ETS models

The ETS models are a family of time series models with an underlying state space model consisting of a level component, a trend component (T), a seasonal component (S), and an error term (E).

This notebook shows how they can be used with statsmodels. For a more thorough treatment we refer to [1], chapter 8 (free online resource), on which the implementation in statsmodels and the examples used in this notebook are based.

statsmodels implements all combinations of: - additive and multiplicative error model - additive and multiplicative trend, possibly dampened - additive and multiplicative seasonality

However, not all of these methods are stable. Refer to [1] and references therein for more info about model stability.

[1] Hyndman, Rob J., and Athanasopoulos, George. Forecasting: principles and practice, 3rd edition, OTexts, 2021. https://otexts.com/fpp3/expsmooth.html

[1]:
import numpy as np
import matplotlib.pyplot as plt
import pandas as pd

%matplotlib inline
from statsmodels.tsa.exponential_smoothing.ets import ETSModel
np.random.seed(1234) # for reproducibility
[2]:
plt.rcParams["figure.figsize"] = (12, 8)

Simple exponential smoothing

The simplest of the ETS models is also known as simple exponential smoothing. In ETS terms, it corresponds to the (A, N, N) model, that is, a model with additive errors, no trend, and no seasonality. The state space formulation of Holt’s method is:

\begin{align} y_{t} &= y_{t-1} + e_t\\ l_{t} &= l_{t-1} + \alpha e_t\\ \end{align}

This state space formulation can be turned into a different formulation, a forecast and a smoothing equation (as can be done with all ETS models):

\begin{align} \hat{y}_{t|t-1} &= l_{t-1}\\ l_{t} &= \alpha y_{t-1} + (1 - \alpha) l_{t-1} \end{align}

Here, \(\hat{y}_{t|t-1}\) is the forecast/expectation of \(y_t\) given the information of the previous step. In the simple exponential smoothing model, the forecast corresponds to the previous level. The second equation (smoothing equation) calculates the next level as weighted average of the previous level and the previous observation.

[3]:
oildata = [
    111.0091,
    130.8284,
    141.2871,
    154.2278,
    162.7409,
    192.1665,
    240.7997,
    304.2174,
    384.0046,
    429.6622,
    359.3169,
    437.2519,
    468.4008,
    424.4353,
    487.9794,
    509.8284,
    506.3473,
    340.1842,
    240.2589,
    219.0328,
    172.0747,
    252.5901,
    221.0711,
    276.5188,
    271.1480,
    342.6186,
    428.3558,
    442.3946,
    432.7851,
    437.2497,
    437.2092,
    445.3641,
    453.1950,
    454.4096,
    422.3789,
    456.0371,
    440.3866,
    425.1944,
    486.2052,
    500.4291,
    521.2759,
    508.9476,
    488.8889,
    509.8706,
    456.7229,
    473.8166,
    525.9509,
    549.8338,
    542.3405,
]
oil = pd.Series(oildata, index=pd.date_range("1965", "2013", freq="YS"))
oil.plot()
plt.ylabel("Annual oil production in Saudi Arabia (Mt)")
[3]:
Text(0, 0.5, 'Annual oil production in Saudi Arabia (Mt)')
../../../_images/examples_notebooks_generated_ets_4_1.png

The plot above shows annual oil production in Saudi Arabia in million tonnes. The data are taken from the R package fpp2 (companion package to prior version [1]). Below you can see how to fit a simple exponential smoothing model using statsmodels’s ETS implementation to this data. Additionally, the fit using forecast in R is shown as comparison.

[4]:
model = ETSModel(oil)
fit = model.fit(maxiter=10000)
oil.plot(label="data")
fit.fittedvalues.plot(label="statsmodels fit")
plt.ylabel("Annual oil production in Saudi Arabia (Mt)")

# obtained from R
params_R = [0.99989969, 0.11888177503085334, 0.80000197, 36.46466837, 34.72584983]
yhat = model.smooth(params_R).fittedvalues
yhat.plot(label="R fit", linestyle="--")

plt.legend()
RUNNING THE L-BFGS-B CODE

           * * *

Machine precision = 2.220D-16
 N =            2     M =           10

At X0         0 variables are exactly at the bounds

At iterate    0    f=  6.27365D+00    |proj g|=  8.99900D-01

At iterate    1    f=  5.31675D+00    |proj g|=  6.49880D-04

At iterate    2    f=  5.30939D+00    |proj g|=  5.55378D-04

At iterate    3    f=  5.29115D+00    |proj g|=  5.80869D-05

At iterate    4    f=  5.29096D+00    |proj g|=  2.04281D-06

           * * *

Tit   = total number of iterations
Tnf   = total number of function evaluations
Tnint = total number of segments explored during Cauchy searches
Skip  = number of BFGS updates skipped
Nact  = number of active bounds at final generalized Cauchy point
Projg = norm of the final projected gradient
F     = final function value

           * * *

   N    Tit     Tnf  Tnint  Skip  Nact     Projg        F
    2      4     13      5     0     1   2.043D-06   5.291D+00
  F =   5.2909564634118551

CONVERGENCE: NORM_OF_PROJECTED_GRADIENT_<=_PGTOL
[4]:
<matplotlib.legend.Legend at 0xadde5de1e8ed>
../../../_images/examples_notebooks_generated_ets_6_2.png

By default the initial states are considered to be fitting parameters and are estimated by maximizing log-likelihood. It is possible to only use a heuristic for the initial values:

[5]:
model_heuristic = ETSModel(oil, initialization_method="heuristic")
fit_heuristic = model_heuristic.fit()
oil.plot(label="data")
fit.fittedvalues.plot(label="estimated")
fit_heuristic.fittedvalues.plot(label="heuristic", linestyle="--")
plt.ylabel("Annual oil production in Saudi Arabia (Mt)")

# obtained from R
params = [0.99989969, 0.11888177503085334, 0.80000197, 36.46466837, 34.72584983]
yhat = model.smooth(params).fittedvalues
yhat.plot(label="with R params", linestyle=":")

plt.legend()
RUNNING THE L-BFGS-B CODE

           * * *

Machine precision = 2.220D-16
 N =            1     M =           10

At X0         0 variables are exactly at the bounds

At iterate    0    f=  6.27365D+00    |proj g|=  8.99900D-01

At iterate    1    f=  5.31675D+00    |proj g|=  0.00000D+00

           * * *

Tit   = total number of iterations
Tnf   = total number of function evaluations
Tnint = total number of segments explored during Cauchy searches
Skip  = number of BFGS updates skipped
Nact  = number of active bounds at final generalized Cauchy point
Projg = norm of the final projected gradient
F     = final function value

           * * *

   N    Tit     Tnf  Tnint  Skip  Nact     Projg        F
    1      1      2      1     0     1   0.000D+00   5.317D+00
  F =   5.3167544390512402

CONVERGENCE: NORM_OF_PROJECTED_GRADIENT_<=_PGTOL
[5]:
<matplotlib.legend.Legend at 0xadde5de1e8ed>
../../../_images/examples_notebooks_generated_ets_8_2.png

The fitted parameters and some other measures are shown using fit.summary(). Here we can see that the log-likelihood of the model using fitted initial states is fractionally lower than the one using a heuristic for the initial states.

[6]:
print(fit.summary())
                                 ETS Results
==============================================================================
Dep. Variable:                      y   No. Observations:                   49
Model:                       ETS(ANN)   Log Likelihood                -259.257
Date:                Thu, 14 Nov 2024   AIC                            524.514
Time:                        09:54:11   BIC                            530.189
Sample:                    01-01-1965   HQIC                           526.667
                         - 01-01-2013   Scale                         2307.767
Covariance Type:               approx
===================================================================================
                      coef    std err          z      P>|z|      [0.025      0.975]
-----------------------------------------------------------------------------------
smoothing_level     0.9999      0.132      7.551      0.000       0.740       1.259
initial_level     110.7864     48.110      2.303      0.021      16.492     205.081
===================================================================================
Ljung-Box (Q):                        1.87   Jarque-Bera (JB):                20.78
Prob(Q):                              0.39   Prob(JB):                         0.00
Heteroskedasticity (H):               0.49   Skew:                            -1.04
Prob(H) (two-sided):                  0.16   Kurtosis:                         5.42
===================================================================================

Warnings:
[1] Covariance matrix calculated using numerical (complex-step) differentiation.
[7]:
print(fit_heuristic.summary())
                                 ETS Results
==============================================================================
Dep. Variable:                      y   No. Observations:                   49
Model:                       ETS(ANN)   Log Likelihood                -260.521
Date:                Thu, 14 Nov 2024   AIC                            525.042
Time:                        09:54:11   BIC                            528.826
Sample:                    01-01-1965   HQIC                           526.477
                         - 01-01-2013   Scale                         2429.964
Covariance Type:               approx
===================================================================================
                      coef    std err          z      P>|z|      [0.025      0.975]
-----------------------------------------------------------------------------------
smoothing_level     0.9999      0.132      7.559      0.000       0.741       1.259
==============================================
              initialization method: heuristic
----------------------------------------------
initial_level                          33.6309
===================================================================================
Ljung-Box (Q):                        1.85   Jarque-Bera (JB):                18.42
Prob(Q):                              0.40   Prob(JB):                         0.00
Heteroskedasticity (H):               0.44   Skew:                            -1.02
Prob(H) (two-sided):                  0.11   Kurtosis:                         5.21
===================================================================================

Warnings:
[1] Covariance matrix calculated using numerical (complex-step) differentiation.

Holt-Winters’ seasonal method

The exponential smoothing method can be modified to incorporate a trend and a seasonal component. In the additive Holt-Winters’ method, the seasonal component is added to the rest. This model corresponds to the ETS(A, A, A) model, and has the following state space formulation:

\begin{align} y_t &= l_{t-1} + b_{t-1} + s_{t-m} + e_t\\ l_{t} &= l_{t-1} + b_{t-1} + \alpha e_t\\ b_{t} &= b_{t-1} + \beta e_t\\ s_{t} &= s_{t-m} + \gamma e_t \end{align}

[8]:
austourists_data = [
    30.05251300,
    19.14849600,
    25.31769200,
    27.59143700,
    32.07645600,
    23.48796100,
    28.47594000,
    35.12375300,
    36.83848500,
    25.00701700,
    30.72223000,
    28.69375900,
    36.64098600,
    23.82460900,
    29.31168300,
    31.77030900,
    35.17787700,
    19.77524400,
    29.60175000,
    34.53884200,
    41.27359900,
    26.65586200,
    28.27985900,
    35.19115300,
    42.20566386,
    24.64917133,
    32.66733514,
    37.25735401,
    45.24246027,
    29.35048127,
    36.34420728,
    41.78208136,
    49.27659843,
    31.27540139,
    37.85062549,
    38.83704413,
    51.23690034,
    31.83855162,
    41.32342126,
    42.79900337,
    55.70835836,
    33.40714492,
    42.31663797,
    45.15712257,
    59.57607996,
    34.83733016,
    44.84168072,
    46.97124960,
    60.01903094,
    38.37117851,
    46.97586413,
    50.73379646,
    61.64687319,
    39.29956937,
    52.67120908,
    54.33231689,
    66.83435838,
    40.87118847,
    51.82853579,
    57.49190993,
    65.25146985,
    43.06120822,
    54.76075713,
    59.83447494,
    73.25702747,
    47.69662373,
    61.09776802,
    66.05576122,
]
index = pd.date_range("1999-03-01", "2015-12-01", freq="3MS")
austourists = pd.Series(austourists_data, index=index)
austourists.plot()
plt.ylabel("Australian Tourists")
[8]:
Text(0, 0.5, 'Australian Tourists')
../../../_images/examples_notebooks_generated_ets_13_1.png
[9]:
# fit in statsmodels
model = ETSModel(
    austourists,
    error="add",
    trend="add",
    seasonal="add",
    damped_trend=True,
    seasonal_periods=4,
)
fit = model.fit()

# fit with R params
params_R = [
    0.35445427,
    0.03200749,
    0.39993387,
    0.97999997,
    24.01278357,
    0.97770147,
    1.76951063,
    -0.50735902,
    -6.61171798,
    5.34956637,
]
fit_R = model.smooth(params_R)

austourists.plot(label="data")
plt.ylabel("Australian Tourists")

fit.fittedvalues.plot(label="statsmodels fit")
fit_R.fittedvalues.plot(label="R fit", linestyle="--")
plt.legend()
RUNNING THE L-BFGS-B CODE

           * * *

Machine precision = 2.220D-16
 N =            9     M =           10

At X0         1 variables are exactly at the bounds

At iterate    0    f=  3.40132D+00    |proj g|=  9.88789D-01

At iterate    1    f=  2.58255D+00    |proj g|=  9.99244D-01

At iterate    2    f=  2.49918D+00    |proj g|=  2.90033D-01

At iterate    3    f=  2.48198D+00    |proj g|=  2.44942D-01

At iterate    4    f=  2.43118D+00    |proj g|=  7.29728D-02

At iterate    5    f=  2.42924D+00    |proj g|=  7.03576D-02

At iterate    6    f=  2.42851D+00    |proj g|=  4.66403D-02

At iterate    7    f=  2.42794D+00    |proj g|=  2.92416D-02

At iterate    8    f=  2.42784D+00    |proj g|=  2.53310D-02

At iterate    9    f=  2.42721D+00    |proj g|=  1.89528D-02

At iterate   10    f=  2.42622D+00    |proj g|=  3.18662D-02

At iterate   11    f=  2.42512D+00    |proj g|=  3.53021D-02

At iterate   12    f=  2.42383D+00    |proj g|=  3.65646D-02

At iterate   13    f=  2.42196D+00    |proj g|=  4.83572D-02

At iterate   14    f=  2.41828D+00    |proj g|=  5.99647D-02

At iterate   15    f=  2.41131D+00    |proj g|=  6.89762D-02

At iterate   16    f=  2.40200D+00    |proj g|=  7.65482D-02

At iterate   17    f=  2.39385D+00    |proj g|=  8.77224D-02

At iterate   18    f=  2.37917D+00    |proj g|=  1.52247D-01

At iterate   19    f=  2.35438D+00    |proj g|=  2.40358D-01

At iterate   20    f=  2.33831D+00    |proj g|=  2.53084D-01

At iterate   21    f=  2.33773D+00    |proj g|=  2.54844D-01

At iterate   22    f=  2.33765D+00    |proj g|=  2.49903D-01

At iterate   23    f=  2.33765D+00    |proj g|=  2.49985D-01
  Positive dir derivative in projection
  Using the backtracking step

At iterate   24    f=  2.32727D+00    |proj g|=  2.23011D-01

At iterate   25    f=  2.29784D+00    |proj g|=  1.16766D-01

At iterate   26    f=  2.29783D+00    |proj g|=  1.19788D-01

At iterate   27    f=  2.29783D+00    |proj g|=  1.21967D-01

At iterate   28    f=  2.29781D+00    |proj g|=  1.25748D-01

At iterate   29    f=  2.29111D+00    |proj g|=  8.38166D-02

At iterate   30    f=  2.27666D+00    |proj g|=  2.21051D-01

At iterate   31    f=  2.27508D+00    |proj g|=  1.19037D-01

At iterate   32    f=  2.27314D+00    |proj g|=  6.98209D-02

At iterate   33    f=  2.27196D+00    |proj g|=  2.35783D-02

At iterate   34    f=  2.27175D+00    |proj g|=  1.71188D-02

At iterate   35    f=  2.27149D+00    |proj g|=  4.14437D-03

At iterate   36    f=  2.27141D+00    |proj g|=  5.89013D-03

At iterate   37    f=  2.27042D+00    |proj g|=  3.16237D-02

At iterate   38    f=  2.26905D+00    |proj g|=  5.01193D-02

At iterate   39    f=  2.26637D+00    |proj g|=  5.94819D-02

At iterate   40    f=  2.26306D+00    |proj g|=  4.38196D-02

At iterate   41    f=  2.25853D+00    |proj g|=  2.54350D-02

At iterate   42    f=  2.25535D+00    |proj g|=  3.73616D-02

At iterate   43    f=  2.25276D+00    |proj g|=  5.86500D-02

At iterate   44    f=  2.25045D+00    |proj g|=  6.65900D-02

At iterate   45    f=  2.24732D+00    |proj g|=  5.40858D-02

At iterate   46    f=  2.24579D+00    |proj g|=  2.35353D-02

At iterate   47    f=  2.24536D+00    |proj g|=  1.19850D-02

At iterate   48    f=  2.24527D+00    |proj g|=  2.61675D-03

At iterate   49    f=  2.24527D+00    |proj g|=  2.58109D-03

At iterate   50    f=  2.24524D+00    |proj g|=  5.36584D-03

At iterate   51    f=  2.24522D+00    |proj g|=  5.77085D-03

At iterate   52    f=  2.24519D+00    |proj g|=  2.31117D-03

At iterate   53    f=  2.24517D+00    |proj g|=  2.61458D-03

At iterate   54    f=  2.24516D+00    |proj g|=  4.23235D-03

At iterate   55    f=  2.24515D+00    |proj g|=  4.10258D-03

At iterate   56    f=  2.24508D+00    |proj g|=  9.58713D-03

At iterate   57    f=  2.24498D+00    |proj g|=  5.84883D-03

At iterate   58    f=  2.24479D+00    |proj g|=  1.33682D-02

At iterate   59    f=  2.24467D+00    |proj g|=  1.07525D-02

At iterate   60    f=  2.24464D+00    |proj g|=  1.75757D-02

At iterate   61    f=  2.24461D+00    |proj g|=  1.97208D-02

At iterate   62    f=  2.24453D+00    |proj g|=  2.22395D-03

At iterate   63    f=  2.24453D+00    |proj g|=  1.61275D-03

At iterate   64    f=  2.24452D+00    |proj g|=  1.45111D-03

At iterate   65    f=  2.24451D+00    |proj g|=  1.36064D-03

At iterate   66    f=  2.24451D+00    |proj g|=  6.19416D-04

At iterate   67    f=  2.24451D+00    |proj g|=  1.19194D-04

At iterate   68    f=  2.24451D+00    |proj g|=  6.74127D-05

At iterate   69    f=  2.24451D+00    |proj g|=  6.44818D-05

At iterate   70    f=  2.24451D+00    |proj g|=  6.63025D-05

At iterate   71    f=  2.24451D+00    |proj g|=  3.41327D-04

At iterate   72    f=  2.24451D+00    |proj g|=  2.87237D-04

At iterate   73    f=  2.24451D+00    |proj g|=  1.38112D-04

At iterate   74    f=  2.24451D+00    |proj g|=  3.04645D-04

At iterate   75    f=  2.24451D+00    |proj g|=  2.65654D-04

At iterate   76    f=  2.24451D+00    |proj g|=  3.79696D-04

At iterate   77    f=  2.24451D+00    |proj g|=  3.42570D-04

At iterate   78    f=  2.24451D+00    |proj g|=  6.34603D-05

At iterate   79    f=  2.24451D+00    |proj g|=  5.87974D-05

           * * *

Tit   = total number of iterations
Tnf   = total number of function evaluations
Tnint = total number of segments explored during Cauchy searches
Skip  = number of BFGS updates skipped
Nact  = number of active bounds at final generalized Cauchy point
Projg = norm of the final projected gradient
F     = final function value

           * * *

   N    Tit     Tnf  Tnint  Skip  Nact     Projg        F
    9     79    101     87     0     1   5.880D-05   2.245D+00
  F =   2.2445126261499118

CONVERGENCE: REL_REDUCTION_OF_F_<=_FACTR*EPSMCH
[9]:
<matplotlib.legend.Legend at 0xadde5de1e8ed>
../../../_images/examples_notebooks_generated_ets_14_2.png
[10]:
print(fit.summary())
                                 ETS Results
==============================================================================
Dep. Variable:                      y   No. Observations:                   68
Model:                      ETS(AAdA)   Log Likelihood                -152.627
Date:                Thu, 14 Nov 2024   AIC                            327.254
Time:                        09:54:11   BIC                            351.668
Sample:                    03-01-1999   HQIC                           336.928
                         - 12-01-2015   Scale                            5.213
Covariance Type:               approx
======================================================================================
                         coef    std err          z      P>|z|      [0.025      0.975]
--------------------------------------------------------------------------------------
smoothing_level        0.3399      0.111      3.069      0.002       0.123       0.557
smoothing_trend        0.0259      0.008      3.154      0.002       0.010       0.042
smoothing_seasonal     0.4011      0.080      5.041      0.000       0.245       0.557
damping_trend          0.9800        nan        nan        nan         nan         nan
initial_level         29.4499   4.18e+04      0.001      0.999   -8.18e+04    8.19e+04
initial_trend          0.6147      0.392      1.567      0.117      -0.154       1.384
initial_seasonal.0    -3.4360   4.18e+04  -8.23e-05      1.000   -8.18e+04    8.18e+04
initial_seasonal.1    -5.9492   4.18e+04     -0.000      1.000   -8.18e+04    8.18e+04
initial_seasonal.2   -11.4855   4.18e+04     -0.000      1.000   -8.19e+04    8.18e+04
initial_seasonal.3          0   4.18e+04          0      1.000   -8.18e+04    8.18e+04
===================================================================================
Ljung-Box (Q):                        5.76   Jarque-Bera (JB):                 7.69
Prob(Q):                              0.67   Prob(JB):                         0.02
Heteroskedasticity (H):               0.46   Skew:                            -0.63
Prob(H) (two-sided):                  0.07   Kurtosis:                         4.05
===================================================================================

Warnings:
[1] Covariance matrix calculated using numerical (complex-step) differentiation.

Predictions

The ETS model can also be used for predicting. There are several different methods available: - forecast: makes out of sample predictions - predict: in sample and out of sample predictions - simulate: runs simulations of the statespace model - get_prediction: in sample and out of sample predictions, as well as prediction intervals

We can use them on our previously fitted model to predict from 2014 to 2020.

[11]:
pred = fit.get_prediction(start="2014", end="2020")
[12]:
df = pred.summary_frame(alpha=0.05)
df
[12]:
mean pi_lower pi_upper
2014-03-01 67.611005 63.136033 72.085977
2014-06-01 42.814363 38.339390 47.289335
2014-09-01 54.106465 49.631492 58.581437
2014-12-01 57.928422 53.453449 62.403394
2015-03-01 68.422014 63.947042 72.896987
2015-06-01 47.278237 42.803265 51.753210
2015-09-01 58.955125 54.480153 63.430098
2015-12-01 63.982634 59.507662 68.457607
2016-03-01 75.905589 71.430616 80.380561
2016-06-01 51.418103 46.653872 56.182334
2016-09-01 63.703409 58.629269 68.777549
2016-12-01 67.978223 62.575546 73.380900
2017-03-01 78.315837 71.735296 84.896377
2017-06-01 53.780146 46.882604 60.677689
2017-09-01 66.018211 58.787455 73.248968
2017-12-01 70.246729 62.667909 77.825549
2018-03-01 80.538973 71.891696 89.186250
2018-06-01 55.958820 46.966873 64.950766
2018-09-01 68.153311 58.803927 77.502695
2018-12-01 72.339127 62.620542 82.057712
2019-03-01 82.589522 71.863228 93.315817
2019-06-01 57.968358 46.874136 69.062580
2019-09-01 70.122659 58.650390 81.594928
2019-12-01 74.269088 62.409419 86.128757
2020-03-01 84.480884 71.654506 97.307262

In this case the prediction intervals were calculated using an analytical formula. This is not available for all models. For these other models, prediction intervals are calculated by performing multiple simulations (1000 by default) and using the percentiles of the simulation results. This is done internally by the get_prediction method.

We can also manually run simulations, e.g. to plot them. Since the data ranges until end of 2015, we have to simulate from the first quarter of 2016 to the first quarter of 2020, which means 17 steps.

[13]:
simulated = fit.simulate(anchor="end", nsimulations=17, repetitions=100)
[14]:
for i in range(simulated.shape[1]):
    simulated.iloc[:, i].plot(label="_", color="gray", alpha=0.1)
df["mean"].plot(label="mean prediction")
df["pi_lower"].plot(linestyle="--", color="tab:blue", label="95% interval")
df["pi_upper"].plot(linestyle="--", color="tab:blue", label="_")
pred.endog.plot(label="data")
plt.legend()
[14]:
<matplotlib.legend.Legend at 0xadde5de1e8ed>
../../../_images/examples_notebooks_generated_ets_21_1.png

In this case, we chose “end” as simulation anchor, which means that the first simulated value will be the first out of sample value. It is also possible to choose other anchor inside the sample.