Hello again, I was away for a few days.

Back to this issue.

I find this interesting too. Periklis, if you've successfully used libsvm for econometric purposes, please give us a (at least) a brief description. Libsvm looks like something we could work with quite easily.

The profile of machine learning has been raised recently by the article by Mullainathan and Spiess in the Spring 2017 issue of the Journal of Economic Perspectives, and it's possible this is something we'd want to explore in gretl.

Allin Cottrell

In my research I have used Support Vector Machines/Regression in forecasting both financial and macroeconomic variables (I attach two relevant papers, one on exchange rate forecasting and one on housing prices ). These applications of ML were coded using matlab.

The thing with the SVM/R is to fine-tune the hyperparameters and the kernel parameters. Libsvm already provides ready wrappers for matlab and R that are used to import values and export results (available in their page). The main optimization routine is written in C so it is quite fast. All you need to build is the wrapper to link libsvm libraries with gretl. Of course you could also develop you own functions for cross validation training and parameter seach. That shouldn't be that hard.

It would be nice to be able to use Gretl for Machine Learning!