The classification tree tries to optimize to pure nodes containing Maximum deviance reduction (also known as Split criterion options are Gini's diversity Try each of the three settings to see if they improve the model with Specify the split criterion measure for deciding when to split nodes. Also, a coarse tree is easy to interpret. Training accuracy can approach that of a representative test But a coarse tree can be more robust in that its In contrast, a coarse tree does not attain high trainingĪccuracy. Lower than its training (or resubstitution) accuracy.

Tends to overtrain, and its validation accuracy is often far However, the tree might not showĬomparable accuracy on an independent test set. The best classifier type dependsĪ fine tree with many leaves is usually highly accurate on Want in speed, flexibility, and interpretability. Table as a guide for your final choice of algorithms. MATLAB ® function called by each one for binary or multiclass data. Typical characteristics of different supervised learning algorithms and the Hyperparameters automatically, see Hyperparameter Optimization in Classification Learner App.įor help choosing the best classifier type for your problem, see the table showing To use optimizable model options and tune model Starting points with different settings, suitable for a range of differentĬlassification problems. Nonoptimizable model options in the Models gallery are preset

Models section to expand the list of classifiers. To see all available classifier options, on the If you want to explore classifiers one at a time, or you already know whatĬlassifier type you want, you can select individual models or train a group of the Trains every type regardless of any prior trained models. Use this option to train all available nonoptimizable model The option also fits a binary GLM logistic This option fits only linear SVM,Įfficient linear SVM, efficient logistic regression, and linearĭiscriminant models. Try this option if you expect linear boundaries between theĬlasses in your data. The app will train all the model typesĪvailable for your data set that are typically fast to

Matlab toolboxes logistic regresssion