Table 1.

Machine-learning algorithms and their example applications to metabolic marker discovery.

Algorithm	Description	Examples
Logistic regression (LR)	Use a logistic function to fit a regression model for categorical outcome prediction.	[59]
Partial least squares-discriminant analysis (PLS-DA)	Find a linear subspace of high-dimensional explanatory variables to maximize the covariance between the input variables and the class label.	[60]
Support vector machine (SVM)	Use various similarity measures of training samples (also known as kernel functions) to perform linear or non-linear separation of two classes.	[61]
Random forest (RF)	Construct an ensemble of decision trees to classify training samples, as well as to assess the variable importance in the classification.	[25]
Gradient boosting machine (GBM)	Build an ensemble of decision trees in a step-wise fashion using boosting and gradient descent algorithms.	[62]
Artificial neural network (ANN)	Construct multi-layered networks of neurons to learn highly non-linear functions that map the explanatory variables to the class label.	[63,64]
Genetic programming (GP)	Use natural evolution mechanisms to automatically search for the most relevant features and classification models.	[7,65]