Chapter 15: Forecasting and Reading Homework
What constitutes a good forecast?
- the forecast should be unbiased; that is, be correct on average. - the forecast should come close to the real outcomes as measured by the mean squared error (MSE) or the mean absolute error (MAE).
Select all the average forecast errors for a biased forecast.
> 0 < 0
unbiased forecast
A forecast that is correct on average, thus an average forecast error equal to zero.
biased forecast
A forecast that is wrong on average, thus an average forecast error different from zero
Mean Squared Error (MSE)
A measure evaluating the quality of a forecast by looking at the average squared forecast error
Forecast Gaming
A purposeful manipulation of a forecast to obtain a certain decision outcome for a decision that is based on the forecast.
Double Exponential Smoothing
A way of forecasting a demand process with a trend that estimates the trend using exponential smoothing and then also uses exponential smoothing to estimate the demand rate net of the trend.
Forecast with consensus building
An iterative discussion among experts about their forecasts and opinions that leads to a single forecast.
True or false: A time series-based forecast of demand will incorporate the "gut feel" of an expert.
False reason: a time series-based forecast only uses historical demand
expert panel forecasting
Forecasts generalized using the subjective opinions of management.
automated forecasting
Forecasts that are created by computers, typically with no human intervention.
long-term forecasts
Forecasts used to support strategic decisions with typical time ranges of multiple years.
Deseasonalize
To remove the seasonal effect from the past data.
Statistical noise
Variables influencing the outcomes of a process in unpredictable ways.
prediction market
a betting game in which forecasters can place financial bets on their forecasts
trend
a continuing increase or decrease in a variable that is consistent over a long period of time
time series-based forecast
a forecast that is obtained based on nothing but old demand data.
Exponential Smoothing
a forecasting method that predicts that the next value will be a weighted average between the last realized value and the old forecast.
naive forecasting method
a forecasting method that predicts that the next value will be like the last realized value
moving average forecast
a forecasting method that predicts that the next value will be the average of the last realized values.
Seasonality
a significant demand change that constitutes a repetitive fluctuation over time
The MAE takes the _____________ values of the forecast errors to ensure they don't cancel each other out.
absolute
MAE MAPE
absolute terms relative terms
momentum-based forecasts
an approach to forecasting that assumes that the trend in the future will be similar to the trend in the past.
Trends are still detected based upon historical data and _______________ guaranteed to continue in the future.
are not
For forecasts that need to be done thousands or millions of times, which type of forecasting is most appropriate?
automated
A forecast that consistently overestimates demand (has a positive forecast error) is _______.
biased
The moving average ___________ forecast a value greater than all historical values seen to date.
can not
Statistical noise ______ be forecasted.
cannot
Forecast combination
combining multiple forecasts that have been generated by different forecasters into one single value
Automated forecasts are typically created by _____________
computers
short-term time horizon mid-term time horizon long-term time horizon
daily to monthly monthly to yearly multiple years
The mean absolute percentage error divides the forecast errors by the actual
demands
The purpose of _______________ data is to try and understand whether a past period was very busy because it was a period of time that is typically very busy (e.g. a coffee shop at 8 am) or there was another factor driving demand.
deseasonalizing
The seasonality index is calculated by ______________ the average demand for the time period that constitutes seasonality by the average total demand.
dividing
The naive, moving average, and exponential smoothing forecasting methods _____________ detect trends or seasonalities.
do not
Deseasonalizing data __________ remove statistical noise
does not
Double exponential smoothing ______________ work well when there is a seasonality pattern to the data
does not
extrapolation
estimation of values beyond the range of the original observations by assuming that some patterns in the values present within the range will also prevail outside the range.
The idea of ___________ smoothing is to put more weight on recent data and less weight on older data.
exponential
The idea of _____________ smoothing is to put more weight on recent data and less weight on older data.
exponential
True or false: The MAE and MSE will always agree on which forecast has the highest error.
false
True or false: The more unbiased a forecast is the smaller the mean squared error.
false
forecast error in t
forecast for t - actual value for t
Demand ____________ is the process of creating statements about future realizations of demand.
forecasting
_________________ is the process of creating statements about outcomes of variables that will only be realized in the future and are currently uncertain.
forecasting
Mid-term forecasts
forecasts used to support capacity planning and financial accounting with typical time ranges from weeks to a year
short-term forecasts
forecasts used to support tactical decision making with typical time ranges from hours to weeks.
One method for choosing the value of the exponential smoothing parameter is to find the value that would have yielded the best forecast for a set of ______________ data.
historical, past, or old
In regression analysis, the variables that influence the dependent variable are the _________________ variables.
independent
The closer the forecast value is to the actual value, the ___________ the forecast error.
lower
When the seasonality index is less than one it indicates that demand during that season is ___________ than average.
lower
Suppose a forecast over-estimates demand by 5,000 units on day 1 and under-estimated demand by 5,000 units on day 2. Which metric better reflects the accuracy of the forecast?
mean squared error
Forecast error in a given period equals the forecast for that period __________ the actual demand value for that period.
minus
The smaller the value of alpha, the _____ smooth the forecast.
more
The smaller the value of alpha, the __________ smooth the forecast.
more
older data newer data
more weight less weight
Reseasonalizing requires _____________ the smoothed forecast for an average day by the appropriate seasonality index.
multiplying
The ___________ forecasting method forecasts next period's demand by using the last realized value.
naive
Taking the moving average reduces the effect of statistical
noise
The statistical ____________ in the demand for a process is the amount of demand that is purely a result of randomness and could not have been forecasted.
noise
Forecasts are often automated when they need to be performed ____________.
often, maybe hundreds of times a day
The naive forecasting method uses how many old data points?
one
When alpha equals _____________ the exponential smoothing forecast becomes the naive forecasting method.
one
When incorporating seasonality into a forecast the first step is to determine the seasonality ___________________.
pattern
When incorporating seasonality into a forecast the second step is to ____________ the effect of seasonality.
quantify
_____________ analysis estimates the relationship of one variable with multiple variables that influence this one variable.
regression
__________________ analysis estimates the relationship of one variable with multiple variables that influence this one variable
regression
Deseasonalizing demand in a period requires dividing the demand in the period by the _____________ index.
seasonality
That a coffee shop regularly sees more customers from 8 to 9 am than from 10 am to 11 am is an example of ____________.
seasonality
That sales for snow shovels are typically much higher in the winter is an example of _______________.
seasonality
The _______________ index is used to deseasonalize old demand, or, to remove the seasonal effect from the data.
seasonality
Unlike statistical noise, ________________ is a pattern we expect to continue into the future.
seasonality
Time _________________ analysis is the process of analyzing old data.
series
The exponential smoothing forecast multiplies the current demand by a _____________ parameter that is between zero and one.
smoothing
The exponential smoothing forecast multiplies the current demand by a ______________ parameter that is between zero and one.
smoothing
A good forecast should come close to the real outcomes as measured by the mean _________ error.
squared or absolute
By ___________________ the errors when calculating the MSE the errors never cancel each other out
squaring
short-term decision mid-term decision long-term decision
staffing levels and scheduling recruiting and machine acquisition entering new markets or launching new products
Regressions analysis is based on ____________________.
statistical analysis
regression analysis
statistical process of estimating the relationship of one variable with multiple variables that influence this one variable
A forecasting method that allows for human input is often referred to as a(n) ____________ forecast.
subjective
Sometimes you want to allow for human input into the forecasting method. In such case, we speak about creating a...
subjective forecast
forecast error
the difference between a forecasted and the realized demand
Overconfidence
the fact that human decision makers are overly confident in their ability to shape a positive outcome.
anchoring bias
the fact that human decision makers are selective in their acquisition of new information, looking for what confirms their initially held beliefs
smoothing parameter
the parameter that determines the weight new realized data have in creating the next forecast with exponential smoothing
time series analysis
the process of analyzing the old (demand) data y1...yt.
demand forecasting
the process of creating statements about future realizations of demand.
Forecasting
the process of creating statements about outcomes of variables that presently are uncertain and will only be realized in the future.
dependent variable (outcome variable)
the variable that we try to explain in a regression analysis
independent variable
the variables influencing the dependent variable
Reseasonalize
to reintroduce the seasonal effect to the forecasted data
Forecast for t+1 considering trend = Forecast for (t+1) + Forecast for _________ in (t+1)
trend
Forecast for t+1 considering trend = Forecast for (t+1) + Forecast for __________________ in (t+1)
trend
It is called double exponential smoothing because exponential smoothing is used to smooth both the demand and the ___________ forecast.
trend
It is called double exponential smoothing because exponential smoothing is used to smooth both the demand and the _______________ forecast.
trend
Unlike moving average or exponential smoothing, detecting a ____________ requires looking for increases or decreases in historical data.
trend
To forecast demand in period t+1 while considering a trend involves how many terms on the right-hand side of the equation?
two
The periods that the moving average forecast is based on are known as the forecast _____________.
window