ETF3231/5231:
Business forecasting
Ch1. Getting started
OTexts.org/fpp3/
Outline
1 What can we forecast?
2 Time series data and random futures
3 Some case studies
2
Outline
1 What can we forecast?
2 Time series data and random futures
3 Some case studies
3
Forecasting is difficult
4
Reputations can be made and lost
“There’s no chance that the iPhone is going to get any significant
market share. No chance.”
(Steve Ballmer, CEO Microsoft, April 2007)
5
Reputations can be made and lost
“There’s no chance that the iPhone is going to get any significant
market share. No chance.”
(Steve Ballmer, CEO Microsoft, April 2007)
“We’re going to be opening relatively soon . . . The virus . . . will go
away in April.”
(Donald Trump, February 2020)
5
Reputations can be made and lost
“There’s no chance that the iPhone is going to get any significant
market share. No chance.”
(Steve Ballmer, CEO Microsoft, April 2007)
“We’re going to be opening relatively soon . . . The virus . . . will go
away in April.”
(Donald Trump, February 2020)
There is a major difference between the first one and the second one.
What is it?
5
Forecasts that are not forecasts
Commonwealth plans to drift back to surplus
show the triumph of experience over hope
Actual and forecast Commonwealth underlying cash balance
per cent of GDP
Forecast made in
1
2011 2012 2013 2014
2015
0
2016 2017
-1
-2
-3
Actual
-4
-5
2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
Financial year ended
Grattan analysis of Commonwealth Budget Papers
6
3
What can we forecast?
7
Which is easiest to forecast?
1
2
3
4
5
6
7
8
Google stock price tomorrow
Tourism demand next summer
Google stock price in 6 months time
Exchange rate of $US/AUS next week
Prison population in 2 years
Time of sunrise this day next year
Maximum temperature tomorrow
Daily electricity demand in 3 days time
8
Which is easiest to forecast?
1
2
3
4
5
6
7
8
Google stock price tomorrow
Tourism demand next summer
Google stock price in 6 months time
Exchange rate of $US/AUS next week
Prison population in 2 years
Time of sunrise this day next year
Maximum temperature tomorrow
Daily electricity demand in 3 days time
https://PollEv.com/georgeathana023
Which one is easiest to forecast?
8
Which is easiest to forecast?
1
2
3
4
5
6
7
8
Google stock price in 6 months time
Exchange rate of $US/AUS next week
Google stock price tomorrow
Prison population in 2 years
Tourism demand next summer
Daily electricity demand in 3 days time
Maximum temperature tomorrow
Time of sunrise this day next year
9
Which is easiest to forecast?
1
2
3
4
5
6
7
8
Google stock price in 6 months time
Exchange rate of $US/AUS next week
Google stock price tomorrow
Prison population in 2 years
Tourism demand next summer
Daily electricity demand in 3 days time
Maximum temperature tomorrow
Time of sunrise this day next year
how do we measure “easiest”?
what makes something easy/difficult to forecast?
9
Factors affecting forecastability
Something is easier to forecast if:
1
2
3
4
we have a good understanding of the factors that contribute to it;
there is lots of data available;
the future is somewhat similar to the past;
the forecasts cannot affect the thing we are trying to forecast.
10
Forecasting, goals and planning
Forecasting: predicting the future as accurately as possible,
given information available (historical data, future events that
might impact the forecasts).
Goals: are what you would like to have happen. Linked to
forecasts and plans, but this does not always occur.
Planning: is a response to forecasts and goals. Determining
appropriate actions to make your forecasts match your goals.
11
Outline
1 What can we forecast?
2 Time series data and random futures
3 Some case studies
12
Types of data and methods
Qualitative forecasts: Judgemental forecasting the only option if
no historical data, new product, new market conditions. See
fpp3 Chapter 6: https://otexts.com/fpp3/judgmental.html.
Quantitative forecasts: most forecasting problems use either
▶
▶
Time series data (collected at regular intervals over time).
Cross-sectional data are for a single point in time.
13
Time series data
Four-yearly Olympic winning times
Annual Google profits
Quarterly sales results for Amazon
Monthly rainfall
Weekly retail sales
Daily IBM stock prices
Hourly electricity demand
5-minute freeway traffic counts
Time-stamped stock transaction data
14
Random futures
A forecast is an estimate of the probabilities of possible futures.
15
Random futures
A forecast is an estimate of the probabilities of possible futures.
Thousands of visitors
Total short−term visitors to Australia
1000
500
0
Jan 2000
Jan 2005
Jan 2010
Month
Jan 2015
Jan 2020
15
Random futures
A forecast is an estimate of the probabilities of possible futures.
Thousands of visitors
Total short−term visitors to Australia
1000
Future 1
500
0
Jan 2000
Jan 2005
Jan 2010
Month
Jan 2015
Jan 2020
Simulated futures
16
from an ETS model
Random futures
A forecast is an estimate of the probabilities of possible futures.
Thousands of visitors
Total short−term visitors to Australia
1000
Future 1
Future 2
500
0
Jan 2000
Jan 2005
Jan 2010
Month
Jan 2015
Jan 2020
Simulated futures
17
from an ETS model
Random futures
A forecast is an estimate of the probabilities of possible futures.
Thousands of visitors
Total short−term visitors to Australia
1000
Future 1
Future 2
Future 3
500
0
Jan 2000
Jan 2005
Jan 2010
Month
Jan 2015
Jan 2020
Simulated futures
18
from an ETS model
Random futures
A forecast is an estimate of the probabilities of possible futures.
Total short−term visitors to Australia
Thousands of visitors
Future 1
Future 2
1000
Future 3
Future 4
Future 5
Future 6
500
Future 7
Future 8
Future 9
Future 10
0
Jan 2000
Jan 2005
Jan 2010
Month
Jan 2015
Jan 2020
Simulated futures
19
from an ETS model
Random futures
A forecast is an estimate of the probabilities of possible futures.
Thousands of visitors
Total short−term visitors to Australia
1000
500
0
Jan 2000
Jan 2005
Jan 2010
Month
Jan 2015
Jan 2020
Simulated futures
20
from an ETS model
Random futures
A forecast is an estimate of the probabilities of possible futures.
Thousands of visitors
Total short−term visitors to Australia
1000
500
0
Jan 2000
Jan 2005
Jan 2010
Month
Jan 2015
Jan 2020
Simulated futures
21
from an ETS model
Random futures
A forecast is an estimate of the probabilities of possible futures.
Thousands of visitors
Total short−term visitors to Australia
1000
500
0
Jan 2000
Jan 2005
Jan 2010
Month
Jan 2015
Jan 2020
Simulated futures
22
from an ETS model
Random futures
A forecast is an estimate of the probabilities of possible futures.
Thousands of visitors
Total short−term visitors to Australia
1000
500
0
Jan 2014
Jan 2016
Jan 2018
Month
Jan 2020
Simulated futures
23
from an ETS model
Random futures
A forecast is an estimate of the probabilities of possible futures.
Thousands of visitors
Total short−term visitors to Australia
1000
500
0
Jan 2014
Jan 2016
Jan 2018
Month
Jan 2020
Simulated futures
23
from an ETS model
Random futures
A forecast is an estimate of the probabilities of possible futures.
Thousands of visitors
Total short−term visitors to Australia
1000
500
0
Jan 2014
Jan 2016
Jan 2018
Month
Jan 2020
Simulated futures
24
from an ETS model
Random futures
A forecast is an estimate of the probabilities of possible futures.
Thousands of visitors
Total short−term visitors to Australia
1000
500
“He who sees the past as surprise-free is
0
bound to
have a future full of surprises.”
Jan 2014
Jan 2016
Jan 2018
Month
(Amos
Tversky)
Jan 2020
Simulated futures
24
from an ETS model
What is a forecast
A whole probability distribution, we call this a forecast distribution,
which we summarise with the mean, we call this a point forecast and
some other quantiles, we call these prediction intervals.
25
Outline
1 What can we forecast?
2 Time series data and random futures
3 Some case studies
26
CASE STUDY: Australian domesting tourism
27
Forecasts published by TFC
Visitor nights: Total
28
Forecasts published by TFC
Visitor nights: Total
Estimates of slope coeffients
Sample -3.16
Forecasts 188.07*
28
Forecasts published by TFC
Visitor nights: Total
Estimates of slope coeffients
Sample -3.16
Forecasts 188.07*
Always plot your data and
your forecasts.
28
Final long-term forecasts
29
Final long-term forecasts
Switch to evaluation slides.
29
CASE STUDY: Airline
30
CASE STUDY: Airline
Economy class passengers
Melbourne−Sydney
Thousands
30000
20000
10000
0
1987 W53
1990 W01
1992 W01
Year
31
CASE STUDY: Airline
Economy class passengers
Melbourne−Sydney
Thousands
30000
20000
10000
Not the real data! Or
is it?
0
1987 W53
1990 W01
1992 W01
Year
31
CASE STUDY: Airline
Economy class passengers
Melbourne−Sydney
Thousands
30000
20000
10000
0
1987 W53
1990 W01
1992 W01
Year
31
CASE STUDY: Airline
Problem: how to forecast passenger traffic on major routes?
Additional information
They can provide a large amount of data on previous routes.
Traffic is affected by school holidays, special events such as the
Grand Prix, advertising campaigns, competition behaviour, etc.
They have a highly capable team of people who are able to do
most of the computing.
32