MATLAB: How to optimize machine-learning model exported from regression learner app

Question

Thank you for visiting this question.

Now I've tried to optimize my machine-learning model exported from regression learner app.

This model has 27 inputs, and there is no certain numerical equation for result.

I googled keywords such as optimization of trained model, or optimization using GAMS, but I've failed to find proper items.

Please help me !

Answer 1

Hi Manhee,

After exporting the model from Regression Learner App, a structure variable is exported to workspace using it prediction on the test dataset can be done.

There are two possible ways to optimize:

• Optimization within regression learner app - Refer to Train Regression Models in Regression Learner App documentation which has example related to optimization in App and Hyperparameter Optimization in Regression Learner App documentation which describes the hyperparameters that can optimize for each type of model.

• Optimization can be done in MATLAB script after using Generate Function under Export. This will generate MATLAB script containing trainRegressionModel function as shown below:

% [trainedModel, validationRMSE] = trainRegressionModel(trainingData) 
% Returns a trained regression model and its RMSE. This code recreates the 
% model trained in Regression Learner app. 
% 



%  Input: 
%      trainingData: A table containing the same predictor and response 
%       columns as those imported into the app. 
% 
%  Output: 
%      trainedModel: A struct containing the trained regression model. The 
%       struct contains various fields with information about the trained 
%       model. 
% 
% For example, to retrain a regression model trained with the original data 
% set T, enter: 
%   [trainedModel, validationRMSE] = trainRegressionModel(T) 
% 
% To make predictions with the returned 'trainedModel' on new data T2, use 
%   yfit = trainedModel.predictFcn(T2) 

The trainRegressionModel function has fit function which can be modified to optimize the model. Refer to Bayesian Optimization Workflow documentation (section - Parameters Available for Fit Functions) for more information related to different kind of fit functions.