Feature selection approaches for predictive modelling of groundwater nitrate pollution: An evaluation of filters, embedded and wrapper methods

RF with SFFS had the best performance (mmce = 0.12 and AUC = 0.92) and good interpretability, where three features related to groundwater polluted areas were selected: i) industries and facilities rating according to their production capacity and total nitrogen emissions to water within a 3 km buffer, ii) livestock farms rating by manure production within a 5 km buffer and, iii) cumulated NDVI for the post-maximum month, being used as a proxy of vegetation productivity and crop yield.