Problem: Southwestern University is experiencing a quickly expanding football program. As a result, attendance for home games is increasing and approaching capacity. It is in the best interest of SWU to forecast attendance to aid them in deciding when the best time to expand the present stadium, which now holds 54,000.
Data: The following data is from the past six seasons, 2002-2007.
Game
Year – Game – Opponent
Attendance
2002-1 Baylor
34200
2002-2 Texas
39800
2002-3 LSU
38200
2002-4 Arkansas
26900
2002-5 USC
35100
2003-1 Oklahoma
36100
2003-2 Nebraska
40200
2003-3 UCLA
39100
2003-4 Nevada
25300
2003-5 Ohio State
36200
2004-1 TCU
35900
2004-2 Texas Tech
46500
2004-3 Alaska
43100
2004-4 Arizona
27900
2004-5 Rice
39200
2005-1 Arkansas
41900
2005-2 Missouri
46100
2005-3 Florida
43900
2005-4 Miami
30100
2005-5 Duke
40500
2006-1 Indiana
42500
2006-2 North Texas
48200
2006-3 Texas A&M
44200
2006-4 Southern
33900
2006-5 Oklahoma
47800
2007-1 LSU
46900
2007-2 Texas
50100
2007-3 Prairie View A&M
45900
2007-4 Montana
36300
2007-5 Arizona State
49900
An important thing to note is that the homecoming game of every season is the second home game (bold), and is always well attended. Also the forth home game always corresponds with a local festival that always draws from attendance (italics).
Summary of Forecasting Methods: Below is a table of the forecasting methods. The correlation coefficient, bias, mean absolute deviation (MAD), mean squared error (MSE), and mean absolute percent error (MAPE) are shown.
Correlation
Bias
MAD
MSE
MAPE
Naïve
--
541.38
6865.52
69,856,200
.19
Moving Average
(3 periods)
--
491.36
6,138.27
59,540,560
.17
Weighted Moving Average
(3 period; .6, .3, .1)
--
424.81
6,501.58
61,107,180
.18
Exponential smoothing
(alpha = 0.5)
--
794.28
5,880.56
50,755,960
.16
Trend Analysis
.54
0.00
4,355.70
31,285,700
.12
Seasonal Additive Decomposition
.97
0.00
1,251.26
2,386,650
.03
It is obvious that the superior