Classification Learner

Train models to classify data using supervised machine learning

Description

The Classification Learner app trains models to classify data. Using this app, you can explore supervised machine learning using various classifiers. You can explore your data, select features, specify validation schemes, train models and optimize hyperparameters, assess results, and investigate how specific predictors contribute to model predictions. Perform automated training to search for the best classification model type, including decision trees, discriminant analysis models, support vector machines, logistic regression models, nearest neighbors, naive Bayes models, kernel approximation models, ensemble models, and neural network classification models. To compare models, use the metric results table and view results plots in the app.

Perform supervised machine learning by supplying a known set of observations of input data (predictors) and known responses (labels or classes). Use the observations to train a model that generates predicted responses for new input data. You can then check model performance using a test data set. To understand how the model uses predictors to make predictions, use global and local interpretability tools, such as partial dependence plots, LIME values, and Shapley values.

To use the trained model with new data, you can export the model to the workspace, Simulink^®, and MATLAB^® Production Server™. You can generate MATLAB code to recreate the trained model outside of the app and explore programmatic classification and further customization of the model training workflow. Export the model training code to Experiment Manager to perform additional tasks, such as changing the training data, adjusting hyperparameter search ranges, and running custom training experiments.

Tip

To get started, in the Models section of the Learn tab, try All Quick-To-Train to train a selection of models. See Automated Classifier Training.

Required Products

• MATLAB

• Statistics and Machine Learning Toolbox™

Open the Classification Learner App

• MATLAB Toolstrip: On the Apps tab, under Machine Learning and Deep Learning, click the app icon.

• MATLAB command prompt: Enter classificationLearner.

Examples

• Start a Classification Learner or Regression Learner Session
• Select Validation Scheme in Classification Learner or Regression Learner
• Train Classification Models in Classification Learner App
• Choose Classifier Options in Classification Learner
• Visualize and Assess Classifier Performance in Classification Learner
• Import Trained Model from Workspace into Classification Learner or Regression Learner
• Test Trained Models in Classification Learner or Regression Learner
• Export Classification Model to Predict New Data

Programmatic Use

classificationLearner opens the Classification Learner app or brings focus to the app if it is already open.

classificationLearner(Tbl,ResponseVarName) opens the Classification Learner app and populates the New Session from Arguments dialog box with the data contained in the table Tbl. The ResponseVarName argument, specified as a character vector or string scalar, is the name of the response variable in Tbl that contains the class labels. The response variable cannot contain more than 500 unique class labels. The remaining variables in Tbl are the predictor variables.

classificationLearner(Tbl,Y) opens the Classification Learner app and populates the New Session from Arguments dialog box with the predictor variables in the table Tbl and the class labels in the vector Y. You can specify the response Y as a categorical array, character array, string array, logical vector, numeric vector, or cell array of character vectors. Y cannot contain more than 500 unique class labels.

classificationLearner(X,Y) opens the Classification Learner app and populates the New Session from Arguments dialog box with the n-by-p predictor matrix X and the n class labels in the vector Y. Each row of X corresponds to one observation, and each column corresponds to one variable. The length of Y and the number of rows of X must be equal. Y cannot contain more than 500 unique class labels.

classificationLearner(___,Name=Value) specifies data partition options for cross-validation and testing using one or more of the following name-value arguments, in addition to any of the input argument combinations in the previous syntaxes. For example, you can specify KFold=10 to use a 10-fold cross-validation scheme.

• CrossVal, specified as "on" (default) or "off", is the cross-validation flag. If you specify "on", the app uses 5-fold cross-validation. If you specify "off", the app uses resubstitution validation.

  You can override the CrossVal cross-validation setting by using the Holdout or KFold name-value argument. You can specify only one of these three arguments at a time.

• Holdout, specified as a numeric scalar in the range [0.05,0.5], is the fraction of the training data set used for holdout validation. The training data set is the data in Tbl or X that is not set aside for testing.

• KFold, specified as a positive integer in the range [2,50], is the number of folds to use for cross-validation.

• TestDataFraction, specified as a numeric scalar in the range [0,0.5], is the fraction of the data in Tbl or X that is set aside for testing.

• ValidationPartition, specified as a cvpartition object, that defines the validation scheme and the indexing for the validation sets. The indices in the object correspond to the rows in Tbl or X. When you specify ValidationPartition:

  • You cannot specify any other name-value argument except TestPartition.

  • You cannot set aside a portion of Tbl or X for testing at the start of the session. To learn how to load a test data set during the session, see Test Trained Models in Classification Learner or Regression Learner.

  For more information about validation schemes, see Select Validation Scheme in Classification Learner or Regression Learner.

• TestPartition, specified as a cvpartition object, that defines the indexing for the rows in Tbl or X to be set aside for testing. When you specify TestPartition:

  • The Type property of the cvpartition object must be 'holdout'.

  • You cannot specify TestDataFraction.

  • You can specify only one of these four arguments: CrossVal, Holdout, KFold, or ValidationPartition.

  For more information about test data sets, see Test Trained Models in Classification Learner or Regression Learner.

classificationLearner(filename) opens the Classification Learner app with the previously saved session in filename. The filename argument, specified as a character vector or string scalar, must include the name of a Classification Learner session file and the path to the file, if it is not in the current folder. The file must have the extension .mat.

Limitations

• Classification Learner does not support model deployment to MATLAB Production Server in MATLAB Online™.

Version History

Introduced in R2015a

R2026a: Import a trained model from the MATLAB workspace

Import a supported trained machine learning model from the MATLAB workspace in one of two ways:

• On the Learn tab, in the File section, select New Session > From Trained Model.

• On the Learn tab, in the File section, click Import Model.

After you import a trained model into the app, you can:

• Assess model performance using a test data set.

• Explain model behavior using interpretability plots.

• Export the model for deployment.

For more information and a list of restrictions, see Import Trained Model from Workspace into Classification Learner or Regression Learner.

R2026a: Specify validation and test partitions

Use the ValidationPartition name-value argument to specify a cvpartition object that defines the validation scheme and the indexing for the validation sets. To define the indexing for the test data set, specify a cvpartition object of type 'holdout' using the TestPartition name-value argument.

R2026a: Export partitions and data sets

Export the partitions used to compute validation and test metrics, as well as the data sets, from the current session to the MATLAB workspace. In the Export section of the Learn tab, select Export > Export Partitions and Data Sets. For more information, see Export Partitions and Data Sets from Classification Learner or Regression Learner.

R2026a: Train customizable neural networks (requires Deep Learning Toolbox)

Create a customizable neural network model by selecting the model preset Fully Connected Customizable Neural Network or Residual Customizable Neural Network in the Models section of the Learn tab. To customize the model's neural network using the Network Editor, select the model in the models pane, and click Customize Network on the model Summary tab. The default model for the fully connected customizable network contains five fully connected layers (excluding the final fully connected layer for prediction). The default model for the residual customizable network contains one residual connection. For more information, see Customizable Neural Network Classifiers and Edit Customizable Neural Network Using Network Editor in Classification Learner or Regression Learner.

R2026a: Export feature ranking plot

Export a feature ranking plot or its data by first creating the plot using the Feature Selection button in the Options section of the Learn tab. Select Export Plot to Figure or Export Plot Data in the Export section.

R2026a: Automatically select the number of predictors to sample in ensemble tree models

When you create a tree model from the Ensemble Classifiers or Ensembles of Trees section of the Models gallery, the default setting for Number of predictors to sample is Auto. For Boosted Tree and RUSBoosted Tree models, the Auto setting is equivalent to the default Select All setting in previous releases. For Bagged Tree models, Auto sets the number of predictors to sample as the square root of the number of predictors.