def _evaluate_garch_model(self, train, test, split_time):
"""
Optimize Garch model
:param train: train part of the signal
:param test: test part of the signal
:param split_time: point when you start to predict
:return: variance volatility
"""
warnings.filterwarnings("ignore")
best_score = float(10000)
best_p = (1)
best_q = int(1)
p_values = range(1, 3)
q_values = range(1, 2)
for p in p_values:
for q in q_values:
try:
am = arch_model(train, vol='Garch', p=p, q=q)
res = am.fit(last_obs=split_time)
aic = SARIMAXResults.aic(res)
if abs(aic) < best_score:
best_score, best_p, best_q = aic, p, q
# print('GARCH : p=%i GARCH : q=%i AIC=%.3f' % (p, q, aic))
except:
continue
print('Best config GARCH model p=%i q=%i AIC=%.3f' %
(best_p, best_q, best_score))
am = arch_model(train, vol='Garch', p=best_p, q=best_q)
res = am.fit(last_obs=split_time)
forecast_analytic = res.forecast(horizon=len(test), method='analytic')
forecast_variance = forecast_analytic.variance[split_time:]
forecast_volatility = forecast_analytic.residual_variance[split_time:]
forecast_garch = forecast_variance.iloc[:, 0]
forecast_vol = forecast_volatility.iloc[:, 0]
return forecast_garch, forecast_vol
def preveProximasCincoSemanas():
loaded = SARIMAXResults.load('model.pkl')
hoje = date.today()
ultimaData = date(2020, 3, 15)
medicoesInicio = round((hoje-ultimaData).days/7)
numMedicoes=medicoesInicio+5
primeiraData='2020-03-22'
predicao=loaded.get_forecast(steps=numMedicoes)
predicao.predicted_mean
##index_date = pandas.date_range(primeiraData, periods = numMedicoes, freq = 'W')
index_date = pandas.bdate_range(primeiraData, periods = numMedicoes, freq = 'C', weekmask='Mon')
##forecast_series = pandas.Series(list(predicao.predicted_mean), index = index_date)
datasStr=numpy.datetime_as_string(index_date, unit='D')
print(numpy.column_stack((datasStr,list(predicao.predicted_mean))))
##conversaoEmLista = pandas.DataFrame(data=forecast_series).tail(5).to_numpy().tolist()
conversaoEmLista = pandas.DataFrame(numpy.column_stack((datasStr,list(predicao.predicted_mean)))).tail(5).to_numpy().tolist()
response = app.response_class(
response=json.dumps(conversaoEmLista),
status=200,
mimetype='application/json'
)
return response
def crearPrediccionMR(fichero_modelo, tipo):
loaded = SARIMAXResults.load(fichero_modelo)
DIAS_A_PREDECIR = 7
inicio = datetime.today().__format__('%Y-%m-%d') # Hoy
final = datetime.today()
for i in range(DIAS_A_PREDECIR):
final += timedelta(days=1)
final = final.__format__('%Y-%m-%d') # x DÍAS EN ADELANTE
inicio = datetime(2018, 3, 1, 0, 0,
0) # fuerzo un día en concreto because boobs
final = datetime(2018, 3, 7, 0, 0,
0) # fuerzo un día en concreto because boobs
predicciones = loaded.predict(
start=inicio, end=final, dynamic=False,
typ='levels').rename('SARIMA(2,1,1)(2,0,1,24) Predicciones')
ristra = pd.date_range(start=inicio, end=final, freq='h')
d = {'Fecha': ristra, tipo: predicciones}
df = pd.DataFrame(data=d)
ruta_fich = settings.MEDIA_ROOT + '\\prediccionesMR\\'
ruta_pred = ruta_fich + 'prediccion' + id_random_generator() + '.csv'
df.to_csv(ruta_pred)
return ruta_pred
def _evaluate_models(self, train, d, d_seasonal, WD):
best_score = float(10000)
best_cfg_arima = (1, d, 0)
best_cfg_seasonal = (1, d_seasonal, 0, WD)
p_values = range(1, 3)
q_values = range(1, 2)
for p in p_values:
for q in q_values:
for p_s in p_values:
for q_s in q_values:
order = (p, d, q)
seasonal_order = (p_s, d_seasonal, q_s, WD)
try:
model = SARIMAX(train,
order=order,
seasonal_order=seasonal_order)
md_fit = model.fit()
aic = SARIMAXResults.aic(md_fit)
if abs(aic) < best_score:
best_score, best_cfg_arima, best_cfg_seasonal = aic, order, seasonal_order
except:
continue
print('Best ARIMA%s Best Seasonal%s AIC=%.3f' %
(best_cfg_arima, best_cfg_seasonal, best_score))
return best_cfg_arima, best_cfg_seasonal, best_score
def predict_with_sarima(self, order=(1, 0, 0), seasonal_order=(1, 0, 0, 24)):
# seasonal_order=(1, 1, 1, 24),
saved_model = SARIMAXResults.load('{}{}{}.pkl'.format(self.model_dir, self.name, self.sarima_models_suffix))
data_used_for_prediction = self.train[-48:]
model = SARIMAX(endog=data_used_for_prediction, order=order, seasonal_order=seasonal_order)
fitted_model = model.filter(params=saved_model.params)
self.prediction = fitted_model.predict(start=len(data_used_for_prediction), end=len(data_used_for_prediction) + self.prediction_length - 1)
return self.prediction
def load(self, filepath):
'''
Loads a model from a pickle file
Arguments:
filepath - Path to the pickle file containing the model
'''
self.model = SARIMAXResults.load(filepath)
def crearPrediccionMRunico(modelo, tipo, rango):
"""
Crea las predicciones de precios para cada tarifa del mercado regulado.
:param modelo: ruta del modelo a partir del cual crear la predicción.
:param tipo: tarifa en concreto de la que se creará la predicción
:param rango: franja temporal en la que crear la predicción.
:return: ruta del fichero que contiene la predicción creada.
"""
loaded = SARIMAXResults.load(modelo)
inicio = pd.to_datetime(rango.get('principio'))
final = pd.to_datetime(rango.get('final'))
if tipo == 'TPD':
predicciones = loaded.predict(
start=inicio, end=final, dynamic=False,
typ='levels').rename('SARIMA(3,1,1)(2,0,1,24) Predicciones')
ristra = pd.date_range(start=inicio, end=final, freq='h')
d = {'Fecha': ristra, tipo: predicciones}
df = pd.DataFrame(data=d)
ruta_fich = settings.MEDIA_ROOT + '\\prediccionesMR\\'
ruta_pred = ruta_fich + 'prediccionMR_' + tipo + '_' + id_random_generator(
) + '.csv'
df.to_csv(ruta_pred)
return ruta_pred
elif tipo == 'EDP':
predicciones = loaded.predict(
start=inicio, end=final, dynamic=False,
typ='levels').rename('SARIMA(1,0,0)(2,0,0,24) Predicciones')
ristra = pd.date_range(start=inicio, end=final, freq='h')
d = {'Fecha': ristra, tipo: predicciones}
df = pd.DataFrame(data=d)
ruta_fich = settings.MEDIA_ROOT + '\\prediccionesMR\\'
ruta_pred = ruta_fich + 'prediccionMR_' + tipo + '_' + id_random_generator(
) + '.csv'
df.to_csv(ruta_pred)
return ruta_pred
elif tipo == 'VE':
predicciones = loaded.predict(
start=inicio, end=final, dynamic=False,
typ='levels').rename('SARIMA(2,0,0)(2,0,0,24) Predicciones')
ristra = pd.date_range(start=inicio, end=final, freq='h')
d = {'Fecha': ristra, tipo: predicciones}
df = pd.DataFrame(data=d)
ruta_fich = settings.MEDIA_ROOT + '\\prediccionesMR\\'
ruta_pred = ruta_fich + 'prediccionMR_' + tipo + '_' + id_random_generator(
) + '.csv'
df.to_csv(ruta_pred)
return ruta_pred
else:
# Creación de modelos para tarifas que no se tienen en cuenta: NOPE.
print('Creación de Predicciones unicas mercado regulado: error')
return False
def load_model(self, path, filename, version):
"""
:param path: Full or relative path for the model file
:param filename: Name of the file
:param version: Version number
:return: unpickled instance of the SARIMAX model
"""
return SARIMAXResults.load(
self.__generate_pkl_filename(path, filename, version))
def load(self, model_path, scaler_path=None):
'''
Loads the model from disk
:param model_path: Path to saved model
'''
if os.path.splitext(model_path)[1] != '.pkl':
raise Exception('Model file path for ' + self.name +
' must have ".pkl" extension.')
self.model = SARIMAXResults.load(model_path)
return
def plot_sarima_forecast(idf_df):
idf_df[c.EnergyConso.CONSUMPTION] = idf_df[c.EnergyConso.CONSUMPTION].fillna(
idf_df[c.EnergyConso.CONSUMPTION].mean())
idf_train = idf_df[md.END_TRAIN_DATE - timedelta(days=365):md.END_TRAIN_DATE]
idf_test = idf_df[md.END_TRAIN_DATE:md.END_TRAIN_DATE + timedelta(hours=md.NB_HOURS_PRED)]
best_model = SARIMAXResults.load(os.path.join(SARIMA_MODELS_PATH, "best_model.pkl"))
plot_sarima(idf_test, idf_train, best_model, md.NB_HOURS_PRED, s=1, d=1)
def price_predict():
product_name = request.args.get("product_name")
num_steps = int(request.args.get("num_steps"))
loaded = SARIMAXResults.load(product_name + ".pkl")
pred_uc = loaded.get_forecast(steps=num_steps)
pred_ci = pred_uc.conf_int()
print(pred_ci)
bytes_obj = get_plot(pred_uc, pred_ci, product_name)
#return send_file(bytes_obj,attachment_filename='plot.png',mimetype='image/png')
return "Done"
def crearPrediccion(fichero):
"""
Creación de la predicción de consumo para un Inmueble.
:param fichero: ruta del modelo predictivo a emplear.
:return: ruta del fichero con la predicción.
"""
print('Has entrado a la función')
# print(File(fichero))
print(fichero)
print('Cargamos el fichero')
loaded = SARIMAXResults.load(fichero)
print('Ya está cargado')
DIAS_A_PREDECIR = 7
inicio = datetime.today().__format__('%Y-%m-%d') # Hoy
final = datetime.today()
for i in range(DIAS_A_PREDECIR):
final += timedelta(days=1)
final = final.__format__('%Y-%m-%d') # x DÍAS EN ADELANTE
inicio = datetime(2018, 3, 1, 0, 0,
0) # fuerzo un día en concreto because boobs
final = datetime(2018, 3, 7, 0, 0,
0) # fuerzo un día en concreto because boobs
# print(inicio)
# print(final)
print('Fechas elegidas. Procedo a predecir.')
predicciones = loaded.predict(
start=inicio, end=final, dynamic=False,
typ='levels').rename('SARIMA(1,1,1)(2,0,3,24) Predicciones')
print('Predicciones hechas.')
#Creo un dataframe con los valores de las predicciones
ristra = pd.date_range(start=inicio, end=final, freq='h')
d = {'Fecha': ristra, 'Consumo_kWh': predicciones}
df = pd.DataFrame(data=d)
ruta_fich = settings.MEDIA_ROOT + '\\predicciones\\'
ruta_pred = ruta_fich + 'prediccion' + id_random_generator() + '.csv'
df.to_csv(ruta_pred)
# prediccion = predicciones.to_csv('LaPrediccion.csv')
print('Guardado en csv. Te lo mando.')
# return predicciones
return ruta_pred
def forecast_sarima(df, n_history, pred_horizon):
start = len(df[:-pred_horizon]) - 1
end = start + pred_horizon - 1
starttime = time.time()
order = (1, 3, 0)
seasonal_order = (1, 1, 1, n_history) #last parameter should be 52 for weekly saisonality
train_sarima(df, n_history = n_history, pred_horizon=pred_horizon, order=order, seasonal_order=seasonal_order)
print("training sarima took ", time.time() - starttime)
model = SARIMAXResults.load('sarima.pkl')
predictions = model.predict(start=start, end=end, dynamic=False).rename('SARIMAX Predictions' % n_history)
df_forecast = pd.DataFrame(predictions.values, columns=['SARIMA Predictions (1, 3, 0)(1, 1, 1, n_history) ' % n_history])
return generate_index_forecast(df, df_forecast, pred_horizon=pred_horizon)
def predict_with_trained_SARIMA_model(self, df, game_id, dep_var, indep_var, X_out_of_sample, target, type, model_type):
# might be mistakes here!
save_path = util.build_model_save_path(game_id, target, type, model_type)
if not util.check_for_model_existence(save_path):
print("Cannot predict for {}. No model has been trained yet.".format(dep_var))
return
saved_model = SARIMAXResults.load(save_path)
Y = df[dep_var]
X = df[indep_var]
X_out_of_sample.index = range(max(Y.index) + 1, max(Y.index) + 24 + 1)
latest_index = max(Y.index)
latest_timeslot = df.iloc[latest_index]['timeslot']
if self.seasonal_order is None:
model = SARIMAX(endog=Y, exog=X, order=self.order)
else:
model = SARIMAX(endog=Y, exog=X, order=self.order, seasonal_order=self.seasonal_order)
fitted_model = model.filter(params=saved_model.params)
prediction = fitted_model.predict(exog=X_out_of_sample, start=latest_index + 1,
end=latest_index + 1 + self.forecast_length - 1)
df_prediction = pd.DataFrame(
{'target_timeslot': range(latest_timeslot + 1, latest_timeslot + 1 + self.forecast_length),
'prediction': prediction})
# set lower and upper bound for prediction
tolerance = 0.2
prediction_upper_bound = max(Y) * (1 + tolerance) if max(Y) > 0 else max(Y) * (1 - tolerance)
prediction_lower_bound = min(Y) * (1 - tolerance) if min(Y) > 0 else min(Y) * (1 + tolerance)
# restrict prediction in lower and upper bound
df_prediction['prediction'] = df_prediction['prediction'].apply(
lambda x: max([min([x, prediction_upper_bound]), prediction_lower_bound]))
df_prediction['prediction_timeslot'] = latest_timeslot
df_prediction['proximity'] = df_prediction['target_timeslot'] - df_prediction['prediction_timeslot']
df_prediction['game_id'] = game_id
df_prediction['target'] = target
df_prediction['type'] = type
return df_prediction
def fit_predict(self, df):
""" Responsible for the train/inference pipeline of the model
Args:
df: the dataframe containing the time series data
"""
logger.info("SARIMAX Modeling")
self.df = df
self.df.index = pd.date_range(df.index[0], df.index[-1], freq='H')
# Log Transform to battle Heteroscedasticity
self.data = np.log(self.df['requests'])
# Train/Test Split
#train, test, exog_train, exog_test = self.train_test_split(self.df, data)
self.train_test_split()
if not self.trainmodel:
logger.info("\tLoading Model from Disk...")
self.result = SARIMAXResults.load(self.modelpath)
else:
self.train_sarimax()
self.predict_sarimax()
from statsmodels.tsa.statespace.sarimax import SARIMAXResults
import pickle
forecast_model_train = SARIMAXResults.load('forecast_model_train.pkl')
# Make a forecast for the next 6 periods after the last period in the training set
future = forecast_model_train.forecast(steps=5)
print('Train forecats...')
print(future)
# Make a forecast for the next 6 periods after the last period in the test set
forecast_model_test = SARIMAXResults.load('forecast_model_test.pkl')
future = forecast_model_test.forecast(steps=5)
print('\nTest forecasts')
print(future)
concentrate_scale=False,
trend_offset=1,
use_exact_diffuse=False,
dates=None,
freq=None,
missing='none'
)
res = mod.fit(disp=False)
if SAVE_MODEL:
res.save('save/prediction_model.pkl')
else:
res = SARIMAXResults.load('save/prediction_model.pkl')
# ------------------------------ Prediction -----------------------------------
def forecast_two_next_weeks(two_last_weeks, res, length):
"""Predicts the two next weeks of a time series from the last two weeks using
the SARIMAX method above
Arguments:
two_last_weeks {float tab} -- Two last weeks of the time series from which we predict
res {statsmodels.tsa.statespace.sarimax.SARIMAXResults} -- prediction model (defined above)
length {int} -- length of the prediction
Returns:
float np.array -- forecast of the next values of the time series
def sarima_test():
data = pd.read_csv('../data/CGMData.csv', header=None)
cgm, bolus = load_data(path=os.path.abspath('..'))
data.reindex(index=data.index[::-1])
t=0
for i in range(len(data.columns)):
data[i][0] = t
t+=5
data.iloc[1] = data.iloc[1].interpolate()
data = data.T
data = data[:len(bolus)]
data.columns=["time","CGMreading"]
y = data.set_index(["time"])
results = SARIMAXResults.load(os.path.join('..', 'model', "mc_sarima.pkl"))
classification_model = pickle.load(open(os.path.join('..', 'model', 'classifer_model.pkl'), 'rb'))
pred = results.get_prediction(start=26255, dynamic=False)
pred_ci = pred.conf_int()
ax = y.plot(label='observed')
pred.predicted_mean.plot(ax=ax, label='One-step ahead Forecast', alpha=.7, figsize=(14, 4))
ax.fill_between(pred_ci.index,
pred_ci.iloc[:, 0],
pred_ci.iloc[:, 1], color='k', alpha=.2)
ax.set_xlabel('Date')
ax.set_ylabel('Retail_sold')
plt.legend()
plt.show()
y_forecasted = pred.predicted_mean
y_truth = y[131275:]['CGMreading']
mse = ((y_forecasted - y_truth) ** 2).mean()
print('The Mean Squared Error is {}'.format(round(mse, 2)))
print('The Root Mean Squared Error is {}'.format(round(np.sqrt(mse), 2)))
testPredict = []
for i in range(0,(len(pred.predicted_mean)-6)):
temp_predict = []
temp_predict.append(pred.predicted_mean.iloc[i])
temp_predict.append(pred.predicted_mean.iloc[i+1])
temp_predict.append(pred.predicted_mean.iloc[i+2])
temp_predict.append(pred.predicted_mean.iloc[i+3])
temp_predict.append(pred.predicted_mean.iloc[i+4])
temp_predict.append(pred.predicted_mean.iloc[i+5])
testPredict.append(temp_predict)
testPredict = np.array(testPredict)
test = cgm[26249:(len(cgm))]
test = test.reshape(-1, 1)
test = regression_scaler.fit_transform(test)
testX, testY = create_dataset_multi_feature(test, 6)
_, _, _, bolus_test, classification_scaler = classification_split(cgm, bolus)
bolus_test = bolus_test[6 - 1:testX.shape[0] + 6 - 2]
predicted_labels = []
regression_predictions = copy.deepcopy(testPredict)
classification_train_data = classification_scaler.transform(regression_predictions)
prediction = classification_model.predict(classification_train_data)
predicted_labels.extend(prediction)
print(classification_report(bolus_test, predicted_labels))
def home():
fcast = ""
if request.method == 'POST':
method = request.form['forecast']
# Forecasting using saved ARMA model
if method == "arma":
result = SARIMAXResults.load('model/arma_model.pkl')
forecast_values = result.get_forecast(steps=test.shape[0])
forecast_values_mean = forecast_values.predicted_mean
conf_interval = forecast_values.conf_int()
arma_forecast_df = pd.DataFrame({
'Date':
test.index,
'Views':
forecast_values.predicted_mean,
'lower_views':
conf_interval['lower Views'].values,
'upper_views':
conf_interval['upper Views'].values
})
arma_forecast_df = arma_forecast_df.set_index('Date')
fig, ax = plt.subplots(figsize=(15, 4))
test.rename(columns={
'Views': 'Actual value'
}).plot(ax=ax, color='blue')
arma_forecast_df[['Views']].rename(columns={
'Views': 'Forecast'
}).plot(ax=ax, label='Forecast', color='red')
plt.fill_between(arma_forecast_df.index, \
arma_forecast_df.lower_views, \
arma_forecast_df.upper_views, \
color='pink', alpha=0.5)
plt.xlabel('Date')
plt.ylabel('Views')
plt.legend(loc='best')
new_arma_plot = "arma_plot_" + str(time.time()) + ".png"
for filename in os.listdir('static/'):
if filename.startswith('arma_plot_'):
os.remove('static/' + filename)
plt.savefig('static/' + new_arma_plot)
return render_template('index.html',
forecast='ARMA',
fcast='static/' + new_arma_plot)
# Forecasting using saved ARIMA model
elif method == "arima":
arima_result = SARIMAXResults.load('model/arima_model.pkl')
arima_forecast_values = arima_result.get_forecast(
steps=test.shape[0])
arima_forecast_mean = arima_forecast_values.predicted_mean
arima_conf_interval = arima_forecast_values.conf_int()
arima_forecast_df = pd.DataFrame({
'Date':
test.index,
'Views':
arima_forecast_values.predicted_mean,
'lower_views':
arima_conf_interval['lower Views'].values,
'upper_views':
arima_conf_interval['upper Views'].values
})
arima_forecast_df = arima_forecast_df.set_index('Date')
fig, ax = plt.subplots(figsize=(15, 4))
test.rename(columns={
'Views': 'Actual value'
}).plot(ax=ax, color='blue')
arima_forecast_df[['Views']].rename(columns={
'Views': 'Forecast'
}).plot(ax=ax, label='Forecast', color='red')
plt.fill_between(arima_forecast_df.index, \
arima_forecast_df.lower_views, \
arima_forecast_df.upper_views, \
color='pink', alpha=0.5)
plt.xlabel('Date')
plt.ylabel('Views')
plt.legend(loc='best')
new_arima_plot = "arima_plot_" + str(time.time()) + ".png"
for filename in os.listdir('static/'):
if filename.startswith('arima_plot_'):
os.remove('static/' + filename)
plt.savefig('static/' + new_arima_plot)
return render_template('index.html',
forecast='ARIMA',
fcast='static/' + new_arima_plot)
# Forecasting using saved Exponential Smoothing model
elif method == 'exp':
exp_model = pickle.load(open('model/exp_smoothing_model.pkl',
'rb'))
exp_smoothing_result = exp_model.fit(smoothing_level=0.5,
optimized=True)
test.index = pd.DatetimeIndex(test.index)
exp_smoothing_forecast = exp_smoothing_result.forecast(
test.shape[0])
exp_smoothing_forecast = exp_smoothing_forecast.reset_index(
).rename(columns={
'index': 'Date',
0: 'Views'
}).set_index('Date')
fig, ax = plt.subplots(figsize=(15, 4))
test.rename(columns={
'Views': 'Actual data'
}).plot(ax=ax, color='blue')
exp_smoothing_forecast.rename(columns={
'Views': 'Forecast'
}).plot(ax=ax, color='red')
plt.xlabel('Date')
plt.ylabel('Views')
plt.legend(loc='best')
new_exp_plot = "exp_plot_" + str(time.time()) + ".png"
for filename in os.listdir('static/'):
if filename.startswith('exp_plot_'):
os.remove('static/' + filename)
plt.savefig('static/' + new_exp_plot)
return render_template('index.html',
forecast='Exponential Smoothing',
fcast='static/' + new_exp_plot)
# Forecasting using saved Prophet model
elif method == 'prophet':
prophet_model = pickle.load(open('model/prophet_model.pkl', 'rb'))
test.index = pd.DatetimeIndex(test.index)
future = prophet_model.make_future_dataframe(periods=test.shape[0])
prophet_forecast = prophet_model.predict(future)
forecast_prophet = prophet_forecast[[
'ds', 'yhat_lower', 'yhat_upper', 'yhat'
]][-test.shape[0]:]
forecast_prophet = forecast_prophet.set_index('ds')
fig, ax = plt.subplots(figsize=(15, 4))
pd.plotting.register_matplotlib_converters()
test.rename(columns={
'Views': 'Actual data'
}).plot(ax=ax, color='blue')
forecast_prophet.rename(columns={'yhat': 'Forecast'})[[
'Forecast'
]].plot(ax=ax, color='red')
plt.fill_between(forecast_prophet.index,
forecast_prophet['yhat_lower'],
forecast_prophet['yhat_upper'],
color='pink',
alpha=0.5)
plt.xlabel('Date')
plt.ylabel('Views')
plt.legend(loc='best')
new_prophet_plot = "prophet_plot_" + str(time.time()) + ".png"
for filename in os.listdir('static/'):
if filename.startswith('prophet_plot_'):
os.remove('static/' + filename)
plt.savefig('static/' + new_prophet_plot)
return render_template('index.html',
forecast='Prophet',
fcast='static/' + new_prophet_plot)
# Forecasting using saved AutoARIMA model
elif method == 'auto_arima':
auto_arima_result = SARIMAXResults.load(
'model/auto_arima_model.pkl')
auto_arima_forecast = auto_arima_result.predict(
n_periods=test.shape[0])
auto_arima_forecast = pd.DataFrame(auto_arima_forecast,
index=test.index,
columns=['Forecast'])
fig, ax = plt.subplots(figsize=(15, 4))
test.rename(columns={
'Views': 'Actual value'
}).plot(ax=ax, color='blue')
auto_arima_forecast[['Forecast']].plot(ax=ax,
label='Forecast',
color='red')
ax.set_xlabel('Date')
ax.set_ylabel('Views')
plt.legend(loc='best')
new_auto_arima_plot = "auto_arima_plot_" + str(
time.time()) + ".png"
for filename in os.listdir('static/'):
if filename.startswith('auto_arima_plot_'):
os.remove('static/' + filename)
plt.savefig('static/' + new_auto_arima_plot)
return render_template('index.html',
forecast='Auto-arima',
fcast='static/' + new_auto_arima_plot)
return render_template('index.html', fcast=fcast)
def test(csv_file, pkl_file, power_test, amb_test):
power_path = str(power_test) # "power.csv"
power = pd.read_csv(power_path,
index_col=[0],
parse_dates=True,
names=["dt", "power"])
# Read Ambient Temperature Data
temp_path = str(amb_test) # "amb_temp.csv"
temp = pd.read_csv(temp_path,
index_col=[0],
parse_dates=True,
names=["dt", "temp"])
df = pd.DataFrame()
df = power.resample('15T').mean()
df["temp"] = temp.resample('15T').mean()
new_predicted = []
output = []
important = pd.read_csv("csv_file")
H = important[0]
L = important[1]
coef = important[2]
intercept = important[3]
load_ = str(pkl_file)
load_model = sresults.load(load_)
start = df.index[0]
end = df.index[-1]
predicted = loaded.predict(start, end).values
actual = power.power.values
df["pred"] = predicted
df["epsilon"] = (df["power"] - df["pred"]) / df["pred"]
epsilon = df.epsilon.values
power_val = df.power.values
temp_val = df.temp.values
for i in range(len(epsilon)):
new_p = 0
if epsilon[i] > H:
output.append("Positive Outlier")
new_p = coef * temp_val[i] + intercept
new_predicted.append(new_p)
elif epsilon[i] < L:
output.append("Negetive Outlier")
new_p = coef * temp_val[i] + intercept
new_predicted.append(new_p)
else:
outout.append("No Outlier")
new_predicted.append(power_val[i])
new_df = pd.DataFrame(new_predicted)
new_df.columns = ["Result"]
new_df["Remark"] = output
new_df.to_csv("Final_result.csv", index=False)
return
import matplotlib.pyplot as plt
from pandas import datetime
import calendar
import seaborn as sns
from statsmodels.tsa.ar_model import AR
from statsmodels.tsa.arima_model import ARMA
from statsmodels.tsa.arima_model import ARIMA
from statsmodels.tsa.statespace.sarimax import SARIMAX
from statsmodels.tsa.statespace.sarimax import SARIMAXResults
def parser(x):
return datetime.strptime(x, '%d-%m-%Y')
loaded = SARIMAXResults.load('model1.pkl')
test = pd.read_csv('validation_data.csv',
header=0,
parse_dates=[1],
index_col=1,
date_parser=parser)
series = pd.read_csv('trenddata.csv',
header=0,
parse_dates=[1],
index_col=1,
date_parser=parser)
print(test.index)
test = test[['Sales']]
test = test[:22]
print(test)
train = series