Table 3.

Accuracy and 95% confidence intervals for Gradient Boosting Machine (GBM), Random Forest and Neural Network (Nnet) predictions with the top 16 features selected using Recursive feature elimination (RFE).

	CSD17
	Continent	City
GBM(AMR)	0.61 (0.55,0.67)	0.40 (0.34,0.47)
Random Forest (AMR)	0.55 (0.48,0.61)	0.46 (0.4,0.52)
Neural Net (AMR)	0.51 (0.45,0.57)	0.38 (0.32,0.44)
GBM (proGenomes)	0.82 (0.77,0.87)	0.67 (0.61,0.72)
Random Forest (proGenomes)	0.81 (0.76,0.86)	0.71 (0.66,0.76)
Neural Net (proGenomes)	0.78 (0.73,0.83)	0.60 (0.54,0.66)

For each method the tuning parameters are selected such that the best accuracy is achieved.