Explainable AI Research

About this project

BSc Dissertation at University of Surrey. Developed a method to extract interpretable decision trees from trained neural networks, making black-box models explainable. Relevant to XAI, model interpretability, and EU AI Act regulatory compliance.

Key Features

• Reduced decision tree complexity by up to 70%
• Maintained accuracy within a few percentage points of original neural networks
• Tested on image classification and banking fraud detection datasets
• First Class Honours dissertation