Comparing Knowledge-Driven and Data-Driven Modeling methods for susceptibility mapping in spatial epidemiology: a case study in Visceral Leishmaniasis

Abstract

The aim of this study is to compare knowledge-driven and data-driven methods for susceptibility mapping in spatial epidemiology. Our comparison focuses on one of the arguably most important requisites in such models, namely predictability. We compare one data-driven modelling method called Radial Basis Functional Link Net (RBFLN - a well-established Neural Network method) with two knowledge-driven modelling methods, Fuzzy AHP_OWA and Fuzzy GIS-based group decision making (multi criteria decision making methods). These methods are compared in the context of a concrete case study, namely the environmental modelling of Visceral Leishmaniasis (VL) for predictive mapping of risky areas. Our results show that, at least in this particular application, RBFLN model offers the best predictive accuracy.