CreateExponentialSmoothingModel

Introduction

Not every time series needs the machinery of ARIMA — a lot of forecasting problems are really just "weight recent history more than distant history, and account for trend and seasonality if they're there." CreateExponentialSmoothingModel covers that ground with Simple, Double (Holt), and Triple (Holt-Winters) exponential smoothing, either picking the right variant for you automatically or letting you set it explicitly, then optimizing the smoothing coefficients, decomposing the series into trend/seasonal/residual components, and comparing training vs. test RMSE so you know how much to trust the forecast.

Teaching Note

Exponential smoothing is essential for:

Generating short-term demand forecasts for inventory management and logistics
Predicting future financial performance and revenue trends for budget planning
Analyzing seasonal patterns in economic indicators or market activity
Monitoring operational metrics and projecting future capacity requirements
Forecasting call volumes or service requests for workforce optimization
Smoothing noisy historical data to identify underlying structural trends
Evaluating the impact of seasonality on business performance metrics

Parameters

Parameter	Default	Description
`dataframe`required	—	The input pandas DataFrame containing the time series data.
`time_column`required	—	The name of the column containing date or time information.
`outcome_column`required	—	The name of the target variable column intended for forecasting.
`smoothing_type`	`'auto'`	The complexity of the model: 'simple' (level only), 'double' (level + trend), 'triple' (level + trend + seasonality), or 'auto' for heuristic-based selection. Defaults to 'auto'.
`alpha`	`None`	The smoothing parameter for the level component (0 < alpha <= 1). Adjusts the weight of recent observations. Defaults to None.
`beta`	`None`	The smoothing parameter for the trend component (0 < beta <= 1). Determines how quickly the model adapts to trend changes. Defaults to None.
`gamma`	`None`	The smoothing parameter for the seasonal component (0 < gamma <= 1). Controls the sensitivity to seasonal fluctuations. Defaults to None.
`seasonal_periods`	`None`	The number of observations that make up a full season (e.g., 12 for monthly seasonality). Required for triple exponential smoothing.
`trend_type`	`'additive'`	The nature of the trend component: 'additive' or 'multiplicative'. Defaults to 'additive'.
`seasonal_type`	`'additive'`	The nature of the seasonal component: 'additive' or 'multiplicative'. Defaults to 'additive'.
`damped_trend`	`False`	Whether to dampen the trend approach to a horizontal line over time. Defaults to False.
`test_size`	`0.2`	Proportion of observations held out as a temporal test set (the last test_size fraction of rows). The model is fit on the remaining earlier observations. Defaults to 0.2.
`auto_optimize`	`True`	If True, utilizes numerical solvers to find optimal smoothing coefficients based on historical data. Defaults to True.
`optimization_method`	`'L-BFGS-B'`	The solver used for parameter tuning. Defaults to 'L-BFGS-B'.
`print_model_performance`	`True`	Whether to print performance metrics (RMSE, MAE, MAPE) to the console. Defaults to True.
`print_parameter_summary`	`True`	Whether to print the optimized smoothing coefficients to the console. Defaults to True.
`print_forecast_summary`	`True`	Whether to print a summary of the future forecast values to the console. Defaults to True.
`forecast_periods`	`12`	The number of future time steps to project beyond the historical data. Defaults to 12.
`data_source_for_plot`	`None`	Source citation string displayed in the caption of all generated plots. Defaults to None.
`plot_model_performance`	`True`	Whether to plot actual vs. fitted values to visualize model accuracy. Defaults to True.
`plot_forecast`	`True`	Whether to plot the future forecast against the historical series. Defaults to True.
`plot_decomposition`	`True`	Whether to plot the trend, seasonal, and residual decomposition of the series. Defaults to True.
`dot_fill_color`	`'#999999'`	Hex color for actual-value points on the model performance plot. Defaults to '#999999'.
`line_color`	`'#b0170c'`	Hex color for the fitted-value line on the model performance plot. Defaults to '#b0170c'.
`forecast_color`	`'#ff6b35'`	Hex color for the forecast line on the forecast plot. Defaults to '#ff6b35'.
`figure_size_for_performance_plot`	`(12, 8)`	Dimensions (width, height) for the model performance plot. Defaults to (12, 8).
`figure_size_for_forecast_plot`	`(12, 6)`	Dimensions (width, height) for the forecast plot. Defaults to (12, 6).
`figure_size_for_decomposition_plot`	`(12, 8)`	Dimensions (width, height) for the decomposition plot. Defaults to (12, 8).
`title_for_performance_plot`	`'Exponential Smoothing Model Performance'`	Title text for the model performance plot.
`subtitle_for_performance_plot`	`'Actual vs. Fitted values showing model accuracy'`	Subtitle text for the model performance plot.
`title_for_forecast_plot`	`'Exponential Smoothing Forecast'`	Title text for the forecast plot.
`subtitle_for_forecast_plot`	`'Future predictions based on historical patterns'`	Subtitle text for the forecast plot.
`title_for_decomposition_plot`	`'Time Series Decomposition'`	Title text for the decomposition plot.
`subtitle_for_decomposition_plot`	`'Breakdown of trend, seasonal, and residual components'`	Subtitle text for the decomposition plot.
`caption_for_performance_plot`	`None`	Optional caption text for the model performance plot. Defaults to None.
`caption_for_forecast_plot`	`None`	Optional caption text for the forecast plot. Defaults to None.
`caption_for_decomposition_plot`	`None`	Optional caption text for the decomposition plot. Defaults to None.
`title_y_indent_for_performance_plot`	`1.1`	Vertical position of the title on the model performance plot. Defaults to 1.1.
`subtitle_y_indent_for_performance_plot`	`1.05`	Vertical position of the subtitle on the model performance plot. Defaults to 1.05.
`title_y_indent_for_forecast_plot`	`1.1`	Vertical position of the title on the forecast plot. Defaults to 1.1.
`subtitle_y_indent_for_forecast_plot`	`1.05`	Vertical position of the subtitle on the forecast plot. Defaults to 1.05.
`title_y_indent_for_decomposition_plot`	`1.1`	Vertical position of the title on the decomposition plot. Defaults to 1.1.
`subtitle_y_indent_for_decomposition_plot`	`1.05`	Vertical position of the subtitle on the decomposition plot. Defaults to 1.05.
`caption_y_indent_for_performance_plot`	`-0.215`	Vertical position of the caption on the model performance plot. Defaults to -0.215.
`caption_y_indent_for_forecast_plot`	`-0.215`	Vertical position of the caption on the forecast plot. Defaults to -0.215.
`caption_y_indent_for_decomposition_plot`	`-0.215`	Vertical position of the caption on the decomposition plot. Defaults to -0.215.
`x_indent_for_performance_plot`	`-0.115`	Horizontal starting position for text on the model performance plot. Defaults to -0.115.
`x_indent_for_forecast_plot`	`-0.115`	Horizontal starting position for text on the forecast plot. Defaults to -0.115.
`x_indent_for_decomposition_plot`	`-0.115`	Horizontal starting position for text on the decomposition plot. Defaults to -0.115.
`plot_training_and_test_performance`	`True`	Whether to render a bar chart comparing training RMSE and test RMSE. Defaults to True.
`training_bar_color`	`'#3a86ff'`	Bar color for the training RMSE bar. Defaults to '#3a86ff'.
`test_bar_color`	`'#b0170c'`	Bar color for the test RMSE bar. Defaults to '#b0170c'.
`figure_size_for_performance_comparison_plot`	`(7, 5)`	Dimensions (width, height) for the training/test RMSE comparison chart. Defaults to (7, 5).
`title_for_performance_comparison_plot`	`None`	Title text for the training/test RMSE comparison chart. Defaults to a sensible built-in title when None.
`subtitle_for_performance_comparison_plot`	`None`	Subtitle text for the training/test RMSE comparison chart. Defaults to a sensible built-in subtitle when None.
`caption_for_performance_comparison_plot`	`None`	Optional caption text for the training/test RMSE comparison chart.
`title_y_indent_for_performance_comparison_plot`	`1.1`	Vertical position of the title on the comparison chart. Defaults to 1.1.
`subtitle_y_indent_for_performance_comparison_plot`	`1.05`	Vertical position of the subtitle on the comparison chart. Defaults to 1.05.
`caption_y_indent_for_performance_comparison_plot`	`-0.15`	Vertical position of the caption on the comparison chart. Defaults to -0.15.
`x_indent_for_performance_comparison_plot`	`-0.115`	Horizontal starting position for text on the comparison chart. Defaults to -0.115.

Returns

A dictionary with model (the fitted statsmodels Holt-Winters object), model_type (the smoothing method used), fitted_values (in-sample predictions), forecast (future predictions), performance_metrics (training_rmse, test_rmse, training_mae, training_mape), parameters (the optimized smoothing coefficients), and data (the cleaned time series used for modeling).

Example

python

from analysistoolbox.predictive_analytics import CreateExponentialSmoothingModel

# Automated model selection with a 12-period forecast
results = CreateExponentialSmoothingModel(
    df,
    time_column='Month',
    outcome_column='Sales',
)
forecast = results['forecast']