forecastit 5. winters' exponential smoothing
DESCRIPTION
This lesson begins with explaining the Winters’ exponential smoothing method characteristics, and uses. Winters’ exponential smoothing method attempts to best fit a smoothing constant, trend constant, and seasonal constants to past data. Using an example and the forecasting process, we apply the winters’ exponential smoothing method to create a model and forecast based upon it.TRANSCRIPT
![Page 1: ForecastIT 5. Winters' Exponential Smoothing](https://reader036.vdocument.in/reader036/viewer/2022083001/55851dbbd8b42aa86c8b4a79/html5/thumbnails/1.jpg)
Copyright 2010 DeepThought, Inc. 1
Winters’ Exponential Smoothing
Lesson #5
Winters’ Exponential Smoothing Method
![Page 2: ForecastIT 5. Winters' Exponential Smoothing](https://reader036.vdocument.in/reader036/viewer/2022083001/55851dbbd8b42aa86c8b4a79/html5/thumbnails/2.jpg)
Copyright 2010 DeepThought, Inc. 2
Winters’ Exponential Smoothing
Model Introduction• A smoothing technique designed to capture trend and seasonality• Estimates a smoothing equation• Uses the estimated smoothing equation to forecast future values• Method format:
– F(t) = a × X(t)/S(t-p) + (1-a) × [F(t-1) + T(t-1)]– S(t) = b × X(t)/F(t) + (1-b) × S(t-p)– T(t) = g × [F(t)-F(t-1)] + (1-g) × T (t-1)– W(t+m) = [F(t) + m × T (t)] × S(t+m-p)
![Page 3: ForecastIT 5. Winters' Exponential Smoothing](https://reader036.vdocument.in/reader036/viewer/2022083001/55851dbbd8b42aa86c8b4a79/html5/thumbnails/3.jpg)
Copyright 2010 DeepThought, Inc. 3
Winters’ Exponential Smoothing
Model Details• Method characteristics
– Fits a smoothing equation to data– Estimating a smoothing equation which minimizes the errors
between actual data points and model estimates• When to use method
– Data has trend and seasonality • When not to use
– Data does not exhibits trend or seasonality
![Page 4: ForecastIT 5. Winters' Exponential Smoothing](https://reader036.vdocument.in/reader036/viewer/2022083001/55851dbbd8b42aa86c8b4a79/html5/thumbnails/4.jpg)
Copyright 2010 DeepThought, Inc. 4
Winters’ Exponential Smoothing
Forecasting Steps1. Set an objective2. Build model3. Evaluate model4. Use model
![Page 5: ForecastIT 5. Winters' Exponential Smoothing](https://reader036.vdocument.in/reader036/viewer/2022083001/55851dbbd8b42aa86c8b4a79/html5/thumbnails/5.jpg)
Copyright 2010 DeepThought, Inc. 5
Winters’ Exponential Smoothing
Objective Setting• Simpler is better• Winters’ exponential smoothing allows to test whether a smoothing
technique with a trend and seasonality works as a model. Objectives should take that principal under consideration
• Example objectives for E-Commerce Retail Sales (see next slide):– Test if E-Commerce Retail Sales can be fit to a trend and
seasonality exponential smoothing model– If E-Commerce Retail Sales exhibits a statistically significant fit,
review and interpret results– If model looks good, create a forecast based off model
![Page 6: ForecastIT 5. Winters' Exponential Smoothing](https://reader036.vdocument.in/reader036/viewer/2022083001/55851dbbd8b42aa86c8b4a79/html5/thumbnails/6.jpg)
Copyright 2010 DeepThought, Inc. 6
Winters’ Exponential Smoothing
Example: E-Commerce Retail Sales
1998-07-24 1999-12-06 2001-04-19 2002-09-01 2004-01-14 2005-05-28 2006-10-10 2008-02-22 2009-07-06 2010-11-180
5000
10000
15000
20000
25000
30000
35000
40000
45000
E-Commerce Retail Sales (Millions of Dollars)
![Page 7: ForecastIT 5. Winters' Exponential Smoothing](https://reader036.vdocument.in/reader036/viewer/2022083001/55851dbbd8b42aa86c8b4a79/html5/thumbnails/7.jpg)
Copyright 2010 DeepThought, Inc. 7
Winters’ Exponential Smoothing
Build Model• Software finds us the best fit line to the data; minimizing the sum of
squared errors
1998-07-24 2001-04-19 2004-01-14 2006-10-10 2009-07-06 2012-04-01 2014-12-270
5000
10000
15000
20000
25000
30000
35000
40000
45000
E-Commerce Retail Sales (Millions of $'s)
Actual Forecast
![Page 8: ForecastIT 5. Winters' Exponential Smoothing](https://reader036.vdocument.in/reader036/viewer/2022083001/55851dbbd8b42aa86c8b4a79/html5/thumbnails/8.jpg)
Copyright 2010 DeepThought, Inc. 8
Winters’ Exponential Smoothing
Evaluate Model• Descriptive Statistics
– Mean– Variance & Standard Deviation
• Accuracy / Error– SSE– RMSE– MAPE– R2; Adjusted R2
• Statistical Significance– F-Test– P-Value F-Test
![Page 9: ForecastIT 5. Winters' Exponential Smoothing](https://reader036.vdocument.in/reader036/viewer/2022083001/55851dbbd8b42aa86c8b4a79/html5/thumbnails/9.jpg)
Copyright 2010 DeepThought, Inc. 9
Winters’ Exponential Smoothing
ExampleDescriptive Statistics
• Mean– 19673.68
• Variance– 99993090.28
• Standard Deviation– 9999.65
![Page 10: ForecastIT 5. Winters' Exponential Smoothing](https://reader036.vdocument.in/reader036/viewer/2022083001/55851dbbd8b42aa86c8b4a79/html5/thumbnails/10.jpg)
Copyright 2010 DeepThought, Inc. 10
Winters’ Exponential Smoothing
ExampleAccuracy / Error
• SSE– 82863394.84
• RMSE– 1439.30
• MAPE– 5%
• R2; Adjusted R2
– 97.9%– 97.8%
![Page 11: ForecastIT 5. Winters' Exponential Smoothing](https://reader036.vdocument.in/reader036/viewer/2022083001/55851dbbd8b42aa86c8b4a79/html5/thumbnails/11.jpg)
Copyright 2010 DeepThought, Inc. 11
Winters’ Exponential Smoothing
ExampleStatistical Significance
• F-Test– 1704.30
• P-Value F-Test– 0.00
![Page 12: ForecastIT 5. Winters' Exponential Smoothing](https://reader036.vdocument.in/reader036/viewer/2022083001/55851dbbd8b42aa86c8b4a79/html5/thumbnails/12.jpg)
Copyright 2010 DeepThought, Inc. 12
Winters’ Exponential Smoothing
Compare Multiple Models• Skip this step until have knowledge of multiple methods• Will use accuracy/error statistics to compare multiple models to
find best models
![Page 13: ForecastIT 5. Winters' Exponential Smoothing](https://reader036.vdocument.in/reader036/viewer/2022083001/55851dbbd8b42aa86c8b4a79/html5/thumbnails/13.jpg)
Copyright 2010 DeepThought, Inc. 13
Winters’ Exponential Smoothing
Use Model
• Understand limitations of model• Answer objectives
![Page 14: ForecastIT 5. Winters' Exponential Smoothing](https://reader036.vdocument.in/reader036/viewer/2022083001/55851dbbd8b42aa86c8b4a79/html5/thumbnails/14.jpg)
Copyright 2010 DeepThought, Inc. 14
Winters’ Exponential Smoothing
Example• Forecasts
2009-10-01 35774.56
2010-01-01 30645.94
2010-04-01 32229.96
2010-07-01 33925.73
2010-10-01 37987.46
2011-01-01 32512.73
2011-04-01 34163.79
2011-07-01 35931.22
2011-10-01 40200.36
2012-01-01 34379.52
2012-04-01 36097.62
2012-07-01 37936.72