This feature is exclusively available to Starter, Research, and Pro. Upgrade when you need this workflow, review pricing, or send a question before changing plans.
What you keep on Free
Empirical · Machine learning / AI forecasting · Subgroup
Linear forecasters that add a penalty to the loss to control variance and select features when there are many candidate predictors -- the gateway ML class for macro forecasting.
What this subgroup is for
Reach for penalized linear models when you have far more candidate predictors than the sample window can support and you want shrinkage or feature selection without leaving the linear world.
Reference coverage
2 models linked in this subgroup's empirical reference coverage.
Models currently covered
Each model below has its own empirical reference page with overview, graph, proof, and comparison material.
L2-penalized linear regression -- shrinks coefficients toward zero to stabilize forecasts when predictors are many or collinear.
Best for: Macro forecasting with many candidate predictors where ordinary least squares overfits and needs shrinkage.
L1-penalized linear regression -- drives some coefficients to zero, doing variable selection inside the fit.
Best for: Wide predictor sets where you want the model to choose a sparse subset of features instead of using all of them.
Back to the cluster
The cluster page lists every model in this empirical class, including the ones that do not yet have a fifth-layer subgroup.