Intelligent Risk Control Forum – Sessions on Graph Algorithms, Pre‑trained GNN, Loop Detection, Active Learning, and Unstructured Data

The model‑algorithm layer is the core of intelligent risk control, enabling probabilistic prediction of risk events.

To build effective models, a variety of algorithms are available, yet practitioners face data sparsity, complex structures, adversarial scenarios, fast‑changing tactics, interpretability and robustness issues, especially with organized fraud.

Speaker: Wang Haoran – Senior Risk Control and Graph Computing Expert British‑educated algorithm specialist with experience at Ant Finance, Alibaba, Tencent, focusing on risk control, social and graph computing across finance, e‑commerce and supply‑chain domains.

Speaker: Han Jifei – Senior Technical Expert, Fabarta Graph Intelligence Tsinghua University graduate, former Huawei and Kuaishou multimodal algorithm developer, now designing complex graph algorithms and graph machine learning.

Talk Title: Loop Detection Algorithms – Optimization Directions for Risk Control Outline: Using loop detection to prevent fraud and illegal transactions, challenges of efficient detection in large graph networks, resource‑aware processing of massive data.

Audience Benefits: How to quickly filter loops, reduce memory consumption in distributed settings, and apply loop detection in financial risk scenarios.

Speaker: Xu Jiarong – Associate Researcher, Fudan University Researches graph data mining and privacy, published in KDD, AAAI, NeurIPS, IJCAI, TKDE, TKDD, and serves as reviewer for major conferences.

Talk Title: Pre‑trained Graph Neural Networks – Data‑Centric Reflections Outline: Motivation for graph pre‑training, handling label sparsity in finance, negative transfer issues, applicability scope, feasibility metrics, and data selection strategies.

Audience Benefits: Understanding pre‑trained GNN concepts, when to apply them, and how data volume impacts performance.

Speaker: Zhou Min – Senior Researcher, Huawei Cloud Algorithm Innovation Lab USTC and NUS graduate, focuses on machine learning and representation learning for sequential and graph data, holds multiple patents and publishes in ICML, KDD, SIGIR, WWW, etc.

Talk Title: Active Learning and Sample Imbalance in Graph Data Outline: Imbalance between normal and fraudulent transactions, labeling challenges, and techniques for graph‑based active learning and imbalance mitigation.

Audience Benefits: Graph active‑learning methods and solutions for sample imbalance.

Speaker: Zeng Libin – Security Algorithm Expert, Ant Group Background in risk management at China UnionPay, now responsible for anti‑theft and international account risk algorithms at Ant Group.

Talk Title: Unstructured Data Intelligent Risk Control Outline: Business background of Ant International risk control, challenges of unstructured data, algorithmic solutions, and intelligent risk control architecture.

Audience Benefits: Extracting accurate information from multimodal unstructured data, ensuring authenticity, and designing intelligent risk control for accounts and transactions.

Speaker: Wang Peng – Algorithm Researcher, Tencent Shanghai Jiao‑Tong University and University of Lyon alumnus, focuses on machine learning, deep learning, content understanding and content risk control.

Talk Title: Exploration and Practice of Content Risk Control Adversarial Systems Outline: Background, problem analysis, solutions (adversarial perception, model automation, model fusion, intelligent decision), and effectiveness.

Audience Benefits: Overview of adversarial systems, challenges in model automation, and countermeasures for text‑based attacks.