def plot_autocorrelation(self, save=False, save_name=None, inline=False):
"""
Plots the autocorrelation of the residuals, options to save the figure with a designated filename are available. If no filename is given, the classifier will generate one.
Args:
x_data : list - default is test data, can choose subset from ["train", "test", "val"]
conf_level : float - confidence level in the coefficients is extracted from the regression results
save : bool - defines whether the figure is saved or just shown
save_name : string - filename if save is set to True
"""
try:
if self._model == None:
raise RuntimeError("No model trained. Please train a model.")
fig = plt.figure(figsize=(15, 6))
residuals = self.get_data(data='Y_train') - nn_out_to_list(
self._model.predict(
self.get_data(data='X_train')[self._current_features]))
autocorrelation_plot(residuals)
x_label_setter('Lag', plt.gca())
y_label_setter('Autocorrelation', plt.gca())
if not inline:
if (save_name == None and save):
save_string = 'SLP_autocorrelation' + str(
self.save_count) + '.png'
self.save_count += 1
else:
save_string = save_name
if save:
plt.savefig(save_string)
else:
plt.show()
except RuntimeError as re:
print(re.args[0])
def plot_autocorrelation(self, save=False, save_name=None, inline=False):
"""
Plots the autocorrelation of the residuals, options to save the figure with a designated filename are available. If no filename is given, the classifier will generate one.
Args:
x_data : list - default is test data, can choose subset from ["train", "test", "val"]
conf_level : float - confidence level in the coefficients is extracted from the regression results
save : bool - defines whether the figure is saved or just shown
save_name : string - filename if save is set to True
"""
try:
if self._was_regularised:
raise RuntimeError(
"Model was regularised, can't extract residuals.")
fig = plt.figure(figsize=(15, 6))
result = self.get_result()
autocorrelation_plot(result.resid)
x_label_setter('Lag', plt.gca())
y_label_setter('Autocorrelation', plt.gca())
if not inline:
if (save_name == None and save):
save_string = 'LinReg_autocorrelation' + str(
self.save_count) + '.png'
self.save_count += 1
else:
save_string = save_name
if save:
plt.savefig(save_string)
else:
plt.show()
except RuntimeError as re:
print(re.args[0])
def plot_residuals(self, save=False, save_name=None, inline=False):
"""
Plots the residuals against the input data to assess the distribution, options to save the figure with a designated filename are available. If no filename is given, the classifier will generate one.
Args:
x_data : list - default is test data, can choose subset from ["train", "test", "val"]
conf_level : float - confidence level in the coefficients is extracted from the regression results
save : bool - defines whether the figure is saved or just shown
save_name : string - filename if save is set to True
"""
try:
if self._model == None:
raise RuntimeError("No model trained, can't extract residuals.")
fig = plt.figure(figsize=(10,10))
residuals = self.get_data(data = 'Y_train') - nn_out_to_list(self._model.predict(self.get_data(data = 'X_train')[self._current_features]))
Y_train_data = self.get_data(data='Y_train')
sns.regplot(Y_train_data, residuals, lowess=True, line_kws={'color': 'red', 'lw': 3, 'alpha': 0.8})
x_label_setter('y training values', plt.gca())
y_label_setter('Residuals', plt.gca())
if not inline:
if (save_name == None and save):
save_string = 'SLP_residuals' + str(self.save_count) + '.png'
self.save_count += 1
else:
save_string = save_name
if save:
plt.savefig(save_string)
else:
plt.show()
except RuntimeError as re:
print(re.args[0])
def plot_training(self, save=False, save_name=None, _inline=False):
"""
Plots the evolution of the training and validation error as a function of the number of epochs.
Args:
save : bool - defines whether the figure is saved or just shown
save_name : string - filename if save is set to True
"""
try:
if self.history != None:
plt.figure(figsize=(15,8))
ax = plt.subplot()
plt.plot(self.history['loss'], label = 'Training Loss')
plt.plot(self.history['val_loss'], color='red', label = 'Validation Loss')
plt.legend()
x_label_setter('Epoch', plt.gca())
y_label_setter('Loss', plt.gca())
else:
raise ValueError("ERROR: No model trained. Please train a model.")
if not _inline:
if (save_name == None and save):
save_string = 'SLP_forecast' + str(self.save_count) + '.png'
self.save_count += 1
else:
save_string = save_name
if save:
plt.savefig(save_string)
else:
plt.show()
except ValueError as ve:
print(ve.args[0])
def plot_forecast(self,
x_data=["test"],
save=False,
save_name=None,
_inline=False):
"""
Plots the selected subset of data along with a pre selected confidence interval, options to save the figure with a designated filename are available. If no filename is given, the classifier will generate one.
Args:
x_data : list - default is test data, can choose subset from ["train", "test", "val"]
save : bool - defines whether the figure is saved or just shown
save_name : string - filename if save is set to True
"""
data = self.get_data()
temp_X = []
temp_Y = []
for _type in x_data:
temp_X.append(data['X_' + _type])
temp_Y.append(data['Y_' + _type])
X_data = pd.concat(temp_X)
Y_data = pd.concat(temp_Y)
try:
if self._model != None:
X_data = X_data[self._current_features]
Y_forecast = nn_out_to_list(self._model.predict(X_data))
# Y_data['Y_forecast'] = Y_forecast
# Y_forecast_df = Y_data['Y_forecast']
temp_df = pd.DataFrame({'Y_act': Y_data})
temp_df['Y_forecast'] = Y_forecast
plt.figure(figsize=(15, 8))
ax = plt.subplot()
plt.plot('Y_act', data=temp_df, color='red', label='Actuals')
plt.plot('Y_forecast',
data=temp_df,
color='blue',
label='Prediction')
plt.legend()
x_label_setter('Date', plt.gca())
y_label_setter('Data Value', plt.gca())
else:
raise ValueError(
"ERROR: No model trained. Please train a model.")
if not _inline:
if (save_name == None and save):
save_string = 'SLP_forecast' + str(
self.save_count) + '.png'
self.save_count += 1
else:
save_string = save_name
if save:
plt.savefig(save_string)
else:
plt.show()
except ValueError as ve:
print(ve.args[0])
def plot_forecast(self,
x_data=["test"],
conf_levl=0.05,
save=False,
save_name=None,
inline=False):
"""
Plots the selected subset of data along with a pre selected confidence interval, options to save the figure with a designated filename are available. If no filename is given, the classifier will generate one.
Args:
x_data : list - default is test data, can choose subset from ["train", "test", "val"]
conf_level : float - confidence level in the coefficients is extracted from the regression results
save : bool - defines whether the figure is saved or just shown
save_name : string - filename if save is set to True
"""
data = self.get_data()
temp_X = []
temp_Y = []
for _type in x_data:
temp_X.append(data['X_' + _type])
temp_Y.append(data['Y_' + _type])
X_data = pd.concat(temp_X)
Y_data = pd.concat(temp_Y)
if self.__result == None:
X_data = X_data[self._current_features]
Y_forecast = (X_data * self.get_latest_params()).sum(axis=1)
plt.figure(figsize=(15, 8))
ax = plt.subplot()
Y_data.plot(ax=ax, color='red', label='Actuals')
Y_forecast.plot(ax=ax, color='blue', label='Prediction')
plt.legend()
x_label_setter('Date', plt.gca())
y_label_setter('Data Value', plt.gca())
else:
conf_int = self.__result.conf_int(alpha=conf_levl)
forecast_lo = (X_data * conf_int[0]).sum(axis=1)
forecast_hi = (X_data * conf_int[1]).sum(axis=1)
Y_forecast = (X_data * self.get_latest_params()).sum(axis=1)
plt.figure(figsize=(15, 8))
ax = plt.subplot()
Y_data.plot(ax=ax, color='red', label='Actuals')
Y_forecast.plot(ax=ax, color='blue', label='Prediction')
plt.legend()
x_label_setter('Date', plt.gca())
y_label_setter('Data Value', plt.gca())
plt.fill_between(forecast_lo.index,
forecast_lo,
forecast_hi,
color='gray',
alpha=0.4)
if not inline:
if (save_name == None and save):
save_string = 'LinReg_forecast' + str(self.save_count) + '.png'
self.save_count += 1
else:
save_string = save_name
if save:
plt.savefig(save_string)
else:
plt.show()
