Forecasting: Common Features
Forecasting enables a manager to anticipate/predict future
demand, sale, production, etc. so s/he can plan accordingly.
It assumes that the same underlying causal system
that existed in the past will continue to exist in the future.
It is rarely perfect; actual results usually differ from the
predicted values, owing to the presence of randomness.
It tends to be more accurate for a group of items than
for individual items, due to cancelling effect in the group.
Its accuracy decreases as the time horizon increases;
short-range forecasts usually tend to be more accurate.
Forecasting: Elements of a Good Forecast
The forecast must be timely;
It must be accurate; its accuracy should be stated;
It has to be reliable, and should work consistently;
Its technique should be simple to understand and use;
It has to be expressed in meaningful units;
It should be in writing, increasing use of same info by all.
Forecasting: Steps in Forecasting Process
Determine the purpose of the forecast; when/how to use;
Establish a time horizon, indicating the time interval(s);
Gather and analyze relevant data before forecasting;
Select a forecasting technique suitable for the data;
Make the forecast, using the selected technique(s);
Monitor the forecast to determine its performance.
Forecasting: Techniques
Judgmental: Rely on analyses of subjective inputs, such
as opinions from consumers, sales staff, executives, etc.
Associative Models: Use techniques consisting of one or
more explanatory variables to predict a future demand.
Time-series: Ident project patterns in recent time-ordered
sequence (using historical data), assuming future like past.
Delphi: A series of relevant questionnaires answered by
experts, managers, staff, with/without info to each other.
Forecasting: Methods
Qualitative: Causal (Judgmental, Associative);
Quantitative: Time-series (Averaging/Smoothing, Trend
projection, Trend projection adjusted for Seasonal
influence, etc.), Delphi, etc.
Naive Method: Simplest, but widely used approach. Uses a
single previous value of time-series as the basis of
Example:
forecast.
Period Actual Change from Previous Value Forecast
Previous, t-1 40
Current, t 60 +20
Next, t+1 60 + 20 = 80
Time-Series Components
Trend: A long-term upward/downward movement in data;
Seasonality: Short-term, fairly regular variations related to
the time of the calendar day, or week, month, quarter, etc.
Cycles: Wave-like variations, lasting for more than a year;
Irregular Variations: Caused by unusual circumstances
not reflective of typical behaviour, e.g. strike, weather, etc;
Random Variations: The residual variations that remain
after all other behaviours have been accounted for.
Averaging/Smoothing Methods
Moving Average: A method that averages a number of
recent actual values, updated as new values are available.
Thus, the forecast for time-period t, Ft = MAn = (At-1)/n,
where, Ft = Forecast value for time period t;
MAn = moving average for n no. of periods;
i = index corresponding to time-period (i = 1~n);
n = no. of periods in moving averages;
At-1 = actual value in period t-1.
Moving Average
(continued)
Examples: Compute a three-period moving average forecast,
given the demand for shopping carts for the last five periods.
Period Demand Solution:
t y
1 44 Taking the three most recent demands,
2 42 F6 = (41 + 40 + 43) / 3 = 124 / 3 = 41.333 nos.
3 43 Again, if the actual demand in period 6 turns
out to be 38,
4 40 F7 = (38 + 41 + 40) / 3 = 119 / 3 = 39.667 nos.
5 41 Ft updated, as the new actual value found.
Forecast Accuracy
Three common measures for summing historical errors are:
A. Mean Absolute Deviation (MAD): The average
absolute forecast error;
MAD = [Actualt - Forecastt] / n, where Et = At Ft;
B. Mean Squared Error (MSE): The average of the
squared forecast errors;
MSE = [Actualt - Forecastt] / (n 1), where n = pds;
C. Mean Absolute Percentage Error (MAPE): The
average absolute percentage error;
MAPE = [Actualt - Forecastt/ Actualt * 100]% / n.
Forecast Accuracy
Example: Compute MAD, MSE and MAPE for the following data:
Period Actual Forecast E=A-F E E/A*100 Solution:
t A F
1 217 215 2 4 0.922% A. MAD
2 213 216 3 9 1.408 = E/ n
3 216 215 1 1 0.463 = 22 / 8 = 2.750
4 210 214 4 16 1.905 B. MSE
5 213 211 2 4 0.939 = E / (n - 1)
6 219 214 5 25 2.283 = 76 / 7 = 10.857
7 216 217 1 1 0.463 C. MAPE
8 212 216 4 16 1.887 =[E/A*100]/n
= - - 22 76 10.27% =10.27/8 =1.284%
