Using a trained regression model in app designer

I used regression learner to create a trained model that can predict future inflation based on inflation data of the past 10 years. I can use this model to predict new data in command window by typing "trainedModel.predictFcn(T)".

I'm trying to do exact same thing in the app designer. I created a private property called model and loaded the trained model into that property in the startupFcn function.

app.model = load("Model\model2.mat");

I put a button in my app to prediction based on the data in the app, I have this code in its function

app.EditField.Value = app.model.predictFcn(data);

but it doesn't work, I get this error "unrecognized field name 'predictFcn". I'm new user and I really don't have any idea what I'm doing wrong.

Answer 1

[Drew]

Short answer

The short answer is that there is probably another layer in the app.model object, so instead of

app.EditField.Value = app.model.predictFcn(data);

you need something like

app.EditField.Value = app.model.trainedModel.predictFcn(data);

Elaboration

As an example, I loaded the fisheriris data, launched Classification Learner, built the default tree model, exported it as trainedModel, and saved it into model2.mat. Note that Classification Learner and Regression Learner work similarly with respect to this behavior, so an example on either will illustrate the point nicely.

t=readtable("fisheriris.csv")
classficationLearner  % in the app, load the fisher iris data, build default tree, and export it as trainedModel
save model2.mat trainedModel

Then I launched AppDesigner, added to the startupFcn as you did

app.model = load("model2.mat");

then set a breakpoint at something like your second line of code in a callback for a button press, and I see this when inspecting app.model at the commandline:

K>> app.model

ans = 

  struct with fields:

    trainedModel: [1×1 struct]

K>> app.model.trainedModel

ans = 

  struct with fields:

            predictFcn: @(x)exportableModel.predictFcn(predictorExtractionFcn(x))
     RequiredVariables: {'PetalLength'  'PetalWidth'  'SepalLength'  'SepalWidth'}
    ClassificationTree: [1×1 ClassificationTree]
                 About: 'This struct is a trained model exported from Classification Learner R2022b.'
          HowToPredict: 'To make predictions on a new table, T, use: ↵  yfit = c.predictFcn(T) ↵replacing 'c' with the name of the variable that is this struct, e.g. 'trainedModel'. ↵ ↵The table, T, must contain the variables returned by: ↵  c.RequiredVariables ↵Variable formats (e.g. matrix/vector, datatype) must match the original training data. ↵Additional variables are ignored. ↵ ↵For more information, see How to predict using an exported model.'
   

This app designer code worked to test random fisheriris observations. For this simple demo the results are just displayed to the command window with the "disp" function.

classdef tutorialApp < matlab.apps.AppBase

 % Properties that correspond to app components
    properties (Access = public)
        UIFigure               matlab.ui.Figure
        TestRandomPointButton  matlab.ui.control.Button
    end

    properties (Access = public)
        model % model saved from classification Learner
        t % fisher iris data for testing
        index % which obvervation index was tested
        predictedLabel % which label was predicted
    end
    
    % Callbacks that handle component events
    methods (Access = private)

        % Code that executes after component creation
        function startupFcn(app)
            app.model=load("model2.mat");
            app.t=readtable('fisheriris.csv');
        end

        % Button pushed function: TestRandomPointButton
        function TestRandomPointButtonPushed(app, event)
            app.index=randi(150);
            app.predictedLabel=app.model.trainedModel.predictFcn(app.t(app.index,:));
            disp(app.predictedLabel)
            disp(app.index)
        end
    end
    

If this answered your question, please remember to accept the answer.