Model Evaluation, Model Selection, and Algorithm Selection in Machine Learning
Sebastian Raschka
TL;DR
This work surveys the core techniques for estimating generalization performance, selecting models and hyperparameters, and comparing learning algorithms in machine learning. It distinguishes between generalization performance and model/algorithm selection, and discusses bias-variance trade-offs, resampling, and uncertainty estimation using holdout, stratified sampling, bootstrap, and cross-validation. It systematically reviews statistical tests (McNemar, Cochran's Q, F-tests) and modern practices (nested CV, 5x2 CV) for robust comparisons on small to moderate datasets, offering practical recommendations and cautions about bias, variance, and data leakage. The conclusion emphasizes that while no single method fits all scenarios, a careful combination of cross-validation, independent testing, and appropriate statistical tests—tailored to dataset size and computation—provides a principled framework for reliable model and algorithm evaluation and selection.
Abstract
The correct use of model evaluation, model selection, and algorithm selection techniques is vital in academic machine learning research as well as in many industrial settings. This article reviews different techniques that can be used for each of these three subtasks and discusses the main advantages and disadvantages of each technique with references to theoretical and empirical studies. Further, recommendations are given to encourage best yet feasible practices in research and applications of machine learning. Common methods such as the holdout method for model evaluation and selection are covered, which are not recommended when working with small datasets. Different flavors of the bootstrap technique are introduced for estimating the uncertainty of performance estimates, as an alternative to confidence intervals via normal approximation if bootstrapping is computationally feasible. Common cross-validation techniques such as leave-one-out cross-validation and k-fold cross-validation are reviewed, the bias-variance trade-off for choosing k is discussed, and practical tips for the optimal choice of k are given based on empirical evidence. Different statistical tests for algorithm comparisons are presented, and strategies for dealing with multiple comparisons such as omnibus tests and multiple-comparison corrections are discussed. Finally, alternative methods for algorithm selection, such as the combined F-test 5x2 cross-validation and nested cross-validation, are recommended for comparing machine learning algorithms when datasets are small.