How to run this

1. Clone this repo
2. Make sure you have the required libraries installed (preferably with pip): pandas, numpy, lightgbm, sklearn
3. Create empty directory data-x-li-data at the same level as the directory where you cloned this repo. It will be used for source CSVs and generated outputs so they won't be committed to this repo.
4. Run ir_nss.py. This will generate the file nss.csv in the data-x-li-data directory. Be patient, it takes about two minutes.
5. Put the following files (you can find them zipped in MS Teams) in the data-x-li-data directory (keep the filenames lowercase to avoid errors on some systems):
   • scen_0001-0200.csv
   • scen_0201-0500.csv
   • scen_0501-0700.csv
   • scen_0701-1000.csv
   • scen_1001-1300.csv
   • scen_1301-1600.csv
   • scen_1601-1900.csv
   • scen_1901-2236.csv
6. Run data_load.py. This will read the source CSVs and nss.csv file and generate df_merged_train_test.pickle, again in the data-x-li-data directory.
7. Run lightgbm_exec.py which reads the df_merged_train_test.pickle file and prints low_MAPE, best_params and best_fit_no (see the bottom of the file). And this takes a looong time.