feature selection for regression

Hello everyone

To do my PhD thesis, I desperately need MATLAB code to select the effective features in the regression problem (time series forecasting) that takes into account the relationship between the features and works as well as possible. Thanks for helping me.

Answer 1

[Sulaymon Eshkabilov]

Test and study these posts and codes posted on MATHWORKS:

https://www.mathworks.com/discovery/feature-selection.html?s_tid=answers_rc2-2_p5_MLT

https://www.mathworks.com/matlabcentral/fileexchange/72177-feature-selection?s_tid=answers_rc2-1_p4_MLT

[sahar rezaei]

Hello Sulaymon

Thank you very much for your answer, but in my research I have compared several time series prediction methods that the accuracy of these methods strongly depends on the type of feature selection method. Accordingly, I need more precise methods to select effective features that not only consider the relationship between the relevant feature and the target, but also the relationship between the features. I ask all my friends if anyone has access to an accurate feature selection method to share it with me.

Thank you very much for guiding me in this direction.

[Sulaymon Eshkabilov]

Most Welcome! Good luck with your studies.

Answer 2

[Image Analyst]

I don't know what you tried already but have you tried the Regression Learner app on the apps tab? If you don't know how, I can show you if you upload your data - both the measurements/predictors, and the ground truth/true answer you're trying to predict with some model. I can try out many models and find the one with the least errors.

[sahar rezaei]

Thank you very much for your help. I will definitely try your suggested method. I have used a combination of the Non-dominated Sorting Genetic Algorithm II (NSGA- ӀӀ) and the MLP Neural network, but this method works very slowly. I think a combination of optimization algorithms and a filter methods like mutual-information method or a combination of filter and wrapper methods can work well. Does anyone have access to a matlab code that can be used for regression?

[Image Analyst]

Yes, almost all of us have regression code of some sort or another.

So. . . did the Regression Learner work for you?

Answer 3

[sahar rezaei]

Hello

I apologize for the delay in responding. I have not tried the Regression Learner App yet but I have used methods such as correlation analysis, mutual-information and PCA. Filter methods are not very suitable for my work because these methods in the feature selection process only consider the relationship between the desired feature and the target and ignore the relationship between the selected features.I think that a combination of optimization algorithms with filter and wrapper methods or a combination of filter and wrapper methods bring more accuracy to the feature selection process. These methods have been mentioned in some authoritative articles. Thanks for guiding me on this.

Sincerely

[Image Analyst]

So, try it. If you don't know how, attach your data and I'll do it.

[sahar rezaei]

Thanks a lot

Answer 4

[David Willingham]

Hi, if you're looking to perform feature engineering with machine learning models, have you tried automl? If not, check out this page for more information:

https://www.mathworks.com/discovery/automl.html

[sahar rezaei]

Hi,

Thank you very much

The following page provides a feature selection method for classification problems. Can anyone guide me on how I can use this method for regression problems?

https://www.mathworks.com/matlabcentral/fileexchange/14608-mrmr-feature-selection-using-mutual-information-computation?s_tid=srchtitle

Answer 5

[sahar rezaei]

Hello everyone

I got the answer to this question. The proposed method can be used for both classification and regression problems.