Performance Evaluation of Machine Learning Models for Lassa Fever Prediction

Lassa fever is a severe viral hemorrhagic illness endemic in parts of West Africa, primarily transmitted through contact with infected rodent excreta. Early detection and accurate diagnosis are critical to reducing mortality and controlling outbreaks. In recent years, machine learning (ML) has shown great potential in enhancing disease prediction and diagnostic accuracy. This study evaluates the performance of various ML models, including Logistic Regression, K-Nearest Neighbors, Support Vector Machine, and Naïve Bayes, in predicting Lassa fever infection. The models were trained and tested on a dataset comprising clinical and demographic features of patients. Key evaluation metrics such as accuracy, precision, recall (sensitivity), F1-score, macroaverage, and weighted-average were employed to assess model performance. The Support Vector Machine (SVM) model outperformed others with an accuracy of 90%, precision of 91%, recall of 96%, and an F1-score of 93%. The findings underscore the effectiveness of SVM in developing a diagnostic model for Lassa fever, providing a foundation for deploying AI-driven diagnostic tools in resource-limited settings. Future research should explore integrating more diverse datasets and incorporating additional clinical parameters to enhance prediction accuracy further. .