Bovine Tuberculosis (bTB) remains one of the most persistent and costly livestock diseases in Ireland, threatening animal health, farm livelihoods, and national biosecurity. In 2024, 5,906 herds experienced a TB breakdown, leading to the removal of 37,964 reactor animals, with significant impacts on agricultural exports—€6.3 billion from dairy and €2.8 billion from beef. This study leverages a comprehensive national dataset spanning 2019–2023, comprising over 84 million records, including individual bTB tests, cattle movement logs, herd demographics, and herd status information. Using this dataset, we applied machine learning enhanced with graph-based network features from cattle movements to predict herd-level bTB breakdowns one year in advance. Traditional herd characteristics were combined with network metrics—including degree, PageRank, closeness, betweenness, and Louvain clustering—and evaluated under three data augmentation strategies: Random Under-Sampling (RUS), Random Over-Sampling, and Synthetic Minority Oversampling. The best-performing model (Random Forest with RUS) achieved a sensitivity of 0.83, specificity of 0.85, and ROC–AUC of 0.89. Feature importance analysis showed conventional risk factors dominated, with PageRank ranking fifth among all features. Despite a moderate precision of 0.23, these results highlight the potential of integrating network analytics with machine learning to enhance targeted bTB surveillance and control strategies.