5/804
(a) Time series plot and pattern
Pattern observation:
- The data shows irregular fluctuations.
- There is no clear trend or seasonality.
- This type of pattern is often referred to as random variation
(b) Three-week moving average and MSE
Week Value MA Forecast
4 11 (18+13+16)/3 = 15.67
5 17 (13+16+11)/3 = 13.33
6 14 (16+11+17)/3 = 14.67
7 ? (11+17+14)/3 = 14.00
MSE (Mean Squared Error):
(c ) Exponential smoothing with α = 0.2
Ft=αAt−1+(1−α)Ft−1
Let the initial forecast F1=A1=18
Week Actual Forecast Ft
1 18 18.00 (initial)
2 13 18.00
3 16 0.2×13 + 0.8×18 = 17.00
4 11 0.2×16 + 0.8×17.00 = 16.80
5 17 0.2×11 + 0.8×16.80 = 15.44
6 14 0.2×17 + 0.8×15.44 = 15.55
7 ? 0.2×14 + 0.8×15.55 = 15.24
(d) Comparison of MA vs. Exponential smoothing
Method MSE Week 7
Forecast
Moving Average (3-week) 11.90 14.00
Exponential Smoothing 12.89 15.24
Conclusion:
- The three-week moving average method provides a lower MSE → more accurate.
- This is likely because moving average smooths out noise better when no strong trend
exists.
(e) Exponential Smoothing with α = 0.4
Week Actual Forecast Ft
1 18 18.00 (initial)
2 13 18.00
3 0.4×13 + 0.6×18 = 15.00
4 0.4×16 + 0.6×15 = 15.40
5 0.4×11 + 0.6×15.4 = 13.64
6 0.4×17 + 0.6×13.64 = 14.78
7 ? 0.4×14 + 0.6×14.78 = 14.47
7/805
calculate the 4-week moving average from week 5 onward:
(a)Moving average, Time series smoothing
● Week 5:
(17+21+19+23)/4=80/4=20.00
(17+21+19+23)/4=80/4=20.00
● Week 6:
(21+19+23+18)/4=81/4=20.25
(21+19+23+18)/4=81/4=20.25
● Week 7:
(19+23+18+16)/4=76/4=19.00
(19+23+18+16)/4=76/4=19.00
● Week 8:
(23+18+16+20)/4=77/4=19.25
(23+18+16+20)/4=77/4=19.25
● Week 9:
(18+16+20+18)/4=72/4=18.00
(18+16+20+18)/4=72/4=18.00
● Week 10:
(16+20+18+17)/4=71/4=17.75
(16+20+18+17)/4=71/4=17.75
● Week 11:
(20+18+17+19)/4=74/4=18.50
(20+18+17+19)/4=74/4=18.50
● Week 12:
(18+17+19+20)/4=74/4=18.50
(18+17+19+20)/4=74/4=18.50
calculate the 5-week moving average from week 6 onward
● Week 6:
(17+21+19+23+18)/5=98/5=19.60
(17+21+19+23+18)/5=98/5=19.60
● Week 7:
(21+19+23+18+16)/5=97/5=19.40
(21+19+23+18+16)/5=97/5=19.40
● Week 8:
(19+23+18+16+20)/5=96/5=19.20
(19+23+18+16+20)/5=96/5=19.20
● Week 9:
(23+18+16+20+18)/5=95/5=19.00
(23+18+16+20+18)/5=95/5=19.00
● Week 10:
(18+16+20+18+17)/5=89/5=17.80
(18+16+20+18+17)/5=89/5=17.80
● Week 11:
(16+20+18+17+19)/5=90/5=18.00
(16+20+18+17+19)/5=90/5=18.00
● Week 12:
(20+18+17+19+20)/5=94/5=18.80
(20+18+17+19+20)/5=94/5=18.80
(b)
● Four-week Moving Average Errors and MSE
Sum = 4.00+18.06+1.00+1.56+1.00+1.56+2.25+0.25 = 29.68
MSE4 = 29.68/8 = 3.71
● Five-week Moving Average Errors and MSE
Sum = 12.96+0.36+1.44+4.00+1.44+4.00+0.64 = 24.84
MSE5 = 24.84/7 = 3.55
(c )
Compare the MSE values
- 3-week moving average MSE = 10.22 (given)
- 4-week moving average MSE = 3.71
- 5-week moving average MSE = 3.55
The moving average with the lowest MSE provides the best fit -> Lowest is 5-week moving
average (MSE = 3.55)
11/806
(a)
- The percentages range between 80% and 85%, showing small fluctuations around a
central value.
- There is no clear upward or downward trend, and no strong seasonality.
- The data appears to be relatively stable with minor random variations.
(b)
Three-Month Moving Average
For month 4 and onward, compute the moving averages:
- Forecast for Month 4 = (80+82+84)/3 = 82.0
- Forecast for Month 5 = (82+84+83)/3 = 83.0
- Forecast for Month 6 = (84+83+83)/3 ≈ 83.33
- Forecast for Month 7 = (83+83+84)/3 ≈ 83.33
- Forecast for Month 8 = (83+84+85)/3 = 84.0
- Forecast for Month 9 = (84+85+84)/3 ≈ 84.33
- Forecast for Month 10 = (85+84+82)/3 ≈ 83.67
- Forecast for Month 11 = (84+82+83)/3 = 83.0
- Forecast for Month 12 = (82+83+84)/3 = 83.0
Sum of squared errors ≈ 11.12
MSE (3MA) ≈ 11.12/9 ≈ 1.24
Exponential Smoothing with α=0.2 using the formula: Ft+1=αXt+(1−α)
with initial forecast F1=80
Compute forecasts sequentially:
● F2=0.2×80+0.8×80=80.00
● F3=0.2×82+0.8×80=80.4
● F4=0.2×84+0.8×80.4=81.12(Error: 83−81.12=1.88, Squared Error≈3.53)
● F5=0.2×83+0.8×81.12≈81.50
● F6=0.2×83+0.8×81.50=81.80(Error for Month 6: 84−81.80=2.20, Squared
Error≈4.84)
● F7=0.2×84+0.8×81.80=82.24 (Error:85−82.24=2.76, Squared Error ≈7.62)
● F8=0.2×85+0.8×82.24≈82.79 (Error:84−82.79≈1.21, Squared Error≈1.46)
● F9=0.2×84+0.8×82.79≈83.03 (Error: 82−83.03≈−1.03, Squared Error ≈1.06)
● F10=0.2×82+0.8×83.03≈82.83 (Error: 83−82.83≈0.17, Squared Error ≈0.03)
● F11=0.2×83+0.8×82.83≈82.86 (Error: 84−82.86≈1.14, Squared Error ≈1.30)
● F12=0.2×84+0.8×82.86≈83.09 (Error: 83−83.09≈−0.09, Squared Error≈0.008)
Sum of squared errors ≈22.10
MSE (Exponential Smoothing) ≈22.10/9≈2.46
Since the three-month moving average method has an MSE of approximately 1.24 compared
to 2.46 for exponential smoothing, the three-month moving average provides more accurate
forecasts.
(c )
- Three-Month Moving Average Forecast: Use Months 10, 11, and 12:
83+84+833 ≈ 83.33.
- Exponential Smoothing Forecast: Using the last computed forecast:
F13=0.2×83+0.8×83.09 ≈ 83.07
Using the more accurate method (moving average), the forecast for next month is 83.33.
13/806
a)
- Three-Month Moving Average Forecasts:
Month 4: (240 + 350 + 230) / 3 = 273.33
Month 5: (350 + 230 + 260) / 3 = 280
Month 6: (230 + 260 + 280) / 3 = 256.67
Month 7: (260 + 280 + 320) / 3 = 286.67
Month 8: (280 + 320 + 220) / 3 = 273.33
Month 9: (320 + 220 + 310) / 3 = 283.33
Month 10: (220 + 310 + 240) / 3 = 256.67
Month 11: (310 + 240 + 310) / 3 = 286.67
Month 12: (240 + 310 + 240) / 3 = 263.33
- Errors:
Month 4: 260 - 273.33 = -13.33
Month 5: 280 - 280 = 0
Month 6: 320 - 256.67 = 63.33
Month 7: 220 - 286.67 = -66.67
Month 8: 310 - 273.33 = 36.67
Month 9: 240 - 283.33 = -43.33
Month 10: 310 - 256.67 = 53.33
Month 11: 240 - 286.67 = -46.67
Month 12: 230 - 263.33 = -33.33
MSE: [(-13.33^2) + (0^2) + 63.33^2 + (-66.67^2) + 36.67^2 + (-43.33²) + 53.33^2 +
(-46.67^2) + (-33.33^2)] / 9 = 2221.23
- Exponential Smoothing Forecasts (α = 0.2):
F2 = 0.2 * 350 + 0.8 * 240 = 70 + 192 = 262
F3 = 0.2 * 230 + 0.8 * 262 = 46 + 209.6 = 255.6
F4 = 0.2 * 260 + 0.8 * 255.6 = 52 + 204.48 = 256.48
F5 = 0.2 * 280 + 0.8 * 256.48 = 56 + 205.184 = 261.184
F6 = 0.2 * 320 + 0.8 * 261.184 = 64 + 208.9472 = 272.9472
F7 = 0.2 * 220 + 0.8 * 272.9472 = 44 + 218.35776 = 262.35776
F8 = 0.2 * 310 + 0.8 * 262.35776 = 62 + 209.886208 = 271.886208
F9 = 0.2 * 240 + 0.8 * 271.886208 = 48 + 217.5089664 = 265.5089664
F10 = 0.2 * 310 + 0.8 * 265.5089664 = 62 + 212.40717312 = 274.40717312
F11 = 0.2 * 240 + 0.8 * 274.40717312 = 48 + 219.525738496 = 267.525738496
F12 = 0.2 * 230 + 0.8 * 267.525738496 = 46 + 214.020590797 = 260.020590797
- Errors:
Month 2: 350 - 262 = 88
Month 3: 230 - 255.6 = -25.6
Month 4: 260 - 256.48 = 3.52
Month 5: 280 - 261.184 = 18.816
Month 6: 320 - 272.9472 = 47.0528
Month 7: 220 - 262.35776 = -42.35776
Month 8: 310 - 271.886208 = 38.113792
Month 9: 240 - 265.5089664 = -25.5089664
Month 10: 310 - 274.40717312 = 35.59282688
Month 11: 240 - 267.525738496 = -27.525738496
Month 12: 230 - 260.020590797 = -30.020590797
MSE: [88^2 + (-25.6^2) + 3.52^2 + 18.816^2 + 47.0528^2 + (-42.35776^2) + 38.113792^2 +
(-25.5089664^2) + 35.59282688^2 + (-27.525738496^2) + (-30.020590797^2)] / 11
=1636.55
Comparison: Exponential smoothing with α = 0.2 (MSE =1636.55) provides more accurate
forecasts than the three-month moving average (MSE= 2221.23) based on MSE
c)
Three-Month Moving Average Forecast: (310 + 240 + 230) / 3 = 260
Exponential Smoothing Forecast (α = 0.2): F13 = 0.2 * 230 + 0.8 * 260.020590797 = 46 +
208.016472638 = 254.016472638
Forecast: The exponential smoothing forecast of 254.02 is recommended due to its lower
MSE
15/806
b)
F2 = 0.2 * 7.40 + 0.8 * 7.35 = 1.48 + 5.88 = 7.36
F3 = 0.2 * 7.55 + 0.8 * 7.36 = 1.51 + 5.888 = 7.398
F4 = 0.2 * 7.56 + 0.8 * 7.398 = 1.512 + 5.9184 = 7.4304
F5 = 0.2 * 7.60 + 0.8 * 7.4304 = 1.52 + 5.94432 = 7.46432
F6 = 0.2 * 7.52 + 0.8 * 7.46432 = 1.504 + 5.971456 = 7.475456
F7 = 0.2 * 7.52 + 0.8 * 7.475456 = 1.504 + 5.9803648 = 7.4843648
F8 = 0.2 * 7.70 + 0.8 * 7.4843648 = 1.54 + 5.98749184 = 7.52749184
F9 = 0.2 * 7.62 + 0.8 * 7.52749184 = 1.524 + 6.021993472 = 7.545993472
F10 = 0.2 * 7.55 + 0.8 * 7.545993472 = 1.51 + 6.0367947776 = 7.5467947776
F11 = 0.2 * 7.55 + 0.8 * 7.5467947776 = 1.51 + 6.03743582208 = 7.54743582208
c)
F2 = 0.3 * 7.40 + 0.7 * 7.35 = 2.22 + 5.145 = 7.365
F3 = 0.3 * 7.55 + 0.7 * 7.365 = 2.265 + 5.1555 = 7.4205
F4 = 0.3 * 7.56 + 0.7 * 7.4205 = 2.268 + 5.19435 = 7.46235
F5 = 0.3 * 7.60 + 0.7 * 7.46235 = 2.28 + 5.323645 = 7.603645
F6 = 0.3 * 7.52 + 0.7 * 7.603645 = 2.256 + 5.3225515 = 7.5785515
F7 = 0.3 * 7.52 + 0.7 * 7.5785515 = 2.256 + 5.30498605 = 7.56098605
F8 = 0.3 * 7.70 + 0.7 * 7.56098605 = 2.31 + 5.292690235 = 7.602690235
F9 = 0.3 * 7.62 + 0.7 * 7.602690235 = 2.286 + 5.3218831645 = 7.6078831645
F10 = 0.3 * 7.55 + 0.7 * 7.6078831645 = 2.265 + 5.32551821515 = 7.59051821515
F11 = 0.3 * 7.55 + 0.7 * 7.59051821515 = 2.265 + 5.313362650615 = 7.578362650615
d)
Errors for α = 0.2:
Week 2: 7.40 - 7.36 = 0.04
Week 3: 7.55 - 7.398 = 0.152
Week 4: 7.56 - 7.4304 = 0.1296
Week 5: 7.60 - 7.46432 = 0.13568
Week 6: 7.52 - 7.475456 = 0.044544
Week 7: 7.52 - 7.4843648 = 0.0356352
Week 8: 7.70 - 7.52749184 = 0.17250816
Week 9: 7.62 - 7.545993472 = 0.074006528
Week 10: 7.55 - 7.5467947776 = 0.0032052224
- MSE for α = 0.2: (0.04^2 + 0.152^2 + 0.1296^2 + 0.13568^2 + 0.044544^2 +
0.0356352^2 + 0.17250816^2 + 0.074006528^2 + 0.0032052224^2) / 9= 0.010978
- Errors for α = 0.3:
Week 2: 7.40 - 7.365 = 0.035
Week 3: 7.55 - 7.4205 = 0.1295
Week 4: 7.56 - 7.46235 = 0.09765
Week 5: 7.60 - 7.603645 = -0.003645
Week 6: 7.52 - 7.5785515 = -0.0585515
Week 7: 7.52 - 7.56098605 = -0.04098605
Week 8: 7.70 - 7.602690235 = 0.097309765
Week 9: 7.62 - 7.6078831645 = 0.0121168355
Week 10: 7.55 - 7.59051821515 = -0.04051821515
- MSE for α = 0.3: [0.035^2 + 0.1295^2 + 0.09765^2 + (-0.003645^2) +
(-0.0585515^2) + (-0.04098605^2) + (0.097309765^2) + (0.0121168355^2) +
(-0.04051821515^2)] / 9= 0.004988
- Comparison: MSE for α = 0.3 (= 0.004988) is lower than MSE for α = 0.2
(=0.010978), indicating α = 0.3 provides more accurate forecasts
- Forecast for Week 11: Use α = 0.3 forecast: 7.578362650615
