaches
1. Enhanced Data Governance. China Customs established a “big data pool”, integrating multiple sources of internal and external data. So far, it has compiled 17,436 internal/external data tables and approximately 270.25 billion data records. To improve data quality and facilitate sharing and reuse, we standardized 9 major categories of customs data—including declarations, manifests, enterprises, logistics, inspections, anti-smuggling cases, audits, fund flows, and insurance—across more than a hundred internal and external customs data sheets to provide complete, accurate, and agile support for AI modeling.
2. Quantified Risk Prioritization. China Customs has collaborated with Peking University and Huawei’s algorithm teams and deployed advanced AI algorithms such as CATBOOST and neural network models to automatically identify risk features in customs declarations and rank their impact. This approach enabled us to conduct real-time, scientific, and rapid (within seconds) quantitative assessments of the risk probability and impact of customs declarations, thereby improving our risk assessment capabilities.
3. Standardization Framework Development. China Customs summarized practical experience and established clear standards for project initiation, R&D, operations, and performance evaluation, in order to provide replicable and promotable experience for building the model library.
II. Achievements
1. More Comprehensive Analysis. China Customs has integrated 13 business areas, including import and export restrictions, taxation, and quarantine, and translated expert knowledge into 132 machine-readable risk indicators for full-spectrum risk screening.
2. More Precise Targeting. China Customs draws on WCO risk management principles and takes into account both the risk level and supervision costs of customs declarations, ranking risks to focus enforcement on higher-risk targets. In 2024, the model’s seizure rate reached 23.7%—2.87 times the average.
3. More Efficient Prevention & Control. After comprehensively identifying risks and completing quantitative risk assessments, any customs declarations exceeding the set threshold were automatically routed for enhanced inspection, significantly improving the timeliness of our risk response. Each customs declaration underwent a "check-up" in just 3 seconds, enabling us to achieve 100% real-time risk monitoring of customs declarations nationwide across land, sea, and air.
4. More Consistent Law Enforcement. China Customs upholds fair and consistent law enforcement as a core value. By implementing a uniform algorithm-based risk evaluation of shipments, China Customs eliminates subjective judgment variances among experts and ensures fairness and consistency in customs controls.
III. Promotion and Application
The project was piloted in Huangpu Customs District and subsequently promoted nationwide. Through 38 iterations and over 600 parameter optimizations, model accuracy and response speed improved significantly. In 2024, targeting volume grew 1.5-fold year-on-year while seizure rates rose by 3.9 percentage points—delivering dual gains in scale and quality.
The project was included in the WCO Risk Management Compendium and presented at WCO multilateral meetings in Jakarta and Brussels, followed by a dedicated study tour by WCO Smart Customs experts, marking a pioneering step toward global AI application standards in customs.