Fig. 5From: Microbial indicators of environmental perturbations in coral reef ecosystemsRandom forest machine learning. a The 30 most important zOTUs reducing the uncertainty in the prediction of seawater temperature classes (low, average, high) based on their mean decrease in accuracy and b their enrichment in the temperature classes. c The 30 most important zOTUs reducing the variance (mean squared error (% Inc. MSE)) in regression-based prediction of seawater temperatures. d Predicted seawater temperature values versus actual seawater temperature values based on random forest regressionBack to article page