Data-driven Modeling as a Tool for Prediction of Future Outbreaks of Ebola Virus Disease in West Africa

Ebola is among the deadliest human diseases and some measures to combat it have been developed, but they are still not able to fully minimize its far-reaching effects. However, the integration of data-driven modeling into existing Ebola prevention and control measures offers hope toward combating this viral disease. Decision trees (DT), gradient boosting (GB), deep neural networks (DNN), and a hybrid model were used to predict Ebola cases and deaths in Guinea, Liberia, and Sierra Leone from 2017 to 2026. These models were evaluated using mean absolute scaled error (MASE), root mean squared error (RMSE), and coefficient of determination (R²). The uncertainties in the prediction of the models were also quantified using the bootstrapping method. Of all the models, the DT model had higher feature importance scores for cases in two countries. The DT model also had a higher accuracy and better performance by RMSE and R ² in predicting Ebola cases and deaths than other models. Predictions of Ebola cases and deaths by the DNN model increased from 2017 to 2026 with higher cases and deaths in 2026 while predictions of Ebola deaths by the hybrid model increased from 2017 to 2026 with higher deaths in 2026. The projected Ebola cases and deaths were higher in Sierra Leone than in other countries. These findings portray the likely number of cases and deaths in case an Ebola outbreak in the three mentioned countries. Furthermore, they show their significance in predicting Ebola virus disease and also have the possibility to help decision-makers in designing effective decisions for the early detection of Ebola incidents. The results of this study show that the DT and DNN models perform better than the other models on the collected Ebola virus disease dataset in the three countries. Therefore, the integration of data-driven infectious disease modeling approaches such as DT and DNN with intervention scenarios such as vaccination can altogether help to reduce the predicted number of Ebola cases and deaths in Guinea, Sierra Leone, and Liberia in the face of an outbreak.