def fit(self, series: TimeSeries):
super().fit(series)
series = self.training_series
# if the model was initially created with `self.seasonal_periods=None`, make sure that
# the model will try to automatically infer the index, otherwise it should use the
# provided `seasonal_periods` value
seasonal_periods_param = (
None if self.infer_seasonal_periods else self.seasonal_periods
)
# set the seasonal periods paramter to a default value if it was not provided explicitly
# and if it cannot be inferred due to the lack of a datetime index
if self.seasonal_periods is None and series.has_range_index:
seasonal_periods_param = 12
hw_model = hw.ExponentialSmoothing(
series.values(),
trend=self.trend if self.trend is None else self.trend.value,
damped_trend=self.damped,
seasonal=self.seasonal if self.seasonal is None else self.seasonal.value,
seasonal_periods=seasonal_periods_param,
freq=series.freq if series.has_datetime_index else None,
dates=series.time_index if series.has_datetime_index else None,
)
hw_results = hw_model.fit(**self.fit_kwargs)
self.model = hw_results
if self.infer_seasonal_periods:
self.seasonal_periods = hw_model.seasonal_periods
return self
def fit(self, series: TimeSeries):
super().fit(series)
self.model = hw.ExponentialSmoothing(
series.values(),
trend=self.trend,
seasonal=self.seasonal,
seasonal_periods=self.seasonal_periods).fit()
def test_convergence_simple():
# issue 6883
gen = np.random.RandomState(0)
e = gen.standard_normal(12000)
y = e.copy()
for i in range(1, e.shape[0]):
y[i] = y[i - 1] - 0.2 * e[i - 1] + e[i]
y = y[200:]
mod = holtwinters.ExponentialSmoothing(y,
initialization_method="estimated")
res = mod.fit()
ets_res = ETSModel(y).fit()
# the smoothing level should be very similar, the initial state might be
# different as it doesn't influence the final result too much
assert_allclose(
res.params["smoothing_level"],
ets_res.smoothing_level,
rtol=1e-4,
atol=1e-4,
)
# the first few values are influenced by differences in initial state, so
# we don't test them here
assert_allclose(res.fittedvalues[10:],
ets_res.fittedvalues[10:],
rtol=1e-4,
atol=1e-4)
def holt_winters(ts, back_steps=10, degree=1, steps=1):
"""
Predicts next value using triple exponential smoothing
(holt-winters method).
Parameters:
-----------
ts: Array of floats
An array of times series data to be used for the polyfit regression
Returns:
--------
x : The predicted value from the holt-winters method.
"""
timeseries = np.array(list(ts.values()))
timeseries = timeseries[-back_steps:]
# exponential smoothing errors when there is only one datapoint
if len(timeseries) == 1:
timeseries = [np.inf, 0]
# exponential smoothing errors when there are five datapoints
# average is appended to the beginning of the timeseries for minimal impact
# https://github.com/statsmodels/statsmodels/issues/4878
elif len(timeseries) == 5:
timeseries = np.append(np.mean(timeseries), timeseries)
model = hw.ExponentialSmoothing(timeseries)
model_fit = model.fit()
x = model_fit.predict(len(timeseries), len(timeseries) + steps - 1)
return x[-1]
def holt_linear(train, l):
model = ets.ExponentialSmoothing(
train, trend='mul', damped=True
) # Use the multiplicative version, unless the data has been logged before.
fit1 = model.fit()
print('Holt Linear method summary\n', fit1.summary())
return fit1.predict(start=l.index[0], end=l.index[-1])
def holt_winter_seasonal(train, l):
model = ets.ExponentialSmoothing(
train, trend='mul', seasonal='mul'
) # Use the multiplicative version, unless the data has been logged before.
fit1 = model.fit()
# print('Holt Winter method summary\n',fit1.summary())
# return fit1.forecast(l)
return fit1.predict(start=l.index[0], end=l.index[-1])
def ses(y_train, y_test):
alpha = .5
model = ets.ExponentialSmoothing(y_train,
trend=None,
damped=False,
seasonal=None).fit(smoothing_level=alpha,
optimized=False)
preds = model.forecast(steps=len(y_test))
return preds
def get_params_model_from_pmml(ts_data, ts_model_obj):
params = dict()
exp_smooth_obj = ts_model_obj.get_ExponentialSmoothing()
level_obj = exp_smooth_obj.get_Level()
trend_obj = exp_smooth_obj.get_Trend_ExpoSmooth()
season_obj = exp_smooth_obj.get_Seasonality_ExpoSmooth()
params['smoothing_level'] = level_obj.get_alpha()
params['initial_level'] = level_obj.get_initialLevelValue()
if trend_obj:
params['smoothing_slope'] = trend_obj.get_gamma()
params['damping_slope'] = trend_obj.get_phi()
params['initial_slope'] = trend_obj.get_initialTrendValue()
trend_type = trend_obj.get_trend()
if trend_type == 'additive':
trend = 'add'
damped = False
elif trend_type == 'multiplicative':
trend = 'mul'
damped = False
elif trend_type == 'damped_additive':
trend = 'add'
damped = True
elif trend_type == 'damped_multiplicative':
trend = 'mul'
damped = True
elif trend_type == 'polynomial_exponential':
pass
else:
trend = None
damped = False
if season_obj:
params['smoothing_seasonal'] = season_obj.get_delta()
seasonal_periods = season_obj.get_Array().get_n()
params['initial_seasons'] = np.array(
season_obj.get_Array().get_valueOf_().strip().split(' '))
season_type = season_obj.get_type()
if season_type == 'additive':
seasonal = 'add'
elif season_type == 'multiplicative':
seasonal = 'mul'
else:
seasonal = None
seasonal_periods = None
# if ts_model_obj.get_Extension(): # if Extension elements exist in pmml file
# for extension in ts_model_obj.get_Extension():
# if extension.get_name() == 'initialLevel':
# params['initial_level'] = extension.get_value()
# if extension.get_name() == 'initialTrend':
# params['initial_slope'] = extension.get_value()
stsmdl = hw.ExponentialSmoothing(ts_data,
trend=trend,
seasonal=seasonal,
seasonal_periods=seasonal_periods,
damped=damped)
return params, stsmdl
def fit(self, series: TimeSeries):
super().fit(series)
hw_model = hw.ExponentialSmoothing(
series.values(),
trend=self.trend,
damped=self.damped,
seasonal=self.seasonal,
seasonal_periods=self.seasonal_periods)
hw_results = hw_model.fit(**self.fit_kwargs)
self.model = hw_results
def test_exponentialSmoothing_02(self):
ts_data = self.getData1()
f_name = 'exponential_smoothing2.pmml'
model_obj = hw.ExponentialSmoothing(ts_data,
trend='add',
damped=False,
seasonal='add',
seasonal_periods=2)
results_obj = model_obj.fit(optimized=True)
statsmodels_to_pmml(ts_data, model_obj, results_obj, f_name)
self.assertEqual(os.path.isfile(f_name), True)
def test_exponentialSmoothing_17(self):
ts_data = self.getData3()
f_name = 'exponential_smoothing17.pmml'
model_obj = hw.ExponentialSmoothing(ts_data,
trend=None,
damped=False,
seasonal=None,
seasonal_periods=None)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(results_obj, f_name)
self.assertEqual(os.path.isfile(f_name), True)
def fit(self, series: TimeSeries, component_index: Optional[int] = None):
super().fit(series, component_index)
series = self.training_series
hw_model = hw.ExponentialSmoothing(
series.values(),
trend=self.trend,
damped=self.damped,
seasonal=self.seasonal,
seasonal_periods=self.seasonal_periods)
hw_results = hw_model.fit(**self.fit_kwargs)
self.model = hw_results
def test_exponentialSmoothing_05(self):
ts_data = self.getData1()
f_name = 'exponential_smoothing5.pmml'
model_obj = hw.ExponentialSmoothing(ts_data,
trend='mul',
damped=True,
seasonal='add',
seasonal_periods=2)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(results_obj, f_name)
self.assertEqual(os.path.isfile(f_name), True)
def test_exponentialSmoothing_03(self):
# data with no trend and no seasonality
# Oil production in Saudi Arabia
data = [
446.6565, 454.4733, 455.663, 423.6322, 456.2713, 440.5881,
425.3325, 485.1494, 506.0482, 526.792, 514.2689, 494.211
]
index = pd.DatetimeIndex(start='1996', end='2008', freq='A')
ts_data = pd.Series(data, index)
ts_data.index.name = 'datetime_index'
ts_data.name = 'oil_production'
f_name = 'exponential_smoothing.pmml'
model_obj = hw.ExponentialSmoothing(ts_data,
trend=None,
damped=False,
seasonal=None,
seasonal_periods=None)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj, f_name)
self.assertEqual(os.path.isfile(f_name), True)
os.remove(f_name)
model_obj = hw.ExponentialSmoothing(ts_data,
trend=None,
damped=False,
seasonal=None,
seasonal_periods=2)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj, f_name)
self.assertEqual(os.path.isfile(f_name), True)
os.remove(f_name)
def call_holt_s(train_s,test_s,lag, a='',b='',c=''):
prediction = ets.ExponentialSmoothing(train_s[a], trend='additive', seasonal='additive', seasonal_periods=1440,damped_trend=True).fit()
forecast = prediction.forecast(steps=len(test_s[a]))
yt = prediction.fittedvalues
final_predn = dataframe_create(yt,train_s,b,a)
final_forec = dataframe_create(forecast,test_s,b,a)
plot_func(train_s[b],test_s[b],final_forec[b],train_s[a],test_s[a],final_forec[a],c)
residuals = residual_avg(train_s[a], final_predn[a])
forec_error = forecast_err(test_s[a], final_forec[a])
mse_p, mse_f = mse(residuals, forec_error)
var_p, var_f = var(residuals, forec_error)
residual_acf = acf_values(residuals, lag)
Q = Q_val(train_s, residual_acf)
acf_plot(residual_acf, c)
return prediction, final_forec, residuals, forec_error, mse_p, mse_f, var_p, var_f, residual_acf, Q
def holt_winters_trainer_full(df,
seasonal='additive',
trend='additive',
seasonal_periods=None):
# optimized ExponentialSmoothing on y_data
print('WARNING: This takes a long time. Recommend a coffee break.')
model = ets.ExponentialSmoothing(df,
seasonal=seasonal,
trend=trend,
seasonal_periods=seasonal_periods).fit()
def holt_winters_predict(h):
return model.forecast(h)
return holt_winters_predict
def holt_forecast(history):
# convert history into a univariate series
series = to_series(history)
# define the model
# model = ARIMA(series, order=(6,0,0))
model = ets.ExponentialSmoothing(series,
trend='add',
damped=True,
seasonal='add',
seasonal_periods=7)
# fit the model
model_fit = model.fit()
# make forecast
yhat = model_fit.predict(len(series), len(series) + 6)
return yhat
def decompose(self):
'''
Croston's decomposition
'''
z = ts.SimpleExpSmoothing(self.demandValues.to_numpy())
p = ts.ExponentialSmoothing(self.intervals.to_numpy())
# Demand level
z = z.fit(smoothing_level=self.alpha,
initial_level = self.demandValues[0],
optimized=False)
# Intervals
p = p.fit(smoothing_level=self.alpha,
initial_level = self.intervals[0],
optimized=False )
self.fittedvalues = z.fittedfcast
self.fittedIntervals = p.fittedfcast
self.fittedForecasts = self.fittedvalues/self.fittedIntervals
self.fcast = z.forecast(1)/p.forecast(1)
def transform(self, seriess, debug):
series = seriess[0]
pdseries = series.pdseries
damped_trend, seasonal_length, output = self.get_params()
if seasonal_length is not None and seasonal_length != '':
calc_seasonal_length = timedelta_to_period(seasonal_length,
series.step())
seasonal = 'add'
else:
calc_seasonal_length = None
seasonal = None
model = tsa_hw.ExponentialSmoothing(
pdseries,
trend='add',
damped_trend=damped_trend,
seasonal=seasonal,
seasonal_periods=calc_seasonal_length,
initialization_method='estimated',
use_boxcox=False)
model_fit = model.fit()
# Debug info
if debug:
debug_info = {
"summary": str(model_fit.summary()),
}
else:
debug_info = {}
result = None
if output == 'predicted':
result = model_fit.fittedvalues
elif output == 'resid':
result = model_fit.resid
else:
raise ValueError('Invalid output: ' + output)
return (result, debug_info)
def holt_winters(y_train, y_test):
holtt = ets.ExponentialSmoothing(y_train,
seasonal_periods=365,
trend=None,
damped=False,
seasonal='mul').fit()
holtf = holtt.forecast(steps=len(y_test))
error = y_test - holtf
holtf = pd.DataFrame(holtf).set_index(y_test.index)
fig, ax = plt.subplots()
#fig.set_figheight(20)
#fig.set_figwidth(50)
ax.plot(y_train, label="Train Data")
ax.plot(y_test, label="Test Data")
ax.plot(holtf, label="Holt-Winters")
plt.legend(loc='upper left')
plt.title('Holt-Winters Method - Flow')
plt.xlabel('Time (weekly)')
plt.ylabel('Weekly Mean Flow')
plt.show()
return error
def expo_smooth(timeseries):
ts = timeseries.copy()
ts.set_index('Date', inplace=True)
import statsmodels.tsa.holtwinters as ets
#Brown Simple Exponential Smoothing
model_es = ets.ExponentialSmoothing(ts.values,
trend='additive',
damped=False,
seasonal=None)
model_es_fit = model_es.fit()
ES_predictions = model_es_fit.forecast(steps=12)
ES_error = round(
sqrt(mean_squared_error(ts.INV_AMOUNT_USD[-12:], ES_predictions)))
predictions['Double Exponential Smoothing Model' + ', ' + 'Error(RMSE):' +
str(ES_error)] = ES_predictions
#MA model
#def MA(timeseries):
#global MA_predictions
#timeseries.set_index('Date', inplace=True)
#timeseries_diff = timeseries.diff(periods=1)
#integrated od order 1 (d =1)
#timeseries_diff = timeseries_diff[1:]
#from statsmodels.graphics.tsaplots import plot_acf
#plot_acf(timeseries)
#plot_acf(timeseries_diff)
#timeseries_diff.plot()
#from statsmodels.tsa.arima_model import ARMA
#model_ma = ARMA(timeseries.values, order=(0,1))
#model_ma_fit = model_ma.fit()
#MA_predictions = model_ma_fit.predict(start=24,end=36)
#MA_predictions = MA_predictions[1:]
#MA_predictions = np.array(MA_predictions)
#error(timeseries.INV_AMOUNT_USD[-12:],AR_predictions)
#create_df(AR_predictions)
import matplotlib.pyplot as plt
import statsmodels.tsa.holtwinters as ets
df = pd.read_csv('ExponentialSmoothing.csv',
index_col='Date',
parse_dates=True)
#print(df)
#Defining the range of the training and testing
spyt = df[:'2020-05-01']
spyf = df['2020-05-02':]
#print(spyt)
#print(spyf)
brownt = ets.ExponentialSmoothing(spyt,
trend=None,
damped=False,
seasonal=None).fit()
brownf = brownt.forecast(steps=len(spyf))
brownf = pd.DataFrame(brownf).set_index(spyf.index)
fig1, ax = plt.subplots()
ax.plot(spyt, label='spyt')
ax.plot(spyf, label='spyf')
ax.plot(brownf, label='brownf')
plt.ylabel('Comments')
plt.xlabel('Timestamp')
plt.show()
# %%
#ExponentialSmoothing
import pandas as pd
def forecast(ts,
alpha=None,
beta=None,
gamma=None,
initial_level=None,
initial_slope=None,
initial_seasons=None,
trend=None,
seasonal=None,
seasonal_periods=None,
debug=False,
use_boxcox=False):
"""Forecasting using exponential smoothing
Usage:
forecast(ts=ts)
"""
smoothing_level = None
smoothing_slope = None
smoothing_seasonal = None
try:
if alpha and (0 <= alpha <= 1):
smoothing_level = alpha
if beta and (0 <= beta <= 1):
smoothing_slope = beta
# if beta is set but trend is None, trend is set equal to add
if smoothing_slope and not trend:
trend = 'add'
if gamma and (0 <= gamma <= 1):
smoothing_seasonal = gamma
if smoothing_seasonal and not seasonal:
seasonal = 'mul'
except TypeError:
print('ERRORE: alpha must be between 0 and 1')
return np.nan
# if trend:
# trend = 'add'
# else:
# trend = None
# if seasonal:
# seasonal = 'add'
# else:
# seasonal = None
if seasonal and not seasonal_periods:
print(
'ERROR: Please, provide the number of periods in a complete seasonal cycle'
)
return np.nan
model = holtwinters.ExponentialSmoothing(
endog=pd.DataFrame(data=ts),
trend=trend,
seasonal=seasonal,
seasonal_periods=seasonal_periods).fit(
smoothing_level=smoothing_level,
smoothing_slope=smoothing_slope,
smoothing_seasonal=smoothing_seasonal,
initial_level=initial_level,
initial_slope=initial_slope,
use_boxcox=use_boxcox)
model.predict(start=1, end=len(ts))
if debug:
print(model.model.params)
# print('TS')
# print(ts)
# print('FORECAST')
# print(model.fittedfcast)
return model.fittedfcast
def simple_forecast_ts(train, test, method, period):
"""
:param ytrain: training data
:param ytest: testing data
:param method: simple forecast method
:param period: only use for Holt-Winter Additive method to define the seasonal_periods
:return: dataframes containing prediction/forecast, errors and square errors
"""
T = len(train) # number of observations in the training set
h = len(
test) # number of observations to be forecast4ed in the testing set
ytrain = train.values
ytest = test.values
xtrain = train.index
xtest = test.index
ytrain_hat = [] # prediction
ytest_hat = [] # forecast
etrain = [] # prediction error
etest = [] # forecast error
setrain = [] # square of prediction error
setest = [] # square of forecast error
if method == 'Average': # using average method
for i in range(0, 1):
ytrain_hat.append(np.nan)
etrain.append(np.nan)
setrain.append(np.nan)
for i in range(1, T):
prediction = np.mean(ytrain[0:i])
ytrain_hat.append(prediction)
error = ytrain[i] - prediction
etrain.append(error)
setrain.append(error**2)
for i in range(0, h):
forecast = np.mean(ytrain)
ytest_hat.append(forecast)
error = ytest[i] - forecast
etest.append(error)
setest.append(error**2)
elif method == 'Naive': # naive method
for i in range(0, 1):
ytrain_hat.append(np.nan)
etrain.append(np.nan)
setrain.append(np.nan)
for i in range(1, T):
prediction = ytrain[i - 1]
ytrain_hat.append(prediction)
error = ytrain[i] - prediction
etrain.append(error)
setrain.append(error**2)
for i in range(0, h):
forecast = ytrain[T - 1]
ytest_hat.append(forecast)
error = ytest[i] - forecast
etest.append(error)
setest.append(error**2)
elif method == 'Drift': # drift method
for i in range(0, 2):
ytrain_hat.append(np.nan)
etrain.append(np.nan)
setrain.append(np.nan)
for i in range(2, T):
prediction = ytrain[i - 1] + (ytrain[i - 1] - ytrain[0]) / (i - 1)
ytrain_hat.append(prediction)
error = ytrain[i] - prediction
etrain.append(error)
setrain.append(error**2)
for i in range(0, h):
forecast = ytrain[T - 1] + (ytrain[T - 1] -
ytrain[0]) * (i + 1) / (T - 1)
ytest_hat.append(forecast)
error = ytest[i] - forecast
etest.append(error)
setest.append(error**2)
elif method == "Simple Exponential Smoothing": # simple exponential smoothing method
alpha = 0.5
ytrain_hat.append(
ytrain[0]) # the first observation is the initial condition
etrain.append(np.nan) # for the initial condition, there is no error
setrain.append(np.nan)
for i in range(1, T):
prediction = alpha * ytrain[i - 1] + (1 - alpha) * ytrain_hat[i -
1]
ytrain_hat.append(prediction)
error = ytrain[i] - prediction
etrain.append(error)
setrain.append(error**2)
for i in range(0, h):
forecast = alpha * ytrain[T - 1] + (1 - alpha) * ytrain_hat[T - 1]
ytest_hat.append(forecast)
error = ytest[i] - forecast
etest.append(error)
setest.append(error**2)
elif method == "Holt's Linear Multiplicative": # Holt linear using multiplicative
holt = ets.ExponentialSmoothing(ytrain,
trend='multiplicative',
damped=True,
seasonal=None).fit()
ytrain_hat = holt.fittedvalues
ytest_hat = holt.forecast(steps=h)
for i in range(0, T):
error = ytrain[i] - ytrain_hat[i]
etrain.append(error)
setrain.append(error**2)
for i in range(0, h):
error = ytest[i] - ytest_hat[i]
etest.append(error)
setest.append(error**2)
elif method == "Holt's Linear":
holt = ets.ExponentialSmoothing(ytrain,
trend=None,
damped=False,
seasonal=None).fit(smoothing_level=0.1)
ytrain_hat = holt.fittedvalues
ytest_hat = holt.forecast(steps=h)
for i in range(0, T):
error = ytrain[i] - ytrain_hat[i]
etrain.append(error)
setrain.append(error**2)
for i in range(0, h):
error = ytest[i] - ytest_hat[i]
etest.append(error)
setest.append(error**2)
elif method == "Holt-Winter":
holt = ets.ExponentialSmoothing(train,
seasonal_periods=period,
trend='add',
damped_trend=True,
seasonal='additive')
holt = holt.fit(smoothing_level=0.1,
smoothing_seasonal=0.2,
smoothing_trend=None)
ytrain_hat = holt.fittedvalues
ytrain_hat = ytrain_hat.values
ytest_hat = holt.forecast(steps=h)
ytest_hat = ytest_hat.values
for i in range(0, T):
error = ytrain[i] - ytrain_hat[i]
etrain.append(error)
setrain.append(error**2)
for i in range(0, h):
error = ytest[i] - ytest_hat[i]
etest.append(error)
setest.append(error**2)
else:
print("Method is not applicable")
# create dataframes containing results after predicting the training data and forecasting the testing data
df_train = pd.DataFrame({
"time": xtrain,
"y_t": ytrain,
"hat_y_t": ytrain_hat,
"error": etrain,
"square_error": setrain
})
df_test = pd.DataFrame({
"time": xtest,
"y_t": ytest,
"hat_y_t": ytest_hat,
"error": etest,
"square_error": setest
})
return df_train, df_test
def test_keras_01(self):
def import_data(trend=False, seasonality=False):
"""
Returns a dataframe with time series values.
:param trend: boolean
If True, returns data with trend
:param seasonality: boolean
If True, returns data with seasonality
:return: ts_data: DataFrame
Index of the data frame is either a time-index or an integer index. First column has time series values
"""
if trend and seasonality:
# no of international visitors in Australia
data = [
41.7275, 24.0418, 32.3281, 37.3287, 46.2132, 29.3463,
36.4829, 42.9777, 48.9015, 31.1802, 37.7179, 40.4202,
51.2069, 31.8872, 40.9783, 43.7725, 55.5586, 33.8509,
42.0764, 45.6423, 59.7668, 35.1919, 44.3197, 47.9137
]
index = pd.DatetimeIndex(start='2005',
end='2010-Q4',
freq='QS')
ts_data = pd.Series(data, index)
ts_data.index.name = 'datetime_index'
ts_data.name = 'n_visitors'
return ts_data
elif trend:
# no. of annual passengers of air carriers registered in Australia
data = [
17.5534, 21.86, 23.8866, 26.9293, 26.8885, 28.8314,
30.0751, 30.9535, 30.1857, 31.5797, 32.5776, 33.4774,
39.0216, 41.3864, 41.5966
]
index = pd.DatetimeIndex(start='1990', end='2005', freq='A')
ts_data = pd.Series(data, index)
ts_data.index.name = 'datetime_index'
ts_data.name = 'n_passengers'
return ts_data
elif seasonality:
pass
else:
# Oil production in Saudi Arabia
data = [
446.6565, 454.4733, 455.663, 423.6322, 456.2713, 440.5881,
425.3325, 485.1494, 506.0482, 526.792, 514.2689, 494.211
]
index = pd.DatetimeIndex(start='1996', end='2008', freq='A')
ts_data = pd.Series(data, index)
ts_data.index.name = 'datetime_index'
ts_data.name = 'oil_production'
return ts_data
ts_data = import_data(trend=False, seasonality=False)
model_obj = hw.ExponentialSmoothing(ts_data,
trend=None,
damped=False,
seasonal=None,
seasonal_periods=None)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj,
'exponential_smoothing.pmml')
self.assertEqual(os.path.isfile("exponential_smoothing.pmml"), True)
ts_data = import_data(trend=True, seasonality=False)
model_obj = hw.ExponentialSmoothing(ts_data,
trend='add',
damped=False,
seasonal=None,
seasonal_periods=None)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj,
'exponential_smoothing.pmml')
self.assertEqual(os.path.isfile("exponential_smoothing.pmml"), True)
model_obj = hw.ExponentialSmoothing(ts_data,
trend='mul',
damped=False,
seasonal=None,
seasonal_periods=None)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj,
'exponential_smoothing.pmml')
self.assertEqual(os.path.isfile("exponential_smoothing.pmml"), True)
model_obj = hw.ExponentialSmoothing(ts_data,
trend='mul',
damped=True,
seasonal=None,
seasonal_periods=None)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj,
'exponential_smoothing.pmml')
self.assertEqual(os.path.isfile("exponential_smoothing.pmml"), True)
ts_data = import_data(trend=True, seasonality=True)
model_obj = hw.ExponentialSmoothing(ts_data,
trend='add',
damped=False,
seasonal='add',
seasonal_periods=4)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj,
'exponential_smoothing.pmml')
self.assertEqual(os.path.isfile("exponential_smoothing.pmml"), True)
model_obj = hw.ExponentialSmoothing(ts_data,
trend='add',
damped=True,
seasonal='add',
seasonal_periods=4)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj,
'exponential_smoothing.pmml')
self.assertEqual(os.path.isfile("exponential_smoothing.pmml"), True)
model_obj = hw.ExponentialSmoothing(ts_data,
trend='add',
damped=False,
seasonal='mul',
seasonal_periods=4)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj,
'exponential_smoothing.pmml')
self.assertEqual(os.path.isfile("exponential_smoothing.pmml"), True)
model_obj = hw.ExponentialSmoothing(ts_data,
trend='add',
damped=False,
seasonal='mul',
seasonal_periods=4)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj,
'exponential_smoothing.pmml')
self.assertEqual(os.path.isfile("exponential_smoothing.pmml"), True)
print(df.shape)
# Fill in nan with forward fill (replace nan with preceding value)
df = df.fillna(method='ffill')
time = pd.date_range('2004-03-10 18:00:00', periods=len(df.Time), freq='h')
lags = 50
# Split the data into 80% train and 20% test
y = df['CO(GT)']
y_train, y_test = train_test_split(y, shuffle=False, test_size=0.2)
x_train, x_test = train_test_split(time, shuffle=False, test_size=0.2)
# ******************************Holt-Winter method*****************************
Holt_Winter = ets.ExponentialSmoothing(y_train, trend='add', seasonal='add', seasonal_periods=24).fit()
Holt_Winter_pr = Holt_Winter.fittedvalues
Holt_Winter_forecast = Holt_Winter.forecast(len(y_test))
# Plot the Holt-Winter Seasonal Method and Forecast
plt.figure()
plt.plot(x_train, y_train, label = 'Training set')
plt.plot(x_test, y_test, label = 'Test set')
plt.plot(x_test, Holt_Winter_forecast, label = 'h-step Forecast')
plt.legend()
plt.xlabel('Time (hourly)')
plt.ylabel('Concentration')
plt.title('CO level - Holt-Winter Seasonal Method and Forecast')
plt.show()
# Residual errors
def test_exponentialSmoothing_01(self):
# data with trend and seasonality present
# no of international visitors in Australia
data = [
41.7275, 24.0418, 32.3281, 37.3287, 46.2132, 29.3463, 36.4829,
42.9777, 48.9015, 31.1802, 37.7179, 40.4202, 51.2069, 31.8872,
40.9783, 43.7725, 55.5586, 33.8509, 42.0764, 45.6423, 59.7668,
35.1919, 44.3197, 47.9137
]
index = pd.DatetimeIndex(start='2005', end='2010-Q4', freq='QS')
ts_data = pd.Series(data, index)
ts_data.index.name = 'datetime_index'
ts_data.name = 'n_visitors'
f_name = 'exponential_smoothing.pmml'
model_obj = hw.ExponentialSmoothing(ts_data,
trend='add',
damped=True,
seasonal='add',
seasonal_periods=2)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj, f_name)
self.assertEqual(os.path.isfile(f_name), True)
os.remove(f_name)
model_obj = hw.ExponentialSmoothing(ts_data,
trend='add',
damped=False,
seasonal='add',
seasonal_periods=2)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj, f_name)
self.assertEqual(os.path.isfile(f_name), True)
os.remove(f_name)
model_obj = hw.ExponentialSmoothing(ts_data,
trend='add',
damped=True,
seasonal='mul',
seasonal_periods=2)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj, f_name)
self.assertEqual(os.path.isfile(f_name), True)
os.remove(f_name)
model_obj = hw.ExponentialSmoothing(ts_data,
trend='add',
damped=False,
seasonal='mul',
seasonal_periods=2)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj, f_name)
self.assertEqual(os.path.isfile(f_name), True)
os.remove(f_name)
model_obj = hw.ExponentialSmoothing(ts_data,
trend='mul',
damped=True,
seasonal='add',
seasonal_periods=2)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj, f_name)
self.assertEqual(os.path.isfile(f_name), True)
os.remove(f_name)
model_obj = hw.ExponentialSmoothing(ts_data,
trend='mul',
damped=False,
seasonal='add',
seasonal_periods=2)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj, f_name)
self.assertEqual(os.path.isfile(f_name), True)
os.remove(f_name)
model_obj = hw.ExponentialSmoothing(ts_data,
trend='mul',
damped=True,
seasonal='mul',
seasonal_periods=2)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj, f_name)
self.assertEqual(os.path.isfile(f_name), True)
os.remove(f_name)
model_obj = hw.ExponentialSmoothing(ts_data,
trend='mul',
damped=False,
seasonal='mul',
seasonal_periods=2)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj, f_name)
self.assertEqual(os.path.isfile(f_name), True)
os.remove(f_name)
def exponentialFit(self, name):
''' Parameters
----------
name: name of model
'''
modelName = name
errorObjs = []
runTimeObj = obj.ModelsRunTime(name)
startTime = None
totalTime = None
# Step 1: fit selected model
if name == 'NAIVE':
# for evaluation
startTime = dt.datetime.now()
self.fittedModel = ts.ExponentialSmoothing(self.trainData)
self.fittedModel = self.fittedModel.fit(smoothing_level=1)
runTimeObj.setTrainingTime(dt.datetime.now() - startTime)
# for real forecasts
self.fittedModelFinal = ts.ExponentialSmoothing(self.data)
self.fittedModelFinal = self.fittedModelFinal.fit(
smoothing_level=1)
elif name == 'SES':
# for evaluation
startTime = dt.datetime.now()
self.fittedModel = ts.SimpleExpSmoothing(self.trainData)
self.fittedModel = self.fittedModel.fit(
optimized=True, use_brute=True) #grid search
runTimeObj.setTrainingTime(dt.datetime.now() - startTime)
# for real forecasts
self.fittedModelFinal = ts.SimpleExpSmoothing(self.data)
self.fittedModelFinal = self.fittedModelFinal.fit(
optimized=True, use_brute=True) #grid search
elif name == 'HOLT':
# Holt's linear trend method
# for evaluation
startTime = dt.datetime.now()
self.fittedModel = ts.Holt(self.trainData)
self.fittedModel = self.fittedModel.fit(
optimized=True, use_brute=True) #grid search
runTimeObj.setTrainingTime(dt.datetime.now() - startTime)
# for real forecasts
self.fittedModelFinal = ts.Holt(self.data)
self.fittedModelFinal = self.fittedModelFinal.fit(
optimized=True, use_brute=True) #grid search
# Step 2: get fitted values for training, test and forecasts
trainingFit = pd.Series(np.ceil(self.fittedModel.fittedvalues))
startTime = dt.datetime.now()
testPredictions = pd.Series(
np.ceil(self.fittedModel.forecast(len(self.testData))))
totalTime = dt.datetime.now() - startTime
forecasts = pd.Series(
np.ceil(self.fittedModelFinal.forecast(self.horizon)))
# Step 3: set error
errorObjs = self.setErrorData(trainingFit, testPredictions, runTimeObj)
runTimeObj.setTestTime(runTimeObj.getTestTime() + totalTime)
self.runTimeList.append(runTimeObj)
# Add to ModelsResult list
self.setModelResults(modelName, errorObjs, trainingFit,
testPredictions, forecasts)
def exponentialFit(self, name):
''' Parameters
----------
name: name of model
'''
modelName = name
errorObjs = []
# Step 1: fit selected model
if name == 'NAIVE':
# for evaluation
self.fittedModel = ts.ExponentialSmoothing(self.trainData)
self.fittedModel = self.fittedModel.fit(smoothing_level=1)
# for real forecasts
self.fittedModelFinal = ts.ExponentialSmoothing(self.data)
self.fittedModelFinal = self.fittedModelFinal.fit(
smoothing_level=1)
elif name == 'SES':
# for evaluation
self.fittedModel = ts.SimpleExpSmoothing(self.trainData)
self.fittedModel = self.fittedModel.fit(
optimized=True, use_brute=True) #grid search
# for real forecasts
self.fittedModelFinal = ts.SimpleExpSmoothing(self.data)
self.fittedModelFinal = self.fittedModelFinal.fit(
optimized=True, use_brute=True) #grid search
elif name == 'HOLT':
# Holt-Winters
# for evaluation
self.fittedModel = ts.Holt(self.trainData)
self.fittedModel = self.fittedModel.fit(
optimized=True, use_brute=True) #grid search
# for real forecasts
self.fittedModelFinal = ts.Holt(self.data)
self.fittedModelFinal = self.fittedModelFinal.fit(
optimized=True, use_brute=True) #grid search
# Step 2: get fitted values for training, test and forecasts
trainingFit = pd.Series(self.fittedModel.fittedvalues)
forecasts = pd.Series(self.fittedModelFinal.forecast(self.horizon))
if self.stepType == 'multi':
testPredictions = pd.Series(
self.fittedModel.forecast(len(self.testData)))
else:
# Compute one-step-ahead forecast over the test data
SESParams = self.fittedModel.params
self.fittedModel = ts.SimpleExpSmoothing(self.data)
if modelName == 'HOLT':
self.fittedModel = ts.Holt(self.data)
self.fittedModel = self.fittedModel.fit(
smoothing_level=SESParams['smoothing_level'],
optimized=False,
smoothing_slope=SESParams['smoothing_slope'],
initial_slope=SESParams['initial_slope'],
initial_level=SESParams['initial_level'],
use_brute=False)
else:
self.fittedModel = self.fittedModel.fit(
smoothing_level=SESParams['smoothing_level'],
optimized=False,
initial_level=SESParams['initial_level'],
use_brute=False)
testPredictions = self.fittedModel.fittedvalues[self.start:]
# Step 3: set error
errorObjs = self.setErrorData(trainingFit, testPredictions)
# Add to ModelsResult list
self.setModelResults(modelName, errorObjs, trainingFit,
testPredictions, forecasts)
def test_exponentialSmoothing_02(self):
# data with trend but no seasonality
# no. of annual passengers of air carriers registered in Australia
data = [
17.5534, 21.86, 23.8866, 26.9293, 26.8885, 28.8314, 30.0751,
30.9535, 30.1857, 31.5797, 32.5776, 33.4774, 39.0216, 41.3864,
41.5966
]
index = pd.DatetimeIndex(start='1990', end='2005', freq='A')
ts_data = pd.Series(data, index)
ts_data.index.name = 'datetime_index'
ts_data.name = 'n_passengers'
f_name = 'exponential_smoothing.pmml'
model_obj = hw.ExponentialSmoothing(ts_data,
trend='add',
damped=True,
seasonal=None,
seasonal_periods=2)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj, f_name)
self.assertEqual(os.path.isfile(f_name), True)
os.remove(f_name)
model_obj = hw.ExponentialSmoothing(ts_data,
trend='add',
damped=True,
seasonal=None,
seasonal_periods=None)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj, f_name)
self.assertEqual(os.path.isfile(f_name), True)
os.remove(f_name)
model_obj = hw.ExponentialSmoothing(ts_data,
trend='add',
damped=False,
seasonal=None,
seasonal_periods=2)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj, f_name)
self.assertEqual(os.path.isfile(f_name), True)
os.remove(f_name)
model_obj = hw.ExponentialSmoothing(ts_data,
trend='add',
damped=False,
seasonal=None,
seasonal_periods=None)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj, f_name)
self.assertEqual(os.path.isfile(f_name), True)
os.remove(f_name)
model_obj = hw.ExponentialSmoothing(ts_data,
trend='mul',
damped=True,
seasonal=None,
seasonal_periods=2)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj, f_name)
self.assertEqual(os.path.isfile(f_name), True)
os.remove(f_name)
model_obj = hw.ExponentialSmoothing(ts_data,
trend='mul',
damped=True,
seasonal=None,
seasonal_periods=None)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj, f_name)
self.assertEqual(os.path.isfile(f_name), True)
os.remove(f_name)
model_obj = hw.ExponentialSmoothing(ts_data,
trend='mul',
damped=False,
seasonal=None,
seasonal_periods=2)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj, f_name)
self.assertEqual(os.path.isfile(f_name), True)
os.remove(f_name)
model_obj = hw.ExponentialSmoothing(ts_data,
trend='mul',
damped=False,
seasonal=None,
seasonal_periods=None)
results_obj = model_obj.fit(optimized=True)
ExponentialSmoothingToPMML(ts_data, model_obj, results_obj, f_name)
self.assertEqual(os.path.isfile(f_name), True)
os.remove(f_name)
