Tuberculosis (TB) continues to be a critical public health crisis in the Eastern Cape of SouthAfrica, a province characterised by high HIV co-infection rates, persistent poverty, andchronically under-resourced district health systems. Despite available epidemiological data,TB prevention resources are overwhelmingly allocated reactively — directed to areas only afterscreening has identified cases — resulting in preventable transmission and inefficient use ofscarce public health capacity. This study proposes a machine learning-based predictiveanalytics framework designed to enable proactive, pre-screening resource allocation for TBprevention across Eastern Cape health districts. Using anonymised district-level data from theSouth African DHIS2 platform and supplementary community-led monitoring data fromRitshidze (ritshidze.org.za), this study will train and evaluate three supervised classifiers —logistic regression, random forest, and XGBoost — to predict which Eastern Cape districts willrequire intensified TB prevention resources before case screening occurs. SHAP analysis willensure model interpretability for non-specialist public health stakeholders. The study runs fromJuly to September 2026 and is expected to produce district-ranked resource-demand profilesdirectly applicable to Eastern Cape health department planning cycles.