Title: Forecasting
1Forecasting
- I think there is a world market for maybe five
computers. Thomas Watson, Chairman of IBM, 1943. - I skate to where the puck is going to be, not to
where it has been. Wayne Gretzky, the greatest
hockey player of all times. - A good forecaster is not smarter than anyone
else, he merely has his ignorance better
organized. Consultants survival tips.
2Forecasting
- Product demand is the basis for all planning
activities in a firm. Product demand drives the
requirements for - Raw material
- Capacity
- People
- However, future demand is almost never certain.
It needs to be predicted using - Past sales history
- Causal factors
- Human judgment
- This is called forecasting (also known as demand
planning)
3Positioning of forecasting process
ISCM, i2 technologies, inc.
4The use of forecasting
5Independent dependent demand
6Characteristics of forecasts
- They are usually wrong
- A good forecast is more than a single number
- Aggregate forecasts are more accurate
- The longer the forecast horizon, the less
accurate the forecast will be - Forecasts should not be used to the exclusion of
known information
7Types of forecasts
- Subjective (qualitative) methods
- Sales force composites
- Customer surveys
- Jury of executive opinion
- Delphi method
- Objective (quantitative) methods
- Causal models
- Time series models
- Blended methods
- A baseline forecast created by quantitative
methods to be adjusted by the user using
subjective methods
8Causal methods
- Predict future using data from other sources than
the series being predicted - Predict spare parts demand using the data about
install base - Predict housing sales using the interest rate for
mortgages - Causal models
- Yf(X1,X2,,Xn)
- Ya0a1X1a2X2anXn
9Time series components
Cycles Randomness
- Average
- Trend
- Seasonality
Seasonal variation
Linear Trend
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
Sales
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
1
2
3
4
Year
10Forecasting notation
- D1, D2,,Dt the observed values of demand during
periods 1,2,,t - If we are forecasting at the end of period t,
- We observed Dt, Dt-1,Dt-2,,
- We did not observe Dt1, Dt2,,
- Ft is the forecast made for period t in period t-1
11Methods for stationary series
- Time series forecasting where the trend,
seasonality or cyclic behavior is hard to predict
or model. - Each observation is a constant plus a random
fluctuation -
-
-
- Mainly two methods
- Moving average
- Exponential smoothing
12Simple Moving Average Formula
- The simple moving average model assumes an
average is a good estimator of future behavior. - The formula for the simple moving average is
N number of periods used in averaging MA(N)
13Example
F4(650678720)/3
F7(650678720 785859920)/6
14Moving average forecast
Weighted moving average
- Multiple-step-ahead forecast
- If you are at the end of period 3, and if you
want to forecast period 6, the forecast would be
same as the forecast for period 4 (stationary
series) - Rounding
- If the numbers are small, a carry forward
rounding (CFR) scheme can be used - Consider the series (,1,0,1,1) and the MA(4) is
used, the forecast for future can be one of - 1,1,0,1,1,1,0,1,1,1,0,.
- 1,0,1,1,1,0,1,1,1,0,1,.
15Finding the right forecasting method
ISCM, i2 technologies, inc.
16Evaluating forecasts
- Forecast error for period, et Ft - Dt
17Forecast errors over time
18Moving average lags behind
19Moving average lags behind
Demand
Is moving average forecast an unbiased forecast,
if there is a known trend?
20Exponential smoothing
- The current forecast is weighted average of the
last forecast and last observed value - New forecast Ft
- a (Current observation of demand)(1-a)(last
forecast) - a (Dt-1)(1-a)(Ft-1)
- a smoothing constant, weight on current
observation - (1- a) weight placed on past observations
- FtFt-1- a(Ft-1-Dt-1)Ft-1- a et-1
21Exponential Smoothing Problem
- Question Given the weekly demand data, what are
the exponential smoothing forecasts for periods
2-10 using a0.10 and a0.60? - Assume F1D1
22Exponential Smoothing Problem
F40.1680(1-0.1)(815.5)
23Smoother forecast with small alpha
24Weights in exponential smoothing
- Ft a Dt-1(1-a)(Ft-1)
- Ft-1 a Dt-2(1-a)(Ft-2)
- Ft a Dt-1 a (1-a)(Dt-2)(1- a)2 Ft-2
25Weights in exponential smoothing
26Example
- Engine failures in a military base
Compare MA(3) and ES(0.1)
27Comparison of two methods
28Comparison of exponential smoothing and moving
average
- Similarities
- Stationary series assumption
- Single parameter
- Lag behind if there is a trend
- When a2/(N1) gives same level of accuracy
- Differences
- Exponential smoothing uses all past data, moving
average uses only the last N - Exponential smoothing require less data storage