Project Report
Project Report
Outline
Model guidance
Data interpretation
Learning Point
Model Guidance
Data & Basic Holt-Winters
Data from 2013, 2014 and 2015
Moving average
Difference of neighboring moving average
Trend
Sensonal Index
Forecast 2016 and 2017 with Basic Formula
Model Guidance
Data & Basic Holt-Winters Result
Year Period Demand MA(4) Differences Trend Seasonal Index Period Seasonal Index
2013 1 8 7.22 1.11 1 1.01
2 10 9.47 1.06 2 0.97
3 7 10 11.72 0.60 3 0.66
4 15 11.75 1.75 13.97 1.07 4 1.18
2014 1 15 13.5 1.75 16.22 0.92
2 17 15.25 1.75 18.47 0.92
3 14 18.5 3.25 20.72 0.68
4 28 21 2.5 22.97 1.22
2015 1 25 23.25 2.25 25.22 0.99
2 26 25 1.75 27.47 0.95
3 21 28 3 29.72 0.71
4 40 31.97 1.25
average 18.47 2.25
Forecast Trend
2016 1 34 34.22
2 36 36.47
3 26 38.72
4 48 40.97
2017 1 44 43.22
2 44 45.47
3 31 47.72
4 59 49.97
Model Guidance
H-W Updating
Data 2016 as the new data that is available
Use Data 2016 to update the parameters
calculate updated values of the stationary, trend and seasonal indices for each period in 2016
Dynamically update the forecasts using new data
Forecast 2017 with Updated Formula
Model Guidance
H-W Updating Result
Smoothing values 0.2 0.2 0.6
Year Period Seasonal Index L Forecast Observation Observed L L Observed Trend T Observed I I
2016 1 1.01 34 34 31 31 34 2 2.1 0.9 0.96
2 0.97 36 35 34 35 35 2 2.1 1.0 0.96
3 0.66 38 25 28 42 39 3 2.3 0.7 0.70
4 1.18 41 48 57 48 42 4 2.6 1.3 1.28
Forecast 2017 with Updated Formula Forecast 2017 with Basic Formula
Year Period Forecast Year Period Forecast
2017 1 43 2017 1 44
2 46 2 44
3 35 3 31
4 67 4 59
Model Guidance
Forecast Errors
Basic Forecast and updated forecast errors for data 2016
Period Actual Demand Basic Forecast Error SE AE PE APE Updated Forecast Error SE AE PE APE
1 31 34 -3 12.19 3.49 -11.26 11.26 34 -3 12.19 3.49 -11.26 11.26
2 34 36 -2 2.34 1.53 -4.50 4.50 35 -1 0.52 0.72 -2.12 2.12
3 28 26 2 6.01 2.45 8.76 8.76 25 3 10.30 3.21 11.46 11.46
4 57 48 9 74.04 8.60 15.10 15.10 48 9 77.08 8.78 15.40 15.40
Mean 2 24 4 2 10 2 25 4 3 10
ME MSE MAE MPE MAPE ME MSE MAE MPE MAPE
Model Guidance
Suitable for data with seasonal pattern
Usage:
When new data for next quarter is available, only need to add data to table of `H-W Updating` sheet which update the parameters and forecast the next quarter dynamically.
Possible Improvement：
Adjust the smoothing coefficients to improve the accuracy.
Data interpretation
Seasonal Pattern
Overall the demand is increasing in each year
Quarter 4 has much larger demand than other quarters
Actual Demand from 2013 to 2015
1 2 3 4 1 2 3 4 1 2 3 4 8 10 7 15 15 17 14 28 25 26 21 40 Time
Quarter
Demand
Data interpretation
The forecast reflects the patterns observed in
actual data in previous slide.
Forecast Demand for 2016 To 2017
34.490744256648547 35.529169000621877 25.547653054111748 48.395132705445235 43.561362064403525 44.296458228498111 31.485558068123378 59.025656770912533 Time Quater
Demand
Learning Point
Acquire the skill to use Holt-Winters method for forecasting seasonal data
Excel operation skills such as math calculation and using built-in functions
VBA coding skills such as writing custom functions to do complex computing
/docProps/thumbnail.jpeg