Prelim 2 – Module 5: Forecasting

Question 1

Method: Historical Analogy
Type:
Data Required:
Relative Cost:
Horizon:

Answer 1

Type: subjective
Data Required: experience
Relative Cost: high
Horizon: medium to long

Question 2

Method: Regression
Type:
Data Required:
Relative Cost:
Horizon:

Answer 2

Type: causal
Data Required: all past data
Relative Cost: medium
Horizon: medium

Question 3

Method: Moving average
Type:
Data Required:
Relative Cost:
Horizon:

Answer 3

Type: time series
Data Required: recent data
Relative Cost: very low
Horizon: short

Question 4

Method: Exponential Smoothing
Type:
Data Required:
Relative Cost:
Horizon:

Answer 4

Type: time series
Data Required: last forecast and smoothed value
Relative Cost: very low
Horizon: short

Question 5

How do we calculate an N-period moving average forecast?
-> what is MAt
-> what is At

Answer 5

MAt = the N-period moving average at the end of period T

At = actual observation for period T

Question 6

Forecast formula

Answer 6

Ft+1 = MAt

Question 7

Strengths of a moving average forecast

Answer 7

1. only need N observations to make a forecast
2. very inexpensive and easy to understand

Question 8

Drawbacks of a moving average forecast

Answer 8

1. does not consider observations older than N periods
2. gives equal weight to last N observations

Question 9

Exponential smoothing advantages

Answer 9

1. old data are never dropped but have progressively less influence
2. don’t need to keep any historical information; only need most recent smoother value

Question 10

St =

Answer 10

Smoothed value at end of period T

𝑆𝑇=𝑆(𝑇−1)+𝛼∗(𝐴𝑇−𝐹𝑇 )=(1−𝛼)∗𝑆(𝑇−1)+𝛼∗𝐴𝑇

Question 11

At =

Answer 11

Actual observation for period T

Question 12

Ft+1 = St

Answer 12

Forecast for period T+1

Question 13

New smoothed value =

Answer 13

Old smoothed value + α*observed error

Question 14

Which of the following two statements is true?
1. When alpha is small (i.e., closer to 0), older data points have more weight in determining the forecast.
2. When alpha is large (i.e., closer to 1), older data points have more weight in determining the forecast.

Answer 14

1. When alpha is small (i.e., closer to 0), older data points have more weight in determining the forecast.

Question 15

𝛼 = 1

Answer 15

Naive forecast (i.e., use last actual value)

Question 16

When is the weighted average age of your data the same under exponential smoothing and moving average forecasts?

Answer 16

WA of data under exponential smoothing is usually lower than moving average forecasts.

They can be made equal by adjusting the smoothing parameter in exponential smoothing to match the time span of the moving average window.

Question 17

How do we measure forecast accuracy?

Answer 17

Cumulative forecast error (CFE)
Mean absolute deviation (MAD)
Mean squared error (MSE)
Mean absolute percentage error (MAPE)

Question 18

Cumulative forecast error (CFE) purpose

Answer 18

Should be close to zero. Otherwise, forecast is biased.

Question 19

Mean absolute deviation (MAD) purpose

Answer 19

Common measure of error. Gives equal weight to all errors.

Question 20

Mean squared error (MSE) purpose

Answer 20

Gives extra penalty to large error values

Question 21

Mean absolute percentage error (MAPE) purpose

Answer 21

Use percentages to prevent large forecast values from dominating the accuracy measure.

Question 22

Exponential smoothing with seasonal adjustment

Data has a ___

Answer 22

cycle of L periods (hours, days, months, seasons, etc.)

Question 23

Exponential smoothing with seasonal adjustment use seasonality index lt to ___

Answer 23

deseasonalize data in each cycle

Question 24

How to use Solver to optimize smoothing parametres

Answer 24

Set objective: choose the performance score that you want to minimize

By changing variable cells: choose your smoothing parametre or smoothing parametres

subject to the constraints:
- all smoothing parametres <= 1
- all smoothing parametres >= 0

Use GRG Nonlinear solving method

Question 25

Optimizing the smoothing parameters minimizes whichever metric we choose ___

Answer 25

on the historical data

Question 26

Overfitting

Answer 26

An overly complex model that tells you more about the idiosyncrasies of historical data without having value for future predictions

Question 27

How do we avoid overfitting to pick among various candidate models?

Answer 27

1. divide our data into training and testing data sets
2. fit our candidate models based on the training dataset
3. see how they perform on the testing dataset