Table 3.

Eleven machine learning models for estimating resting energy expenditure ranked by the lowest root mean squared error.

	RMSE		Optimized features (N = 65)	Run time
Model	(Kcal)	R 2	(n)	(Seconds)
SVR	103.6	0.918	35	35.0
Linear regression	119.0	0.892	37	19.8
Elastic Net	121.1	0.888	24	25.5
Linear SVR	128.9	0.874	37	22.3
MLP regressor	131.2	0.869	41	1239.0
Bayesian ridge	134.6	0.862	24	22.8
Ridge	136.4	0.858	35	17.7
SGD regressor	139.1	0.853	29	25.6
Lasso	140.9	0.849	26	19.1
Gradient boosting	156.0	0.815	8	170.1
Random forest	161.6	0.801	15	412.4

MLP: multi layer perception; R²: correlation coefficient; RMSE: route mean squared error; SGD: stochastic gradient descent; SVR: support vector regression.