How to Use Classification in Simulink for Data Science

To use classification in Simulink, import or create a trained classification model in MATLAB, then integrate it using the MATLAB Function block or Predict block from the Statistics and Machine Learning Toolbox. Connect your input data signals to the block and output the predicted class labels for real-time classification.

Syntax

In Simulink, classification is typically done by embedding a trained model using the MATLAB Function block or the Predict block. The general syntax inside the MATLAB Function block is:

• label = predict(model, inputData); - where model is your trained classification model and inputData is the feature vector.
• The label output is the predicted class.

This syntax calls the predict function from MATLAB's Statistics and Machine Learning Toolbox.

matlab

function label = fcn(inputData)
    %#codegen
    persistent model;
    if isempty(model)
        model = coder.load('trainedModel.mat');
    end
    label = predict(model, inputData);
end

Example

This example shows how to classify input data in Simulink using a trained classification model saved in MATLAB. The model is loaded once and used to predict class labels for incoming data signals.

matlab

% MATLAB script to create a simple classification model and save it
% Create sample data
X = [randn(50,2)+1; randn(50,2)-1];
Y = [ones(50,1); zeros(50,1)];
% Train a simple classification tree
model = fitctree(X,Y);
% Save the model
save('trainedModel.mat','model');

% Simulink MATLAB Function block code:
function label = fcn(inputData)
    %#codegen
    persistent model;
    if isempty(model)
        model = coder.load('trainedModel.mat');
    end
    label = predict(model, inputData);
end

Output

No direct output; Simulink block outputs predicted class labels for each inputData vector.

Common Pitfalls

• Not loading the model properly: Forgetting to load the trained model inside the MATLAB Function block causes errors.
• Input data shape mismatch: The input feature vector must match the model's expected size.
• Using unsupported functions: Some MATLAB functions are not supported for code generation in Simulink.
• Not setting #codegen directive: This directive is needed for code generation compatibility.

matlab

function label = fcn_wrong(inputData)
    % This will fail because model is not persistent and loaded every call
    load('trainedModel.mat','model');
    label = predict(model, inputData);
end

% Correct way:
function label = fcn_right(inputData)
    %#codegen
    persistent model;
    if isempty(model)
        model = coder.load('trainedModel.mat');
    end
    label = predict(model, inputData);
end

Quick Reference

Tips for classification in Simulink:

• Train and save your model in MATLAB before using it in Simulink.
• Use persistent variables to load the model once inside MATLAB Function blocks.
• Ensure input data matches the model's expected features.
• Use #codegen directive for compatibility.
• Test your model in MATLAB before integrating into Simulink.

Key Takeaways

Load your trained classification model once using persistent variables inside MATLAB Function blocks.

Use the predict function to classify input data signals in Simulink.

Ensure input data shape matches the model's expected features to avoid errors.

Include the #codegen directive for code generation compatibility.

Test your classification model in MATLAB before Simulink integration.