Método de Holt-Winters

5.11.3

Método de Holt-Winters

Actualizado 2020-02-26 Lectura 3 min
1
2
3
4
5
6
7
8

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns

from statsmodels.tsa.holtwinters import ExponentialSmoothing
from statsmodels.tsa.holtwinters import Holt
from statsmodels.tsa.holtwinters import SimpleExpSmoothing

Datos de muestra #

Se están utilizando los datos de Consumo de Gas Natural (NATURALGAS). La unidad de medida de los valores es BCF (mil millones de pies cúbicos).

  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127

data = {
    "value": {
        "2013-01-01": 2878.8,
        "2013-02-01": 2567.2,
        "2013-03-01": 2521.1,
        "2013-04-01": 1967.5,
        "2013-05-01": 1752.5,
        "2013-06-01": 1742.9,
        "2013-07-01": 1926.3,
        "2013-08-01": 1927.4,
        "2013-09-01": 1767.0,
        "2013-10-01": 1866.8,
        "2013-11-01": 2316.9,
        "2013-12-01": 2920.8,
        "2014-01-01": 3204.1,
        "2014-02-01": 2741.2,
        "2014-03-01": 2557.9,
        "2014-04-01": 1961.7,
        "2014-05-01": 1810.2,
        "2014-06-01": 1745.4,
        "2014-07-01": 1881.0,
        "2014-08-01": 1933.1,
        "2014-09-01": 1809.3,
        "2014-10-01": 1912.8,
        "2014-11-01": 2357.5,
        "2014-12-01": 2679.2,
        "2015-01-01": 3115.0,
        "2015-02-01": 2925.2,
        "2015-03-01": 2591.3,
        "2015-04-01": 2007.9,
        "2015-05-01": 1858.2,
        "2015-06-01": 1899.9,
        "2015-07-01": 2067.7,
        "2015-08-01": 2052.7,
        "2015-09-01": 1901.3,
        "2015-10-01": 1987.3,
        "2015-11-01": 2249.1,
        "2015-12-01": 2588.2,
        "2016-01-01": 3091.7,
        "2016-02-01": 2652.3,
        "2016-03-01": 2356.3,
        "2016-04-01": 2083.9,
        "2016-05-01": 1965.8,
        "2016-06-01": 2000.7,
        "2016-07-01": 2186.6,
        "2016-08-01": 2208.4,
        "2016-09-01": 1947.8,
        "2016-10-01": 1925.2,
        "2016-11-01": 2159.5,
        "2016-12-01": 2866.3,
        "2017-01-01": 2913.8,
        "2017-02-01": 2340.2,
        "2017-03-01": 2523.3,
        "2017-04-01": 1932.0,
        "2017-05-01": 1892.0,
        "2017-06-01": 1910.4,
        "2017-07-01": 2141.6,
        "2017-08-01": 2093.8,
        "2017-09-01": 1920.5,
        "2017-10-01": 2031.5,
        "2017-11-01": 2357.3,
        "2017-12-01": 3086.0,
        "2018-01-01": 3340.9,
        "2018-02-01": 2710.7,
        "2018-03-01": 2796.7,
        "2018-04-01": 2350.5,
        "2018-05-01": 2055.0,
        "2018-06-01": 2063.1,
        "2018-07-01": 2350.7,
        "2018-08-01": 2313.8,
        "2018-09-01": 2156.1,
        "2018-10-01": 2285.9,
        "2018-11-01": 2715.9,
        "2018-12-01": 2999.5,
        "2019-01-01": 3424.3,
        "2019-02-01": 3019.1,
        "2019-03-01": 2927.8,
        "2019-04-01": 2212.4,
        "2019-05-01": 2134.0,
        "2019-06-01": 2119.3,
        "2019-07-01": 2393.9,
        "2019-08-01": 2433.9,
        "2019-09-01": 2206.3,
        "2019-10-01": 2306.5,
        "2019-11-01": 2783.8,
        "2019-12-01": 3170.7,
        "2020-01-01": 3320.6,
        "2020-02-01": 3058.5,
        "2020-03-01": 2722.0,
        "2020-04-01": 2256.9,
        "2020-05-01": 2072.2,
        "2020-06-01": 2127.9,
        "2020-07-01": 2464.1,
        "2020-08-01": 2399.5,
        "2020-09-01": 2151.2,
        "2020-10-01": 2315.9,
        "2020-11-01": 2442.0,
        "2020-12-01": 3182.8,
        "2021-01-01": 3343.9,
        "2021-02-01": 3099.2,
        "2021-03-01": 2649.4,
        "2021-04-01": 2265.1,
        "2021-05-01": 2117.4,
        "2021-06-01": 2238.4,
        "2021-07-01": 2412.2,
        "2021-08-01": 2433.8,
        "2021-09-01": 2142.3,
        "2021-10-01": 2262.6,
        "2021-11-01": 2693.3,
        "2021-12-01": 3007.3,
        "2022-01-01": 3612.1,
        "2022-02-01": 3064.2,
        "2022-03-01": 2785.4,
        "2022-04-01": 2379.3,
        "2022-05-01": 2247.8,
        "2022-06-01": 2326.9,
        "2022-07-01": 2597.9,
        "2022-08-01": 2566.1,
        "2022-09-01": 2263.3,
    }
}

data = pd.DataFrame(data)
data.rename(columns={"value": "Natural Gas Consumption(BCF)"}, inplace=True)
data.index = pd.to_datetime(data.index)
data = data.asfreq("MS")
data.head()

Natural Gas Consumption(BCF)
2013-01-012878.8
2013-02-012567.2
2013-03-012521.1
2013-04-011967.5
2013-05-011752.5
1
2
3
4

plt.figure(figsize=(12, 6))
sns.lineplot(data=data)
plt.grid()
plt.show()

Datos de muestra (figura)

División de datos #

Se divide el conjunto de datos para utilizar los datos más recientes (a partir de 2020) con el fin de verificar el rendimiento de las predicciones.

1
2
3
4
5
6
7
8

data_train = data[data.index < "2020-1-1"]
data_test = data[data.index >= "2020-1-1"]

plt.figure(figsize=(12, 6))
plt.plot(data_train.index, data_train.values, label="train", linewidth=2)
plt.plot(data_test.index, data_test.values, "-.", label="test")
plt.legend()
plt.grid()

Se divide el conjunto de datos para utilizar los datos más r… (figura)

Entrenamiento y predicción del modelo #

Se especifican los parámetros mientras se consulta la documentación de statsmodels.tsa.holtwinters.ExponentialSmoothing.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13

ses = SimpleExpSmoothing(data_train)
ses = ses.fit(smoothing_level=0.1)
ses_pred = ses.forecast(33)

holt = Holt(data_train)
holt = holt.fit(smoothing_level=0.1, smoothing_trend=0.2)
holt_pred = holt.forecast(33)

hw = ExponentialSmoothing(
    data_train, trend="additive", seasonal="add", seasonal_periods=12
)
hw = hw.fit()
hw_pred = hw.forecast(33)

1
2
3
4
5
6
7
8

plt.figure(figsize=(12, 6))
plt.plot(data_train.index, data_train.values, label="train", linewidth=2)
plt.plot(data_test.index, data_test.values, "-.", label="test")
plt.plot(ses_pred.index, ses_pred.values, "-.", label="prediction(ses)")
plt.plot(holt_pred.index, holt_pred.values, "-.", label="prediction(Holt)")
plt.plot(hw_pred.index, hw_pred.values, "-o", label="prediction(Holt-Winters)")
plt.legend()
plt.grid()

ExponentialSmoothing (figura)