holt_winters()

Observability

Meaning

Holt-Winters method is a triple exponential smoothing technique for forecasting time series data with trend and seasonality. Use when you have seasonal patterns and need short-to-medium term forecasts.

Primary Function

Time series forecasting

Communicative Purpose

Predict future values of a seasonal time series by modeling level, trend, and seasonal components.

Pattern

holt_winters(data, seasonal_periods, trend='add', seasonal='add')

Core Structure

holt_winters(...)

Função primária

Time series forecasting

Propósito comunicativo

Predict future values of a seasonal time series by modeling level, trend, and seasonal components.

Situações de gatilho

Forecasting monthly sales with yearly seasonality; predicting electricity demand with daily/weekly patterns; estimating inventory needs for products with periodic demand.

Contextos

Appears in statistical libraries like statsmodels (Python), forecast package (R), Prophet, and in econometrics or data science codebases.

Padrão

holt_winters(data, seasonal_periods, trend='add', seasonal='add')

Estrutura central

holt_winters(...)

Slots de substituição

data: array-like time series, seasonal_periods: int length of seasonality, trend: 'add'|'mul'|None, seasonal: 'add'|'mul'|None

Colocados típicos

train_test_split
forecast
plot
evaluate with MAE/RMSE

Substituições comuns

SimpleExponentialSmoothing
ARIMA
Prophet

Erros comuns

Using wrong seasonal period, forgetting to set trend/seasonal components, not checking stationarity.

Similar / contraste

ARIMA (models linear trends and seasonality via differencing), ETS (exponential smoothing state space model) – Holt-Winters is a special case of ETS.

Interferências

Coming from ARIMA background: may assume differencing needed; Holt-Winters handles seasonality directly.

Família do chunk

Exponential smoothing
ETS
ARIMA
Prophet

Nuance

Assumes additive or multiplicative seasonality; not suitable for multiple seasonalities or non-stationary variance; performance O(n).

Efeito pragmático

Provides interpretable components and fast forecasts.

Dica de memória

Holt-Winters: level + trend + season

Nota

statsmodels' ExponentialSmoothing requires non‑missing values and the seasonal_periods argument must match the true season length; different libraries may name parameters differently (e.g., "seasonal" vs "seasonal_periods")

Upgrade path

Use statsmodels.tsa.holtwinters.ExponentialSmoothing with fitted model and .forecast().

Frequência: MediumFormulaicidade: Semi-fixedTipo de construção: functionPrioridade de aquisição: Active recallPrioridade de output: BothTag de espaçamento: Medium-term

Examples

import pandas as pd from statsmodels.tsa.holtwinters import ExponentialSmoothing # monthly sales data sales = pd.Series([120, 135, 150, 160, 170, 180, 190, 200, 210, 220, 230, 240], index=pd.date_range(start='2020-01', periods=12, freq='M')) model = ExponentialSmoothing(sales, seasonal_periods=12, trend='add', seasonal='add') fit = model.fit() forecast = fit.forecast(6) print(forecast)

import numpy as np from statsmodels.tsa.holtwinters import ExponentialSmoothing # synthetic weekly demand np.random.seed(0) demand = np.random.poisson(lam=50, size=52) * np.sin(np.linspace(0, 2*np.pi, 52)) + 100 model = ExponentialSmoothing(demand, seasonal_periods=7, trend='add', seasonal='mul') fit = model.fit() print(fit.forecast(14))

import pandas as pd from statsmodels.tsa.holtwinters import ExponentialSmoothing # quarterly revenue revenue = pd.Series([400, 420, 460, 480, 500, 530, 560, 590], index=pd.date_range(start='2019Q1', periods=8, freq='Q')) model = ExponentialSmoothing(revenue, seasonal_periods=4, trend='add', damped_trend=True) fit = model.fit() print(fit.forecast(4))