def fit_expsmooth(ts_train,
ts_test,
trend="additive",
seasonal="multiplicative",
s=None,
alpha=0.94,
figsize=(15, 10)):
## checks
check_seasonality = "Seasonal parameters: No Seasonality" if (
seasonal is None) & (s is None) else "Seasonal parameters: " + str(
seasonal) + " Seasonality every " + str(s) + " observations"
print(check_seasonality)
## train
#alpha = alpha if s is None else 2/(s+1)
model = smt.ExponentialSmoothing(
ts_train, trend=trend, seasonal=seasonal,
seasonal_periods=s).fit(smoothing_level=alpha)
dtf_train = ts_train.to_frame(name="ts")
dtf_train["model"] = model.fittedvalues
## test
dtf_test = ts_test.to_frame(name="ts")
dtf_test["forecast"] = model.predict(start=len(ts_train),
end=len(ts_train) + len(ts_test) - 1)
## evaluate
dtf = dtf_train.append(dtf_test)
dtf = utils_evaluate_forecast(dtf,
figsize=figsize,
title="Holt-Winters (" + str(alpha) + ")")
return dtf, model
def time_series_udf(data):
data.set_index('Date',inplace = True)
time_series_data = data['Weekly_Sales']
##the model
model_monthly = sm.ExponentialSmoothing(np.asarray(time_series_data),trend='add').fit()
##forecast values
forecast_values = pd.Series(model_monthly.forecast(13),name = 'fitted_values')
#forecast_values.show()
return pd.DataFrame({'Store': [str(data.Store.iloc[0])],'category': [str(data.category.iloc[0])] ,'weekly_forecast_1': [forecast_values[0]], 'weekly_forecast_2':[forecast_values[1]],'weekly_forecast_3': [forecast_values[2]],'weekly_forecast_4': [forecast_values[3]],'weekly_forecast_5': [forecast_values[4]],'weekly_forecast_6': [forecast_values[5]],'weekly_forecast_7': [forecast_values[6]],'weekly_forecast_8': [forecast_values[7]],'weekly_forecast_9': [forecast_values[8]],'weekly_forecast_10': [forecast_values[9]],'weekly_forecast_11': [forecast_values[10]],'weekly_forecast_12': [forecast_values[11]],'weekly_forecast_13': [forecast_values[12]] })
def holt_winters_time_series_udf(data):
data.set_index('Date', inplace=True)
time_series_data = data['Weekly_Sales']
##the model
model_monthly = sm.ExponentialSmoothing(np.asarray(time_series_data),
trend='add').fit()
##forecast values
forecast_values = pd.Series(model_monthly.forecast(2),
name='fitted_values')
return pd.DataFrame({
'Store': [str(data.Store.iloc[0])],
'Dept': [str(data.Dept.iloc[0])],
'weekly_forecast_1': [forecast_values[0]],
'weekly_forecast_2': [forecast_values[1]]
})
print("No standard input provided to Holt-Winters algorithm, exiting", file=sys.stderr)
sys.exit(1)
try:
algorithm_input = AlgorithmInput.from_json(stdin)
except JSONDecodeError as ex:
print("Invalid JSON provided: {0}, exiting".format(str(ex)), file=sys.stderr)
sys.exit(1)
except KeyError as ex:
print("Invalid JSON provided: missing {0}, exiting".format(str(ex)), file=sys.stderr)
sys.exit(1)
if len(algorithm_input.series) < 10 + 2 * (algorithm_input.seasonal_periods // 2):
print("Invalid data provided, must be at least 10 + 2 * (seasonal_periods // 2) observations, exiting",
file=sys.stderr)
sys.exit(1)
model = sm.ExponentialSmoothing(algorithm_input.series, trend=algorithm_input.trend, seasonal=algorithm_input.seasonal,
seasonal_periods=algorithm_input.seasonal_periods,
initialization_method=algorithm_input.initialization_method,
damped_trend=algorithm_input.damped_trend, initial_level=algorithm_input.initial_level,
initial_trend=algorithm_input.initial_trend,
initial_seasonal=algorithm_input.initial_seasonal)
fitted_model = model.fit(smoothing_level=algorithm_input.alpha, smoothing_trend=algorithm_input.beta,
smoothing_seasonal=algorithm_input.gamma)
# Predict the value one ahead
print(math.ceil(fitted_model.forecast(steps=1)[0]), end="")
import statsmodels.api as sm
import statsmodels.tsa.api as smt
import matplotlib.pyplot as plt
# dataset
target = sm.datasets.sunspots.load_pandas().data['SUNACTIVITY'] + 10
target.plot(marker='o', color='black', legend=True, figsize=(12, 5))
# Holt-Winter's TS(A,A)
model = smt.ExponentialSmoothing(target,
seasonal_periods=24,
trend='add',
seasonal='add',
damped_trend=False).fit(use_boxcox=True)
model.fittedvalues.plot(style='--',
color='blue',
label=r'$\alpha=%s$' %
model.model.params['smoothing_level'])
forecast = model.forecast(24).rename(r'$\alpha=%s$' %
model.model.params['smoothing_level'])
forecast.plot(marker='o', color='blue')
plt.legend()
plt.show()
# [smt.ExponentialSmoothing]
# trend :(str) “add”, “mul”, “additive”, “multiplicative”, None
# damped_trend :(bool)
# seasonal :(str) “add”, “mul”, “additive”, “multiplicative”, None
# initialization_method :(str) None, ‘estimated’, ‘heuristic’, ‘legacy-heuristic’, ‘known’
# initial_level :(float)
# *************************************************************************************************************************************************************************************************************************************************************** # * Tópico - Separação Treino e Validação * # *************************************************************************************************************************************************************************************************************************************************************** # Separação em conjuntos treino e validação treino = historico.drop(historico.tail(120).index, inplace=False) validacao = historico.tail(120) # *************************************************************************************************************************************************************************************************************************************************************** # * Tópico - ETS * # *************************************************************************************************************************************************************************************************************************************************************** ets_pesos = smt.ExponentialSmoothing(endog=treino['adj close'], seasonal_periods=5, trend='mul', seasonal='add', damped=True).fit(use_brute=True) ets_previsao = pd.Series([x for x in ets_pesos.forecast(120)], index=validacao.index) # *************************************************************************************************************************************************************************************************************************************************************** # * Tópico - * # *************************************************************************************************************************************************************************************************************************************************************** arima_pesos = smt.arima.ARIMA(endog=treino['adj close'], order=(1, 1, 1)).fit() temp = arima_pesos.forecast(120) forecast_previsao = pd.Series([x for x in arima_pesos.forecast(120)], index=validacao.index) # [0]= predições
