chapter 3 powerpoint: forecasting Flashcards
A demand forecast
is an estimate of demand expected over a future time period
3 Uses for Forecasts
Design the System
Use of the System
Schedule the System
Features of Forecasts
Assumes causal system(past ==> future)
Forecasts rarely perfect because of randomness
Forecasts more accurate forgroups vs. individuals
Forecast accuracy decreases as time horizon increases
Elements of a Good Forecast
reliable
meaningful
compatible
useful time horizon
easy to understand and use
Steps in the Forecasting Process
- Determine purpose of forecast
- Establish a time horizon
- Select a forecasting technique
- Obtain, clean and analyze data
- Make the forecast
- Monitor the forecast
Approaches to Forecasting
Judgmental forecasting
Quantitative forecasting
Judgmental forecasting
non-quantitative analysis of subjective inputs
considers “soft” information such as human factors, experience, gut instinct
what do we use for quantitative forecasting
Time series models
–> extends historical patterns of numerical data
Associative models
–> create equations with explanatory variables to predict the future
Judgmental forecasting methods
Executive opinions
Expert opinions
Sales force opinions
Consumer surveys
Historical analogies
Executive opinions
pool opinions of high-level executives
long term strategic or new product development
Expert opinions
Delphi method
technological forecasting
Delphi method
iterative questionnaires circulated until consensus is reached
Sales force opinions
based on direct customer contact
Consumer surveys
questionnaires or focus groups
Historical analogies
use demand for a similar product
What is a Time Series?
a time ordered sequence of observations
the 6 patterns of time series
Level
Trend
Seasonality
Cycles
Irregular variations
Random variations
Level
(average) horizontal pattern
Trend
steady upward or downward movement
Seasonality
regular variations related to time of year or day
Cycles
wavelike variations lasting more than one year
Irregular variations
caused by unusual circumstances, not reflective of typical behaviour
Time series models
Naive methods
Averaging methods
Trend models
Techniques for seasonality
Averaging methods
Moving average
Weighted moving average
Exponential smoothing
Trend models
Linear and non-linear trend
Trend adjusted exponential smoothing
Techniques for seasonality
Techniques for cycles
Naive Methods
Next period = last period
if there is a trend, follow the trend
Simple to use and understand
Very low cost
Low accuracy
Moving average
Forecast = (EActual) / n
average of last few actual data values, updated each period
fewer data points = more sensitive to changes
more data points = smoother, less responsive
Weighted moving average
ex: 0.5 · 36 + 0.3 · 32 + 0.2 · 38
usually, the most recent actual demand is the one with heaviest weight
Exponential smoothing
Ft = Ft-1 + a(At-1 - Ft-1)
sophisticated weighted moving average
weights decline exponentially
most recent data weighted most
subjectively choose smoothing constant a which ranges from 0 to 1
when do we use a smaller smoothing constant (a) in the exponential smoothing?
When demand is fairly stable
smoothes out random fluctuations
when do we use a higher smoothing constant (a) in the exponential smoothing?
When demand increasing or decreasing
more responsive to real changes
True or False?
A moving average forecast tends to be more responsive to changes in the data series when more data points are included in the average
False
True or False?
As compared to a simple moving average, the weighted moving average is more reflective of the recent changes
True
True or False?
A smoothing constant of .1 will cause an exponential smoothing forecast to react more quickly to a sudden change than a value of .3 will
False
Techniques for Trend
Develop an equation that describes the trend
Look at historical data
Linear Trend Equation
yt = a + bt
b = n(Eyt -Et · Ey) / n(Et^2 - (Et)^2)
a = (Ey - bEt) / n
Trend-Adjusted Exponential Smoothing
a = smoothing constant for average
B = smoothing constant for trend
estimate starting smoothed average and smoothed trend by using most recent data
Trend-Adjusted Exponential Smoothing formula
TAFt+1 = St + Tt
St = TAFt + a(At - TAFt)
Tt = Tt-1 + B(st - St-1 - Tt-1)
Techniques for Seasonality
Additive or Multiplicative Model
Additive Model
Demand = Trend + Seasonality
Multiplicative Model
Demand = Trend x Seasonality
Seasonal Relative (or index)
= proportion of average or trend for a season in the multiplicative model
ex: seasonal relative of 1.2 = 20% above average
Deseasonalizing
removing seasonal component to more clearly see other components
dividing by seasonal relative
Reseasonalizing
adjusting the forecast for seasonal component
multiplying by seasonal relative
Times Series Decomposition
- Compute the seasonal relatives.
- Deseasonalize the demand data.
- Fit a model to deseasonalized demand data,
- -> e.g., moving average or trend. - Forecast using this model and the deseasonalized demand data.
- Reseasonalize the deseasonalized forecasts.
firecast error
Actual value - Forecast value
positive is due to a forecast that was too low compared to actual
negative is due to a forecast that was too high compared to actual
Three measures of forecasts errors are used
Mean absolute deviation (MAD)
Mean squared error (MSE)
Mean absolute percent error (MAPE)
Control charts
plot errors to see if within pre-set control limits
A visual tool for monitoring forecast errors
Used to detect non-randomness in errors
Set limits that are multiples of the √MSE
Tracking signal
Ratio of cumulative error and MAD
Mean absolute deviation (MAD)
(E|Actual - Forecast|) / n
Easy to compute
Weights errors linearly
Mean squared error (MSE)
((Actual - Forecast)^2) / n
Squares error
More weight to large errors
Mean absolute percent error (MAPE)
(E[|Actual - Forecast| / Actual] / n
Puts errors in perspective
above 70% satisfactory
bias
the sum of the forecast errors
positive bias = frequent underestimation
negative bias = frequent overestimation
possible sources of error include:
Model may be inadequate (things have changed)
Incorrect use of forecasting technique
Irregular variations
when are forecasting errors “in control”?
when only random errors are present
no errors from identifiable causes
All errors are within control limits
No patterns (e.g. trends or cycles) are present
errors outside limit = need corrective action
Control Limits
Standard deviation of error = s = √MSE
control limits = 2s
68% of all errors should be within 1s
95% of all errors should be within 2s
99.7% of all errors should be within 3s
Tracking signal
ratio of cumulative error to MAD
can be plotted on a control chart
investigate if Tracking Signal > 4
Tracking Signal = E(Actual - forecast) / MAD
True or False?
When error values fall outside the limits of a control chart, this signals a need for corrective action
True
True or False?
When all errors plotted on a control chart are either all positive, or all negative, this shows that the forecasting technique is performing adequately
False
True or False?
A random pattern of errors within the limits of a control chart signals a need for corrective action.
False
Two most important factors to choosing a forecasting technique?
Cost
Accuracy