To reproduce the results of Section 5.2.5 (electricity consumption forecasting)
you can obtain the mentioned train, validation and test set from running the 
`prep_data_elec.R` file which takes the raw data (`elec.csv`) as input. This
produces the `electricity.RDS` file which serves as the input for `neat_on_ts.R`
and `arima_on_ts.R` which, in turn, produce the log-scores reported for the
DRIFT model and the ARIMA model respectively. The `utils.R` file holds corresponding
helper functions for both DRIFT and ARIMA.