This book presents a practical, end-to-end approach to risk management in modern e-commerce platforms using machine learning. It covers buyer abuse, seller governance, collusion networks, and payment and credit risk, showing how behavioural, transactional, and network data are translated into actionable risk signals. The book discusses a range of modelling techniquesincluding tree-based ensembles, sequence and transformer models, graph neural networks, and anomaly detectionfocusing on their application under real-world constraints such as class imbalance, latency, adversarial drift, fairness, and regulatory requirements. Extending beyond model development, it addresses deployment, monitoring, governance, and operational assurance, making it relevant to industry practitioners and suitable for advanced courses in applied, industry-focused machine learning.

Comprehensive guide that unites payment fraud, buyer-side exploits, seller governance, and collusion detection Leverages the latest advancements in transformer-based sequence models and graph neural networks for patterns of misuse Designed as both a hands-on handbook for data scientists in e-commerce and a practical course text for AI/ML students

Autorentext

Dr Simon Liu is a senior leader in data science and risk management, with experience spanning academia, financial services, and large-scale e-commerce plat forms. He is currently Chief Data and AI Officer at TrustDecision, where he leads the company’s data and artificial intelligence strategy in decision intelligence and risk analytics. Dr Liu is also an Adjunct Associate Professor at Nanyang Technological University (NTU), teaching graduate-level courses in natural language processing and ensemble learning, and is coauthor of the textbook Analytic Learning Methods for Pattern Recognition. Previously, he served as Senior Vice President and Head of Data Science for Risk and Security at Lazada and earlier held leadership roles at Scotiabank, where he was Director of AML Modelling and Analytics and led the development of Canada’s first AI-powered human trafficking detection model. He is a strong advocate of academic–industry collaboration, focusing on translating advanced machine learning research into practical, production-grade risk systems.

Zusammenfassung

This book offers a comprehensive guide to leveraging machine learning and predictive analytics for managing the full spectrum of risks inherent in modern e-commerce platformsfrom payment fraud and chargebacks to platform-specific abuses like voucher manipulation, collusion rings, and promotional gaming. Grounded in real-world practice, it shows how to transform raw transaction logs, user clickstreams, and promotional datasets into actionable risk scores that power both real-time interdiction and strategic oversight. Readers discover a suite of methodsranging from gradient-boosted trees and deep sequence models to graph neural networks and unsupervised anomaly detectorseach chosen for its ability to detect subtle, evolving patterns of misuse.

Inhalt

The E-Commerce Risk Landscape.- Data Foundations for E-Commerce.- Tree-based Methods.- Deep Learning Models.- Unsupervised & Anomaly Detection Methods.- Buyer Journey Models.- BuyerSeller Collusion Models.