Australian Border Force: AI System Evaluation and Recommendations

This project evaluates two commercial AI solutions for the Australian Border Force's next-generation border control system, which utilizes iris recognition for identity verification. The assignment involves calculating conditional probabilities using Bayes' rule to assess the risk of imposters. It also includes recommendations based on the analysis, addressing uncertainty in the context of minimizing false positives. Furthermore, the project analyzes data visualization techniques, critiques existing graphics, and redesigns them to better represent data. It also assesses the performance of different AI security systems, recommending the optimal choice for various airport environments using ROC curves and machine learning methods. Finally, the project investigates the feasibility of activity recognition using wearable sensors, analyzing acceleration data to classify activities and assess the generalizability of machine learning models across subjects. The project uses MATLAB code and visualization to support the analysis and recommendations.