This repository contains the codes used to obtain the results presented in Section 4 of ...

To reproduct the results, one needs to first download the google mobility data in csv form for the region "France" on https://www.gstatic.com/covid19/mobility

The preprocessing is done using the files "Preprocessing_first_part.R" (which aggregates the data from the different sources), then using the file "Preprocessing_second_part.R"
(which detects outlier data in regional consumption and attributes them a weight equal to 0).

Then, the regional and national GAM are computed using the file "compute_GAM.R".

The staked GAM and RF (individual and commons) are computed using the file "Stacked_RF.R"

The aggregation is performed using the file "Aggregation.R"

The remainder of the files are used to produce the analyses presented Section 4. The file "ALE" produces the ALE plots of the differents variables. The file 
"RF_importance.R" computes the importance of the different variables. The file "Variable_Selection.R" compares the different variable selections methods.

The files "Data_regional_*.RDS" are data at the regional level, divided into 3 period for reasons of size constraints (2013-2015, 2016-2017, 2018-2020).
The file "Data_Data_RTE_janv2012_oct2020_V3.RDS" contains data at the national level.
The file "CGRT.RDS" contains Government Response Index for France, collected at https://www.bsg.ox.ac.uk/research/research-projects/covid-19-government-response-tracker .