# LEAR¶

The LEAR model is a parameter-rich ARX model estimated using the LASSO as an implicit feature selection that was originally proposed by Uniejewski (2016). It has been used in multiple studies and it has often shown state-of-the-art results in electricity price forecasting, e.g. see Uniejewski (2016) or Lago (2018).

The LEAR model is provided in the library as a sigle LEAR class. The class receives as parameter the calibration window of the method, and has three four main function: a function to recalibrate the model, a function to make predictions, a function to recalibrate and predict, and a function that can perform daily recalibration and prediction using pandas DataFrames.

Besides the LEAR class, the library also includes the evaluate_lear_in_test_dataset function. This function can be used as a simplified interface to evaluate a pandas DataFrame by simply specific the dates of the training and test datasets.

The library also includes a couple of LEAR Examples to get users familiar with the syntax and capabilities of the model.