The increasing frequency of zoonotic disease events underscores a need to develop forecasting tools toward a more preemptive approach to outbreak investigation. We apply machine learning to data describing the traits and zoonotic pathogen diversity of the most speciose group of mammals, the rodents, which also comprise a disproportionate number of zoonotic disease reservoirs. Our models predict reservoir status in this group with over 90% accuracy, identifying species with high probabilities of harboring undiscovered zoonotic pathogens based on trait profiles that may serve as rules of thumb to distinguish reservoirs from nonreservoir species. Key predictors of zoonotic reservoirs include biogeographical properties, such as range size, as well as intrinsic host traits associated with lifetime reproductive output. Predicted hotspots of novel rodent reservoir diversity occur in the Middle East and Central Asia and the Midwestern United States.machine learning | disease forecasting | prediction | pace-of-life hypothesis | generalized boosted regression trees I nfectious agents transmitted from animals to humans account for most outbreaks of novel pathogens worldwide (1-3). With over 1 billion cases of human illness attributable to zoonotic disease each year, identifying wild reservoirs of zoonotic pathogens is a perennial public health priority (4). Until now, investigations of disease outbreaks have mostly been reactive, with surveillance efforts targeting a broad host range (5), but because human activities precipitating these events continue to accelerate (4, 6), a more proactive approach is necessary (7,8). Identifying which wildlife species are most likely to serve as reservoirs of future zoonotic diseases and in which regions new outbreaks are most likely to occur are necessary steps toward a preemptive approach to minimizing zoonotic disease risk in humans. To this end, trait profiles inferred from large datasets that distinguish reservoirs from nonreservoir species can play a major role in guiding the search for novel disease reservoirs in the wild. Identifying these distinguishing, intrinsic features of zoonotic reservoirs also has the potential to generate testable hypotheses that can explain why some host species are more permissive to zoonotic infections.To accomplish these goals, we applied generalized boosted regressions (9, 10), a type of machine learning that builds ensembles of classification/regression trees to identify variables that are most important for prediction-in our case, predicting zoonotic reservoir status and hyperreservoir status (species known to carry two or more zoonotic infections). These methods and similar methods have particular use for comparative ecological studies because they accommodate multiple data types as covariates, nonrandom patterns of data missingness, and hidden, nonlinear interactions. The explanatory power of decision tree methods is unaffected by variations in data coverage that may arise because of sampling bias or when species share a particular trait because...