Machine Learning for Panel Data

Can anyone please confirm whether Matlab supports machine learning methods for panel/longitudinal data?

I have not been able to find anything around trees/random forests etc.

Many thanks

Answer 1

[Hiro Yoshino]

Yes, it does.

table type variable is supported by many ML models. For example, SVM (https://jp.mathworks.com/help/stats/fitcecoc.html) accepts "Tbl" (table) as a first argument.

This is where "table" is elaborated https://jp.mathworks.com/help/matlab/tables.html

[Nick]

Thanks very much Hiro.

Does that also accept unbalanced panel data or do they have to be balanced?

Also, in my case I have a dataset broadly similar to this one:

load RetailCreditPanelData.mat

I couldn't see in the links you provided how or where I'd let the model know to use the ID (Loan Identifier column in this example) such that it understands that the data is indeed panel data with the ID variable being the variable that distinguishes it between individuals.

[Hiro Yoshino]

Looks like you need to get yourself familiarized with the MATLAB documentation a bit more.

Let's us take https://www.mathworks.com/help/stats/fitcecoc.html#namevaluepairarguments as an example.

You will see the available syntaxes as follows:

Mdl = fitcecoc(Tbl,ResponseVarName)

Mdl = fitcecoc(Tbl,formula)

Mdl = fitcecoc(Tbl,Y)

Mdl = fitcecoc(X,Y)

Mdl = fitcecoc(___,Name,Value)

[Mdl,HyperparameterOptimizationResults] = fitcecoc(___,Name,Value)

then you may want to puress the link to the "Name" & "Value" option. This option allows you to tune the model finely as you wish. In the list of this option, you'll find "PredictorNames" and "ResponseName". These are where you specify the predictors and the response respectively.

[Nick]

Thank you Hiro. Apologies for being repetitive but having gone through the links I still have trouble understanding this.

In this simple example below how exactly do I add the ID variable (data.ID) in this model given this is a panel data set where there are different borrowers (i.e. different IDs) with several observations under each ID?

load RetailCreditPanelData.mat
data = data(1:1000,:);

X = data(:,[2:3,5]);
Y = data.Default;
Model1 = fitcecoc(X,Y);