Radius fractures are one of the most common types of fractures, and better prediction tools could help with early prevention. In this project, I used data from the UK Biobank, which includes about 500,000 participants. I focused on 8,653 people with radius fractures and 29,430 without fractures. Predictors included demographic, lifestyle, and genetic factors, including a polygenic risk score for bone mineral density. I excluded age to avoid bias, since fracture cases were generally older. I used PyCaret AutoML to compare several machine learning models and chose AdaBoost, which had the best cross-validation performance (AUC ≈ 0.96, Accuracy ≈ 94%). When I tested the model on a separate hold-out set, performance dropped (AUC = 0.68, Accuracy = 0.50). The model caught almost all fractures (recall ≈ 99%) but also predicted many fractures that didn’t happen (precision ≈ 50%). These results show that while the model is good at finding fracture cases, it over-predicts, which may be due to weak predictors or overfitting. In the future, I plan to add stronger predictors like bone density scans and test the model on other datasets to improve generalization.