ABC computer company is engaged in repairs and service of computers. It is trying to forecast the demand for next 12 months, based on the accumulated demand data as given below. The company wants to follow exponential smoothing forecasts using smoothing constant (α) equal to 0·30 and 0·50 :
Demand for Repair
Period | Month | Demand
1 | January | 40
2 | February | 43
3 | March | 44
4 | April | 40
5 | May | 45
6 | June | 50
7 | July | 45
8 | August | 48
9 | September | 57
10 | October | 55
11 | November | 56
12 | December | 55
The forecast for period one may be taken as 40. The present your result in a table.
20 marks · directive: Forecast
The core of a good answer
With F1 = 40, simple exponential smoothing gives a forecast for period 13 (the next January) of about 53.19 repairs with α = 0.30 and 54.87 with α = 0.50; α = 0.50 tracks the rising demand better (MAD 3.20 against 3.97).
An outline written by upsc.help, not an official UPSC answer; every fact checked by a separate review. How we work
What the question asks
Apply the simple exponential smoothing recurrence F(t+1) = F(t) + α[D(t) - F(t)] period by period for both values of α, starting from F1 = 40, and tabulate demand, forecasts and errors. Comparing the mean absolute deviation shows which constant suits this data.
Answer outline
Introduction
Given: monthly repair demand for 12 periods, F1 = 40, α = 0.30 and 0.50. Forecasts are rounded to two decimals for display; the working carries full precision.
Step 1: Formula
- F(t+1) = α D(t) + (1 - α) F(t) = F(t) + α [D(t) - F(t)].
Step 2: First few periods, α = 0.30
- F2 = 40 + 0.3(40 - 40) = 40.00
- F3 = 40 + 0.3(43 - 40) = 40.90
- F4 = 40.90 + 0.3(44 - 40.90) = 41.83
- F5 = 41.83 + 0.3(40 - 41.83) = 41.28
Step 3: First few periods, α = 0.50
- F2 = 40 + 0.5(40 - 40) = 40.00
- F3 = 40 + 0.5(43 - 40) = 41.50
- F4 = 41.50 + 0.5(44 - 41.50) = 42.75
- F5 = 42.75 + 0.5(40 - 42.75) = 41.38
Step 4: Results table
- Period 1 (Jan): D = 40; F(α=0.30) = 40.00; F(α=0.50) = 40.00
- Period 2 (Feb): D = 43; F(α=0.30) = 40.00; F(α=0.50) = 40.00
- Period 3 (Mar): D = 44; F(α=0.30) = 40.90; F(α=0.50) = 41.50
- Period 4 (Apr): D = 40; F(α=0.30) = 41.83; F(α=0.50) = 42.75
- Period 5 (May): D = 45; F(α=0.30) = 41.28; F(α=0.50) = 41.38
- Period 6 (Jun): D = 50; F(α=0.30) = 42.40; F(α=0.50) = 43.19
- Period 7 (Jul): D = 45; F(α=0.30) = 44.68; F(α=0.50) = 46.59
- Period 8 (Aug): D = 48; F(α=0.30) = 44.77; F(α=0.50) = 45.80
- Period 9 (Sep): D = 57; F(α=0.30) = 45.74; F(α=0.50) = 46.90
- Period 10 (Oct): D = 55; F(α=0.30) = 49.12; F(α=0.50) = 51.95
- Period 11 (Nov): D = 56; F(α=0.30) = 50.88; F(α=0.50) = 53.47
- Period 12 (Dec): D = 55; F(α=0.30) = 52.42; F(α=0.50) = 54.74
- Period 13 (next January): F(α=0.30) = 53.19; F(α=0.50) = 54.87
Step 5: Accuracy
- α = 0.30: sum of absolute errors over periods 1 to 12 = 47.64, MAD = 3.97; sum of errors (bias) = +43.98.
- α = 0.50: sum of absolute errors = 38.42, MAD = 3.20; sum of errors = +29.74.
- Both forecasts lag because demand trends upward (positive bias); the higher α reacts faster.
Conclusion
Forecast for the next period: 53.19 (α = 0.30) and 54.87 (α = 0.50). Simple exponential smoothing gives a flat forecast, so the same value serves for each of the next 12 months until new actuals arrive; with a clear trend, Holt's trend-adjusted smoothing would be better. Check: all values were recomputed in Python.
Examples and data to cite
- The demand rises from about 40 to about 55 over the year, which is why both series under-forecast in the second half.
Facts to remember
- Simple exponential smoothing: F(t+1) = α D(t) + (1 - α) F(t), 0 < α < 1.
- Weights on past demand decline geometrically: α, α(1 - α), α(1 - α)², ...
- Higher α gives more weight to recent demand and responds faster; lower α smooths random variation more.
- MAD = Σ |D - F| / n; bias (running sum of forecast errors) shows systematic under or over forecasting.
- Trend-adjusted (Holt) smoothing adds a second constant β for the trend.