Schistosomiasis Prevention Undermined by Water Point Forecasting: An Approach Based on Data Fusion Model

---

Résumé

In this study, we present an approach called FuMalMMo for the Fusion of Machine learning and Mathematical Models. It is an approach based on data fusion which leverages information coming from different data sources to make a decision. The approach we present follows a “Y” pattern where in the left branch there is a machine learning model in charge of forecasting the water quality of a water point. In the right branch, there is an epidemiological model responsible for making a forecast of the evolution of the density of parasites and snails causing schistosomiasis. In the middle branch, we rely on the theory of belief functions or evidence theory to combine the forecasts made by the two models in order to infer one day ahead the state of infestation of a water point with an accuracy of 0.75.

Mots-clés

schistosomiasis prevention, data fusion, water quality prediction, epidemic modeling, belief functions, Dempster-Shafer theory, day-ahead forecasting