Lassa fever remains a persistent endemic disease in Nigeria, characterized by recurring seasonal outbreaks and spatial clustering across states. Despite improvements in surveillance systems, outbreak monitoring remains largely descriptive, limiting proactive response strategies. This study develops a spatio-temporal predictive framework using Nigeria Centre for Disease Control (NCDC) surveillance data from 2020 to 2025. The study focuses on two objectives: examining spatio-temporal outbreak patterns and developing predictive machine learning models for outbreak forecasting. A retrospective quantitative modelling approach was adopted, integrating feature engineering techniques such as lag variables and rolling averages. Regression models (Random Forest and Gradient Boosting) were used to predict case counts, while classification models (Logistic Regression, Random Forest, and XGBoost) were used to predict outbreak occurrence. Results revealed strong temporal patterns with recurrent seasonal peaks. Gradient Boosting achieved the best regression performance (R² = 0.96), while Logistic Regression demonstrated the highest classification accuracy (1.00). The findings indicate that machine learning models can effectively capture outbreak dynamics and provide reliable early warning signals. The study concludes that integrating predictive analytics into disease surveillance systems can significantly enhance outbreak preparedness and resource allocation efficiency in Nigeria.