Credit card fraud detection using a hierarchical behavior-knowledge space model.

Area of Science:

• Computer Science
• Financial Technology
• Data Science

Background:

• Credit card fraud results in billions of dollars in annual losses for merchants globally.
• Existing fraud detection methods face challenges with noisy data and evolving fraud tactics.
• Advancements in machine learning offer new opportunities for developing intelligent fraud detection systems.

Purpose of the Study:

• To design and evaluate a novel multi-classifier framework for enhanced credit card fraud detection.
• To leverage the Behavior-Knowledge Space (BKS) for combining predictions from diverse machine learning classifiers.
• To demonstrate the model's effectiveness using both public and real-world financial datasets.

Main Methods:

• Development of an ensemble model integrating multiple machine learning classification algorithms.
• Implementation of the Behavior-Knowledge Space (BKS) to aggregate classifier outputs.
• Performance evaluation using statistical tests on publicly available and real financial datasets.

Main Results:

• The developed ensemble model demonstrated significant effectiveness in credit card fraud detection.
• The BKS-based combination method proved superior to the majority voting method.
• The model successfully tackled challenges associated with noisy data classification.

Conclusions:

• The proposed multi-classifier ensemble model with BKS offers a robust solution for credit card fraud detection.
• This approach provides a valuable advancement over conventional prediction combination techniques.
• The findings support the application of intelligent methods for securing financial transactions.