def main():
data = dataset.UNDevGoalsDataset()
X, Y = data.preprocess_simple()
# Run simple baseline model
print('Predicting 2007 from 1972:2006')
status_quo_preds = status_quo_model(X)
status_quo_rmse = data.evaluate(status_quo_preds)
print('Status quo model RMSE:', status_quo_rmse)
def main():
data = dataset.UNDevGoalsDataset()
X, y = data.preprocess_simple()
sq = models.status_quo_model(X)
# Currently the arima model doesn't output a DataFrame
#arima = models.arima(X)
# This function returns a GroupBy object where each group is a development index
gb = data.preprocess_for_viz()
for name, group in gb:
visualize_all(name, group)
visualize_worst(name, group, sq)
def main():
data = dataset.UNDevGoalsDataset()
X, Y = data.preprocess(pp_fn=preprocess_avg_NANs)
X5, Y5 = data.preprocess(pp_fn=preprocess_avg_NANs, years_ahead=5)
preds = data.predictions(model_name=arima,
order=(1, 1, 1),
lookback=5,
preprocessed_data=X)
preds5 = data.predictions(model_name=arima,
order=(1, 1, 1),
lookback=5,
preprocessed_data=X5)
preds = pd.Series(preds, index=X.index)
preds5 = pd.Series(preds5, index=X.index)
# This function returns a GroupBy object where each group is a development index
gb = data.preprocess(pp_fn=preprocess_for_viz)
for name, group in gb:
visualize_worst(name, group, preds, preds5)
def main():
data = dataset.UNDevGoalsDataset()
print('Predicting 2007 from 1972:2006')
X_simple, Y_simple = data.preprocess(pp_fn=preprocess_simple)
status_quo_predictions_simple = data.predictions(
model_name=status_quo_model, preprocessed_data=X_simple)
status_quo_simple_rmse = data.error(
error_fn=RMSE, predictions=status_quo_predictions_simple)
print('Status quo model RMSE with simple preprocessing:',
status_quo_simple_rmse)
X_improved, Y_improved = data.preprocess(pp_fn=preprocess_avg_NANs)
status_quo_predictions_improved = data.predictions(
model_name=status_quo_model, preprocessed_data=X_improved)
status_quo_improved_rmse = data.error(
error_fn=RMSE, predictions=status_quo_predictions_improved)
print('Status quo model RMSE with better preprocessing:',
status_quo_improved_rmse)
arima_predictions_simple = data.predictions(model_name=arima,
order=(1, 1, 1),
lookback=5,
preprocessed_data=X_simple)
arima_rmse_simple = data.error(error_fn=RMSE,
predictions=arima_predictions_simple)
print('ARIMA model RMSE with simple preprocessing:', arima_rmse_simple)
arima_predictions_improved = data.predictions(model_name=arima,
order=(1, 1, 1),
lookback=5,
preprocessed_data=X_improved)
arima_rmse_improved = data.error(error_fn=RMSE,
predictions=arima_predictions_improved)
print('ARIMA model RMSE with better preprocessing:', arima_rmse_improved)
def main():
data = dataset.UNDevGoalsDataset()
X_simple, Y_simple = data.preprocess(pp_fn=preprocess_simple)
X_simple_5, Y_simple_5 = data.preprocess(pp_fn=preprocess_simple,
years_ahead=5)
X_with_global_avg, Y_with_global_avg = data.preprocess(
pp_fn=preprocess_avg_NANs)
X_with_global_avg_5, Y_with_global_avg_5 = data.preprocess(
pp_fn=preprocess_avg_NANs, years_ahead=5)
X_with_cont_avg, Y_with_cont_avg = data.preprocess(
pp_fn=preprocess_with_continent_interpolation)
X_with_cont_avg_5, Y_with_cont_avg_5 = data.preprocess(
pp_fn=preprocess_with_continent_interpolation, years_ahead=5)
status_quo_predictions_simple = data.predictions(
model_name=status_quo_model, preprocessed_data=X_simple)
status_quo_simple_rmse = data.error(
error_fn=RMSE, predictions=status_quo_predictions_simple)
print('Status quo model RMSE with simple preprocessing (1 yr):',
status_quo_simple_rmse)
status_quo_predictions_simple_5 = data.predictions(
model_name=status_quo_model, preprocessed_data=X_simple_5)
status_quo_simple_rmse_5 = data.error(
error_fn=RMSE, predictions=status_quo_predictions_simple_5)
print('Status quo model RMSE with simple preprocessing (5 yr):',
status_quo_simple_rmse_5)
status_quo_rmse_avg = (status_quo_simple_rmse +
status_quo_simple_rmse_5) / 2
print('Status quo model RMSE with simple preprocessing (avg):',
status_quo_rmse_avg)
print()
status_quo_predictions_with_global_avg = data.predictions(
model_name=status_quo_model, preprocessed_data=X_with_global_avg)
status_quo_with_global_avg_rmse = data.error(
error_fn=RMSE, predictions=status_quo_predictions_with_global_avg)
print('Status quo model RMSE with global average imputation:',
status_quo_with_global_avg_rmse)
status_quo_predictions_with_cont_avg = data.predictions(
model_name=status_quo_model, preprocessed_data=X_with_cont_avg)
status_quo_with_cont_avg_rmse = data.error(
error_fn=RMSE, predictions=status_quo_predictions_with_cont_avg)
print('Status quo model RMSE with continent average imputation:',
status_quo_with_cont_avg_rmse)
print()
X_improved, Y_improved = data.preprocess(pp_fn=preprocess_avg_NANs)
status_quo_predictions_improved = data.predictions(
model_name=status_quo_model, preprocessed_data=X_improved)
status_quo_improved_rmse = data.error(
error_fn=RMSE, predictions=status_quo_predictions_improved)
print('Status quo model RMSE with better preprocessing (1 yr):',
status_quo_improved_rmse)
X_improved_5, Y_improved_5 = data.preprocess(pp_fn=preprocess_avg_NANs,
years_ahead=5)
status_quo_predictions_improved_5 = data.predictions(
model_name=status_quo_model, preprocessed_data=X_improved_5)
status_quo_improved_rmse_5 = data.error(
error_fn=RMSE, predictions=status_quo_predictions_improved_5)
print('Status quo model RMSE with better preprocessing (5 yr):',
status_quo_improved_rmse_5)
status_quo_improved_rmse_avg = (status_quo_improved_rmse +
status_quo_improved_rmse_5) / 2
print('Status quo model RMSE with better preprocessing (avg):',
status_quo_improved_rmse_avg)
print()
arima_predictions_simple = data.predictions(model_name=arima,
order=(1, 1, 1),
lookback=5,
preprocessed_data=X_simple)
arima_rmse_simple = data.error(error_fn=RMSE,
predictions=arima_predictions_simple)
print('ARIMA model RMSE with simple preprocessing (1 yr):',
arima_rmse_simple)
arima_predictions_simple_5 = data.predictions(model_name=arima,
order=(1, 1, 1),
lookback=5,
preprocessed_data=X_simple_5)
arima_rmse_simple_5 = data.error(error_fn=RMSE,
predictions=arima_predictions_simple_5)
print('ARIMA model RMSE with simple preprocessing (5 yr):',
arima_rmse_simple_5)
arima_rmse_avg = (arima_rmse_simple + arima_rmse_simple_5) / 2
print('ARIMA model RMSE with simple preprocessing (avg):', arima_rmse_avg)
print()
arima_predictions_with_global_avg = data.predictions(
model_name=arima,
order=(1, 1, 1),
lookback=5,
preprocessed_data=X_with_global_avg)
arima_rmse_with_global_avg = data.error(
error_fn=RMSE, predictions=arima_predictions_with_global_avg)
print('ARIMA model RMSE with global average imputation (1 yr):',
arima_rmse_with_global_avg)
arima_predictions_with_global_avg_5 = data.predictions(
model_name=arima,
order=(1, 1, 1),
lookback=5,
preprocessed_data=X_with_global_avg_5)
arima_rmse_with_global_avg_5 = data.error(
error_fn=RMSE, predictions=arima_predictions_with_global_avg_5)
print('ARIMA model RMSE with global average imputation (5 yr):',
arima_rmse_with_global_avg_5)
arima_rmse_with_global_avg_avg = (arima_rmse_with_global_avg +
arima_rmse_with_global_avg_5) / 2
print('ARIMA model RMSE with global average imputation (avg):',
arima_rmse_with_global_avg_avg)
print()
arima_predictions_with_cont_avg = data.predictions(
model_name=arima,
order=(1, 1, 1),
lookback=5,
preprocessed_data=X_with_cont_avg)
arima_rmse_with_cont_avg = data.error(
error_fn=RMSE, predictions=arima_predictions_with_cont_avg)
print('ARIMA model RMSE with continent average imputation (1 yr):',
arima_rmse_with_cont_avg)
arima_predictions_with_cont_avg_5 = data.predictions(
model_name=arima,
order=(1, 1, 1),
lookback=5,
preprocessed_data=X_with_cont_avg_5)
arima_rmse_with_cont_avg_5 = data.error(
error_fn=RMSE, predictions=arima_predictions_with_cont_avg_5)
print('ARIMA model RMSE with continent average imputation (5 yr):',
arima_rmse_with_cont_avg_5)
arima_rmse_with_cont_avg_avg = (arima_rmse_with_cont_avg +
arima_rmse_with_cont_avg_5) / 2
print('ARIMA model RMSE with continent average imputation (avg):',
arima_rmse_with_cont_avg_avg)
def main():
data = dataset.UNDevGoalsDataset()
# X, Y = data.preprocess(pp_fn=preprocess_by_country_all_years)
# assert False
# Xtr, Ytr, Xval, Yval = data.preprocess(pp_fn=preprocess_by_country_one_year)
# preds = data.predictions(model_name=mlp, preprocessed_data=(Xtr, Ytr, Xval, Yval))
# assert False
#rmse = data.error(error_fn=RMSE, predictions=preds)
#print(rmse)
# Get all the preprocessed data
X_simple, Y_simple = data.preprocess(pp_fn=preprocess_simple)
X_simple_5, Y_simple_5 = data.preprocess(pp_fn=preprocess_simple,
years_ahead=5)
X_with_global_avg, Y_with_global_avg = data.preprocess(
pp_fn=preprocess_avg_NANs)
X_with_global_avg_5, Y_with_global_avg_5 = data.preprocess(
pp_fn=preprocess_avg_NANs, years_ahead=5)
X_with_cont_avg, Y_with_cont_avg = data.preprocess(
pp_fn=preprocess_with_continent_interpolation)
X_with_cont_avg_5, Y_with_cont_avg_5 = data.preprocess(
pp_fn=preprocess_with_continent_interpolation, years_ahead=5)
X_with_cont_avg_and_lin_interp, Y_with_cont_avg_and_lin_interp = data.preprocess(
pp_fn=preprocess_with_continent_and_linear_interpolation)
X_with_cont_avg_and_lin_interp_5, Y_with_cont_avg_and_lin_interp_5 = data.preprocess(
pp_fn=preprocess_with_continent_and_linear_interpolation,
years_ahead=5)
# Status quo with simple preprocessing
status_quo_predictions_simple = data.predictions(
model_name=status_quo_model, preprocessed_data=X_simple)
status_quo_simple_rmse = data.error(
error_fn=RMSE, predictions=status_quo_predictions_simple)
print('Status quo model RMSE with simple preprocessing (1 yr):',
status_quo_simple_rmse)
status_quo_predictions_simple_5 = data.predictions(
model_name=status_quo_model, preprocessed_data=X_simple_5)
status_quo_simple_rmse_5 = data.error(
error_fn=RMSE, predictions=status_quo_predictions_simple_5)
print('Status quo model RMSE with simple preprocessing (5 yr):',
status_quo_simple_rmse_5)
status_quo_rmse_avg = (status_quo_simple_rmse +
status_quo_simple_rmse_5) / 2
print('Status quo model RMSE with simple preprocessing (avg):',
status_quo_rmse_avg)
print()
# Status quo with global avg interpolation
status_quo_predictions_with_global_avg = data.predictions(
model_name=status_quo_model, preprocessed_data=X_with_global_avg)
status_quo_with_global_avg_rmse = data.error(
error_fn=RMSE, predictions=status_quo_predictions_with_global_avg)
print('Status quo model RMSE with global average imputation (1 yr):',
status_quo_with_global_avg_rmse)
status_quo_predictions_with_global_avg_5 = data.predictions(
model_name=status_quo_model, preprocessed_data=X_with_global_avg_5)
status_quo_with_global_avg_rmse_5 = data.error(
error_fn=RMSE, predictions=status_quo_predictions_with_global_avg_5)
print('Status quo model RMSE with global average imputation (5 yr):',
status_quo_with_global_avg_rmse_5)
status_quo_predictions_with_global_avg_rmse_avg = (
status_quo_with_global_avg_rmse +
status_quo_with_global_avg_rmse_5) / 2
print('Status quo model RMSE with global average imputation (avg):',
status_quo_predictions_with_global_avg_rmse_avg)
print()
# Status quo with continent average interpolation
status_quo_predictions_with_cont_avg = data.predictions(
model_name=status_quo_model, preprocessed_data=X_with_cont_avg)
status_quo_with_cont_avg_rmse = data.error(
error_fn=RMSE, predictions=status_quo_predictions_with_cont_avg)
print('Status quo model RMSE with continent average imputation (1yr):',
status_quo_with_cont_avg_rmse)
status_quo_predictions_with_cont_avg_5 = data.predictions(
model_name=status_quo_model, preprocessed_data=X_with_cont_avg_5)
status_quo_with_cont_avg_rmse_5 = data.error(
error_fn=RMSE, predictions=status_quo_predictions_with_cont_avg_5)
print('Status quo model RMSE with continent average imputation (5 yr):',
status_quo_with_cont_avg_rmse_5)
status_quo_predictions_with_cont_avg_rmse_avg = (
status_quo_with_cont_avg_rmse + status_quo_with_cont_avg_rmse_5) / 2
print('Status quo model RMSE with continent average imputation (avg):',
status_quo_predictions_with_cont_avg_rmse_avg)
print()
# Status quo with continent average and linear interpolation
status_quo_predictions_with_cont_avg_lin_interp = data.predictions(
model_name=status_quo_model,
preprocessed_data=X_with_cont_avg_and_lin_interp)
status_quo_with_cont_avg_lin_interp_rmse = data.error(
error_fn=RMSE,
predictions=status_quo_predictions_with_cont_avg_lin_interp)
print(
'Status quo model RMSE with continent average imputation and linear interpolation (1yr):',
status_quo_with_cont_avg_lin_interp_rmse)
status_quo_predictions_with_cont_avg_lin_interp_5 = data.predictions(
model_name=status_quo_model,
preprocessed_data=X_with_cont_avg_and_lin_interp_5)
status_quo_with_cont_avg_lin_interp_rmse_5 = data.error(
error_fn=RMSE,
predictions=status_quo_predictions_with_cont_avg_lin_interp_5)
print(
'Status quo model RMSE with continent average imputation and linear interpolation (5 yr):',
status_quo_with_cont_avg_lin_interp_rmse_5)
status_quo_predictions_with_cont_avg_lin_interp_rmse_avg = (
status_quo_with_cont_avg_lin_interp_rmse +
status_quo_with_cont_avg_lin_interp_rmse_5) / 2
print(
'Status quo model RMSE with continent average imputation and linear interpolation (avg):',
status_quo_predictions_with_cont_avg_lin_interp_rmse_avg)
print()
# Arima with simple preprocessing
arima_predictions_simple = data.predictions(model_name=arima,
preprocessed_data=X_simple)
arima_rmse_simple = data.error(error_fn=RMSE,
predictions=arima_predictions_simple)
print('ARIMA model RMSE with simple preprocessing (1 yr):',
arima_rmse_simple)
arima_predictions_simple_5 = data.predictions(model_name=arima,
preprocessed_data=X_simple_5,
forward=5)
arima_rmse_simple_5 = data.error(error_fn=RMSE,
predictions=arima_predictions_simple_5)
print('ARIMA model RMSE with simple preprocessing (5 yr):',
arima_rmse_simple_5)
arima_rmse_avg = (arima_rmse_simple + arima_rmse_simple_5) / 2
print('ARIMA model RMSE with simple preprocessing (avg):', arima_rmse_avg)
print()
# Arima with global avg interpolation
arima_predictions_with_global_avg = data.predictions(
model_name=arima, preprocessed_data=X_with_global_avg)
arima_rmse_with_global_avg = data.error(
error_fn=RMSE, predictions=arima_predictions_with_global_avg)
print('ARIMA model RMSE with global average imputation (1 yr):',
arima_rmse_with_global_avg)
arima_predictions_with_global_avg_5 = data.predictions(
model_name=arima, preprocessed_data=X_with_global_avg_5, forward=5)
arima_rmse_with_global_avg_5 = data.error(
error_fn=RMSE, predictions=arima_predictions_with_global_avg_5)
print('ARIMA model RMSE with global average imputation (5 yr):',
arima_rmse_with_global_avg_5)
arima_rmse_with_global_avg_avg = (arima_rmse_with_global_avg +
arima_rmse_with_global_avg_5) / 2
print('ARIMA model RMSE with global average imputation (avg):',
arima_rmse_with_global_avg_avg)
print()
# VAR with global avg interpolation
VAR_predictions_with_global_avg = data.predictions(
model_name=var, lookback=4, preprocessed_data=X_with_global_avg)
VAR_rmse_with_global_avg = data.error(
error_fn=RMSE, predictions=VAR_predictions_with_global_avg)
print('VAR model RMSE with global average imputation (1 yr):',
VAR_rmse_with_global_avg)
VAR_predictions_with_global_avg_5 = data.predictions(
model_name=var,
lookback=4,
preprocessed_data=X_with_global_avg_5,
forward=5)
VAR_rmse_with_global_avg_5 = data.error(
error_fn=RMSE, predictions=VAR_predictions_with_global_avg_5)
print('VAR model RMSE with global average imputation (5 yr):',
VAR_rmse_with_global_avg_5)
VAR_rmse_with_global_avg_avg = (VAR_rmse_with_global_avg +
VAR_rmse_with_global_avg_5) / 2
print('VAR model RMSE with global average imputation (avg):',
VAR_rmse_with_global_avg_avg)
print()
# Arima with continent avg interpolation
arima_predictions_with_cont_avg = data.predictions(
model_name=arima, preprocessed_data=X_with_cont_avg)
arima_rmse_with_cont_avg = data.error(
error_fn=RMSE, predictions=arima_predictions_with_cont_avg)
print('ARIMA model RMSE with continent average imputation (1 yr):',
arima_rmse_with_cont_avg)
arima_predictions_with_cont_avg_5 = data.predictions(
model_name=arima, preprocessed_data=X_with_cont_avg_5, forward=5)
arima_rmse_with_cont_avg_5 = data.error(
error_fn=RMSE, predictions=arima_predictions_with_cont_avg_5)
print('ARIMA model RMSE with continent average imputation (5 yr):',
arima_rmse_with_cont_avg_5)
arima_rmse_with_cont_avg_avg = (arima_rmse_with_cont_avg +
arima_rmse_with_cont_avg_5) / 2
print('ARIMA model RMSE with continent average imputation (avg):',
arima_rmse_with_cont_avg_avg)
print()
# Arima with continent average and linear interpolation
arima_predictions_with_cont_avg_lin_interp = data.predictions(
model_name=arima, preprocessed_data=X_with_cont_avg_and_lin_interp)
arima_with_cont_avg_lin_interp_rmse = data.error(
error_fn=RMSE, predictions=arima_predictions_with_cont_avg_lin_interp)
print(
'Arima model RMSE with continent average imputation and linear interpolation (1yr):',
arima_with_cont_avg_lin_interp_rmse)
arima_predictions_with_cont_avg_lin_interp_5 = data.predictions(
model_name=arima,
preprocessed_data=X_with_cont_avg_and_lin_interp_5,
forward=5)
arima_with_cont_avg_lin_interp_rmse_5 = data.error(
error_fn=RMSE,
predictions=arima_predictions_with_cont_avg_lin_interp_5)
print(
'Arima model RMSE with continent average imputation and linear interpolation (5 yr):',
arima_with_cont_avg_lin_interp_rmse_5)
arima_predictions_with_cont_avg_lin_interp_rmse_avg = (
arima_with_cont_avg_lin_interp_rmse +
arima_with_cont_avg_lin_interp_rmse_5) / 2
print(
'Arima model RMSE with continent average imputation and linear interpolation (avg):',
arima_predictions_with_cont_avg_lin_interp_rmse_avg)
print()
def main():
num_args = len(sys.argv)
if num_args > 1:
if sys.argv[1] == 'arima_global_avg':
data = dataset.UNDevGoalsDataset()
X_submit = data.preprocess(
pp_fn=preprocess_for_submission_with_global_avg_and_lin_interp)
arima_predictions_with_global_avg_lin_interp = data.predictions(
model_name=arima, preprocessed_data=X_submit, lookback=4)
arima_predictions_with_global_avg_lin_interp_5 = data.predictions(
model_name=arima,
preprocessed_data=X_submit,
lookback=4,
forward=5)
one_yr_output = pd.Series(
arima_predictions_with_global_avg_lin_interp,
index=X_submit.index,
name='2008 [YR2008]')
five_yr_output = pd.Series(
arima_predictions_with_global_avg_lin_interp_5,
index=X_submit.index,
name='2012 [YR2012]')
out_df = pd.concat([one_yr_output, five_yr_output], axis=1)
out_df.to_csv('SubmissionRows.csv')
elif sys.argv[1] == 'arima_continent_avg':
data = dataset.UNDevGoalsDataset()
X_submit = data.preprocess(
pp_fn=preprocess_for_submission_with_cont_avg_and_lin_interp)
arima_predictions_with_cont_avg_lin_interp = data.predictions(
model_name=arima, preprocessed_data=X_submit, lookback=6)
arima_predictions_with_cont_avg_lin_interp_5 = data.predictions(
model_name=arima,
preprocessed_data=X_submit,
lookback=6,
forward=5)
one_yr_output = pd.Series(
arima_predictions_with_cont_avg_lin_interp,
index=X_submit.index,
name='2008 [YR2008]')
five_yr_output = pd.Series(
arima_predictions_with_cont_avg_lin_interp_5,
index=X_submit.index,
name='2012 [YR2012]')
out_df = pd.concat([one_yr_output, five_yr_output], axis=1)
out_df.to_csv('SubmissionRows.csv')
elif sys.argv[1] == 'arima_and_var':
data = dataset.UNDevGoalsDataset()
X_submit = data.preprocess(
pp_fn=preprocess_for_submission_with_global_avg_and_lin_interp)
arima_predictions_with_global_avg_lin_interp = data.predictions(
model_name=arima_and_var,
preprocessed_data=X_submit,
var_lookback=6,
forward=1,
do_arima=False)
arima_predictions_with_global_avg_lin_interp_5 = data.predictions(
model_name=arima_and_var,
preprocessed_data=X_submit,
arima_lookback=4,
forward=5,
do_arima=True)
one_yr_output = pd.Series(
arima_predictions_with_global_avg_lin_interp,
index=X_submit.index,
name='2008 [YR2008]')
five_yr_output = pd.Series(
arima_predictions_with_global_avg_lin_interp_5,
index=X_submit.index,
name='2012 [YR2012]')
out_df = pd.concat([one_yr_output, five_yr_output], axis=1)
out_df.to_csv('SubmissionRows.csv')
else:
raise Exception(
"You have to submit a valid type of submittable as a command line argument"
)