def fit(self, series: TimeSeries):
super().fit(series)
ts = self.training_series
self.length = len(ts)
# Check for statistical significance of user-defined season period
# or infers season_period from the TimeSeries itself.
if self.season_mode is SeasonalityMode.NONE:
self.season_period = 1
else:
self.season_period = self.seasonality_period
if self.season_period is None:
max_lag = len(ts) // 2
self.is_seasonal, self.season_period = check_seasonality(
ts, self.season_period, max_lag=max_lag
)
else:
# force the user-defined seasonality to be considered as a true seasonal period.
self.is_seasonal = self.season_period > 1
new_ts = ts
# Store and remove seasonality effect if there is any.
if self.is_seasonal:
_, self.seasonality = extract_trend_and_seasonality(
ts, self.season_period, model=self.season_mode
)
new_ts = remove_from_series(ts, self.seasonality, model=self.season_mode)
# SES part of the decomposition.
self.model = hw.SimpleExpSmoothing(new_ts.values(copy=False)).fit()
# Linear Regression part of the decomposition. We select the degree one coefficient.
b_theta = np.polyfit(
np.array([i for i in range(0, self.length)]),
(1.0 - self.theta) * new_ts.values(copy=False),
1,
)[0]
# Normalization of the coefficient b_theta.
self.coef = b_theta / (-self.theta)
self.alpha = self.model.params["smoothing_level"]
if self.alpha == 0.0:
self.model = hw.SimpleExpSmoothing(new_ts.values(copy=False)).fit(
initial_level=ALPHA_START
)
self.alpha = self.model.params["smoothing_level"]
return self
def exp_smoothing(ts, back_steps=10, degree=1, steps=1):
"""
Predicts next value using simple exponential smoothing.
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 exponential smoothing method.
"""
timeseries = np.array(list(ts.values()))
timeseries = timeseries[-back_steps:]
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.SimpleExpSmoothing(timeseries)
model_fit = model.fit()
x = model_fit.predict(len(timeseries), len(timeseries) + steps - 1)
return x[-1]
def fit(self, ts, season_period: int = None):
"""
Fits the Theta method to the TimeSeries `ts`.
The model decomposition is defined by the parameters `theta`, and the TimeSeries `ts`
is de-seasonalized according to `season_period`.
:param ts: The TimeSeries to fit.
:param season_period: User-defined seasonality period. Default to None.
"""
super().fit(ts)
self.length = len(ts)
# Check for statistical significance of user-defined season period
# or infers season_period from the TimeSeries itself.
if season_period is None:
max_lag = 24
self.is_seasonal, self.season_period = check_seasonality(ts, season_period, max_lag=max_lag)
else:
self.season_period = season_period
self.is_seasonal = True # force the user-defined seasonality to be considered as a true seasonal period.
new_ts = ts
# Store and remove seasonality effect if there is any.
if self.is_seasonal:
_, self.seasonality = extract_trend_and_seasonality(ts, self.season_period, model=self.mode)
new_ts = remove_seasonality(ts, self.season_period, model=self.mode)
# SES part of the decomposition.
self.model = hw.SimpleExpSmoothing(new_ts.values()).fit()
# Linear Regression part of the decomposition. We select the degree one coefficient.
b_theta = np.polyfit(np.array([i for i in range(0, self.length)]), (1.0 - self.theta) * new_ts.values(), 1)[0]
# Normalization of the coefficient b_theta.
self.coef = b_theta / (2.0 - self.theta)
self.alpha = self.model.params["smoothing_level"]
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 fit(self, series: TimeSeries, component_index: Optional[int] = None):
super().fit(series, component_index)
ts = self.training_series
self.length = len(ts)
self.season_period = self.seasonality_period
# Check for statistical significance of user-defined season period
# or infers season_period from the TimeSeries itself.
if self.season_period is None:
max_lag = len(ts) // 2
self.is_seasonal, self.season_period = check_seasonality(
ts, self.season_period, max_lag=max_lag)
logger.info(
'Theta model inferred seasonality of training series: {}'.
format(self.season_period))
else:
self.is_seasonal = True # force the user-defined seasonality to be considered as a true seasonal period.
new_ts = ts
# Store and remove seasonality effect if there is any.
if self.is_seasonal:
_, self.seasonality = extract_trend_and_seasonality(
ts, self.season_period, model=self.mode)
new_ts = remove_seasonality(ts,
self.season_period,
model=self.mode)
# SES part of the decomposition.
self.model = hw.SimpleExpSmoothing(new_ts.values()).fit()
# Linear Regression part of the decomposition. We select the degree one coefficient.
b_theta = np.polyfit(np.array([i for i in range(0, self.length)]),
(1.0 - self.theta) * new_ts.values(), 1)[0]
# Normalization of the coefficient b_theta.
self.coef = b_theta / (2.0 - self.theta)
self.alpha = self.model.params["smoothing_level"]
def fit(self, series):
super().fit(series)
self.length = len(series)
# normalization of data
if self.normalization:
self.mean = series.pd_dataframe(copy=False).mean().mean()
raise_if_not(
not np.isclose(self.mean, 0),
"The mean value of the provided series is too close to zero to perform normalization",
logger,
)
new_ts = series / self.mean
else:
new_ts = series
# Check for statistical significance of user-defined season period
# or infers season_period from the TimeSeries itself.
if self.season_mode is SeasonalityMode.NONE:
self.season_period = 1
else:
self.season_period = self.seasonality_period
if self.season_period is None:
max_lag = len(series) // 2
self.is_seasonal, self.season_period = check_seasonality(
series, self.season_period, max_lag=max_lag
)
else:
# force the user-defined seasonality to be considered as a true seasonal period.
self.is_seasonal = self.season_period > 1
# Store and remove seasonality effect if there is any.
if self.is_seasonal:
_, self.seasonality = extract_trend_and_seasonality(
new_ts, self.season_period, model=self.season_mode
)
new_ts = remove_from_series(
new_ts, self.seasonality, model=self.season_mode
)
ts_values = new_ts.univariate_values()
if (ts_values <= 0).any():
self.model_mode = ModelMode.ADDITIVE
self.trend_mode = TrendMode.LINEAR
logger.warning(
"Time series has negative values. Fallback to additive and linear model"
)
# Drift part of the decomposition
if self.trend_mode is TrendMode.LINEAR:
linreg = ts_values
else:
linreg = np.log(ts_values)
self.drift = np.poly1d(np.polyfit(np.arange(self.length), linreg, 1))
theta0_in = self.drift(np.arange(self.length))
if self.trend_mode is TrendMode.EXPONENTIAL:
theta0_in = np.exp(theta0_in)
if (theta0_in > 0).all() and self.model_mode is ModelMode.MULTIPLICATIVE:
theta_t = (ts_values**self.theta) * (theta0_in ** (1 - self.theta))
else:
if self.model_mode is ModelMode.MULTIPLICATIVE:
logger.warning("Negative Theta line. Fallback to additive model")
self.model_mode = ModelMode.ADDITIVE
theta_t = self.theta * ts_values + (1 - self.theta) * theta0_in
# SES part of the decomposition.
self.model = hw.SimpleExpSmoothing(theta_t).fit()
theta2_in = self.model.fittedvalues
if (theta2_in > 0).all() and self.model_mode is ModelMode.MULTIPLICATIVE:
self.fitted_values = theta2_in**self.wses * theta0_in**self.wdrift
else:
if self.model_mode is ModelMode.MULTIPLICATIVE:
self.model_mode = ModelMode.ADDITIVE
logger.warning("Negative Theta line. Fallback to additive model")
theta_t = self.theta * ts_values + (1 - self.theta) * theta0_in
self.model = hw.SimpleExpSmoothing(theta_t).fit()
theta2_in = self.model.fittedvalues
self.fitted_values = self.wses * theta2_in + self.wdrift * theta0_in
if self.is_seasonal:
if self.season_mode is SeasonalityMode.ADDITIVE:
self.fitted_values += self.seasonality.univariate_values(copy=False)
elif self.season_mode is SeasonalityMode.MULTIPLICATIVE:
self.fitted_values *= self.seasonality.univariate_values(copy=False)
# Fitted values are the results of the fit of the model on the train series. A good fit of the model
# will lead to fitted_values similar to ts. But one cannot see if it overfits.
if self.normalization:
self.fitted_values *= self.mean
return self
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 fit(self):
self.fittedModel = holtwinters.SimpleExpSmoothing(
self.ts_train).fit(optimized=True)
# root_path=r"C:\Users\Kang\Desktop\model_management\saved_models"
file_name=id+'_'+model_name+'.m' # e.g., 44117_WMA.m
model_path=os.path.join(root_path, file_name)
# 模型的加载
# model = joblib.load(model_path)
try:
model = joblib.load(model_path)
except Exception as e:
print('[ERROR]:sku_id({})的{}模型加载失败({})'.format(id,model_name,e))
return None
return model
if __name__=='__main__':
import numpy as np
import pandas as pd
from statsmodels.tsa import holtwinters
# 测试指数平滑
index=pd.date_range('5/1/2018',periods=20,freq='d')
ts_train=pd.Series([1.0,2,3,4,3,6,3,7,3,5,1,2,3,4,3,6,3,7,3,5],index=index)
train_model = holtwinters.SimpleExpSmoothing(ts_train).fit(optimized=True)
save_forecast_model(id='11111',model_name='SES',model=train_model)
loaded_model=load_forecast_model(id='11111',model_name='SES')
fittedvalues=loaded_model.predict(start=0,end=len(ts_train)-1)
y_predict=loaded_model.forecast(4)
print(fittedvalues)
print(y_predict)
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)