DY0-001 practice question 4

The Akaike information criterion (AIC) compares models fitted by maximum likelihood, trading goodness of fit against the number of parameters, and it does not require the models to be linear or nested, so it suits comparing several nonlinear models. ANOVA-based comparison and the chi-squared test assume specific (often nested or linear) structures or test independence and goodness of fit rather than ranking competing models. MCC (Matthews correlation coefficient) is a single classification metric from a confusion matrix and only applies to classifiers, not to nonlinear models in general.