Files:

dem.panel.tab : original dataset

(6)Analysis_sect_5_2_Tables_2_3.R : original R script

Oren_replication.R : cleaned up version of the original R script, used to produce the replicated table.

Oren_bigger_test.R : same as above, only with bigger test set, used to produce the table with the bigger test set.

clr.py : cyclical learning rate scheduler for Keras, taken from https://github.com/bckenstler/CLR.

train.csv, test.csv : exported the preprocessed datasets from R

Final Project - Python Part.ipynb : used together with the authors’ dataset for most of the first Python table

Final Project - Python Part with fastai.ipynb : used together with the authors’ dataset for the PyTorch-NN row in the first Python table

train_raw.csv, test_raw.csv : exported raw datasets from R

Final Project - Python Part with preprocessing.ipynb : used together with the raw dataset in order to preprocess it for most of the second Python table

Final Project - Python Part with preprocessing + fastai.ipynb : used together with the raw dataset in order to preprocess it for the Pytorch-NN row in the second Python table

train_preds.csv, test_preds.csv : exported preprocessed datasets from R to use for stacking, did not make it into the final project analysis

Final Project - Python Part With Stacking.ipynb : My attempt at adding the results of the R models to the data and feeding that into the models (stacking) - did not improve the results so was not included in the final project analysis.