Establishment of Machine Learning Datasets for better healthcare outcomes: A focus on rabies

Introduction: Rabies remains a major public health concern in low- and middle-income countries (LMICs), despite being vaccine-preventable. Limitations in timely diagnosis, surveillance, and outbreak prediction hinder progress toward global elimination targets. Artificial Intelligence (AI) and Machine Learning (ML) offer promising solutions, but their uptake in LMICs has been constrained by the lack of structured, high-quality datasets. This study aimed to establish machine learning-ready datasets to support rabies diagnosis, outbreak prediction, and improved allocation of preventive resources in Tanzania. Methods: Data were sourced from Integrated Bite Case Management (IBCM) and contact tracing systems, including over 15,000 bite patient records and 3,000 confirmed rabies cases. Data preprocessing, annotation, and feature engineering were conducted to create structured datasets suitable for ML applications. Several baseline models were developed and evaluated using ROC curve analysis. Results: Gradient Boosting achieved the highest diagnostic performance with an Area Under the Curve (AUC) of 0.82, followed by Logistic Regression at 0.81. The processed datasets also revealed clear spatiotemporal trends in rabies outbreaks, highlighting their potential for predictive analytics and strategic planning. Conclusion: This work provides a foundational dataset for AI-driven rabies surveillance and control in Tanzania. Public access to the dataset is expected to enhance research, support data-driven health policy, and contribute to achieving the WHO's goal of eliminating dog-mediated human rabies deaths by 2030.