def predict(self, data, gap=0, predtill=1):
assert predtill - 1 <= gap
true = data[:, -predtill:, :]
pred = []
varm = VAR(data[0, :-gap - 1, :])
fitted = varm.fit()
pred = varm.predict(fitted.params,
start=data.shape[1] - gap - 2,
end=data.shape[1] - 1)[gap + 1 - predtill:gap + 1]
pred = np.expand_dims(np.array(pred), axis=0)
mae = np.mean(np.abs(pred - true))
mape = np.mean(np.abs(pred - true) / true) * 100
return mae, mape, pred
def var(data, gap=0, predtill=1):
assert predtill - 1 <= gap
true = data[:, -predtill:, :]
pred = []
# varm = VAR(data[0,:-1,:])
varm = VAR(data[0, :-gap - 1, :])
fitted = varm.fit()
# print("Lag", fitted.k_ar)
print("Coefficients", fitted.params.shape)
pred = varm.predict(fitted.params,
start=data.shape[1] - gap - 2,
end=data.shape[1] - 1)[gap + 1 - predtill:gap + 1]
print(pred.shape)
pred = np.expand_dims(np.array(pred), axis=0)
mae = np.mean(np.abs(pred - true))
mape = np.mean(np.abs(pred - true) / true) * 100
return mae, mape, pred
#1010025
all_data_scaled_sorted = pd.concat([all_data_scaled,labels],axis=1,sort=False).sort_values(['galaxy','galactic year'])[['galaxy','galactic year','y']]
#%%
X_train_with_labels = pd.concat([X_train,labels],axis=1,sort=False).sort_values(['galaxy','galactic year'])[['galaxy','galactic year','y']]
# %%
predicted_data = pd.DataFrame([])
for galaxy in tqdm_notebook(X_test['galaxy'].unique(), desc='Galaxy Loop'):
galaxy_data = X_train_with_labels[X_train_with_labels['galaxy']==galaxy]
galaxy_data_to_predict = X_test[X_test['galaxy']==galaxy][['galaxy','galactic year']]
model = GradientBoostingRegressor(n_estimators=300)
model.fit(np.array(galaxy_data['galactic year']).reshape(-1,1),galaxy_data['y'])
galaxy_data_to_predict['y'] = model.predict(np.array(galaxy_data_to_predict['galactic year']).reshape(-1,1))
# trend_line = np.poly1d(np.polyfit(galaxy_data['galactic year'],galaxy_data['y'],deg=4))
# galaxy_data_to_predict['y'] = galaxy_data_to_predict['galactic year'].apply(lambda x: trend_line(x))
eei_test = test_df[test_df['galaxy']==galaxy][['galactic year','existence expectancy index']]
eei_test.fillna(method='ffill',axis=0,inplace=True)
galaxy_data_to_predict['existence expectancy index'] = eei_test['existence expectancy index'].values
predicted_data = pd.concat([predicted_data,galaxy_data_to_predict[['y','existence expectancy index']]])
# data_to_plot = galaxy_data[['galactic year','y']]
# data_to_plot = pd.concat([data_to_plot,galaxy_data_to_predict[['galactic year','y']]]).sort_values('galactic year')
# plt.Line2D(xdata=data_to_plot['galactic year'],ydata=data_to_plot['y'])
# plt.show()
