Method Number of Variable Variable Selection necessary Applications Implementation
MLR Small, uncorrelated yes Simple, linear relationships Fast and simple
Partial Least Squares Regression PLS Large, typically correlated no First idea of a prediction model, process development Fast and simple
Structured Additive Regression STAR moderate yes Production process Computationally very intensive
Random Forests RF large no Quick idea of a prediction model, variable importance, process development Rather fast, some parameter tuning required
Support Vector Machines Regression SVM large no Production process A lot of parameter tuning required
Neural Networks NN moderate yes Production process Rather slow
Deep Learning DL huge no Production process Computationally very intensive
Gaussian Process Regression GPR moderate no Process development Rather fast