How does MATLAB perform the "automated feature engineering for regression"?

I am exploring MATLAB's new feature, "genrfeatures", to generate more features for my multiple non-linear regression problem (I have sensor data that is collected every minute for a year). This function can generate new features using my current predictors, but I wanted to know how it came up with a new feature based on the different responses? For example, for response "a", I can see it generated different features such as "Euclidean distance to centroid 10 - kmd10", but it did not generate this feature (kmd10) with another response "b" and the same predictors.

I was wondering if anyone knows any references for further study regarding this?

Answer 1

[Shivansh]

Hi Smoa,

I understand that you are using “genrfeatures” to generate more features for non-linear regression problem. The “genrfeatures” function enables you to automate the feature engineering process in the context of a machine learning workflow.

It is difficult to explain the exact reason behind the generation of different features for ‘a’ and ‘b’ without knowing more about the data and machine learning environment. You can read more about the “genrfeatures” by referring to the MATLAB documentation here https://in.mathworks.com/help/stats/genrfeatures.html?searchHighlight=genrfeatures&s_tid=srchtitle_support_results_1_genrfeatures.