Forecasting: principles

and practice

Rob J Hyndman

Introduction to Forecasting
Forecasting: principles and practice 1
Outline

1 Background

2 Case studies

Forecasting: principles and practice Background 2

Resources

Slides
Exercises
Textbook
Useful links

robjhyndman.com/uwa2017

Forecasting: principles and practice Background 3

Brief bio
Professor of Statistics, Monash University
Editor-in-Chief, International Journal of Forecasting

How my forecasting methodology is used:

Pharmaceutical Benefits Scheme
Cancer incidence and mortality
Electricity demand
Ageing population
Fertilizer sales

robjhyndman.com
Forecasting: principles and practice Background 4
Brief bio
Professor of Statistics, Monash University
Editor-in-Chief, International Journal of Forecasting

How my forecasting methodology is used:

Pharmaceutical Benefits Scheme
Cancer incidence and mortality
Electricity demand
Ageing population
Fertilizer sales

robjhyndman.com
Forecasting: principles and practice Background 4
Assumptions

This is not an introduction to R. I assume you are

broadly comfortable with R code and the RStudio
environment.
This is not a statistics course. I assume you are
familiar with concepts such as the mean, standard
deviation, quantiles, regression, normal distribution,
likelihood, etc.
This is not a theory course. I am not going to derive
anything. I will teach you forecasting tools, when to
use them and how to use them most effectively.

Forecasting: principles and practice Background 5

Key reference

Hyndman, R. J. & Athanasopoulos, G. (2017)

Forecasting: principles and practice, 2nd ed.

OTexts.org/fpp2/

Free and online

Data sets in associated R package
R code for examples

Forecasting: principles and practice Background 6

Key reference

Hyndman, R. J. & Athanasopoulos, G. (2017)

Forecasting: principles and practice, 2nd ed.

OTexts.org/fpp2/

Free and online

Data sets in associated R package
R code for examples

Forecasting: principles and practice Background 6

Key reference

Hyndman, R. J. & Athanasopoulos, G. (2017)

Forecasting: principles and practice, 2nd ed.

OTexts.org/fpp2/

Free and online

Data sets in associated R package
R code for examples

Forecasting: principles and practice Background 6

forecasting

1 Guru: I wrote the book, done it for decades, now I do

the conference circuit.
2 Expert: It has been my full time job for more than a
decade.
3 Skilled: I have been doing it for years.
4 Comfortable: I understand it and have done it.
5 Learner: I am still learning.
6 Beginner: I have heard of it and would like to learn
more.
7 Unknown: What is forecasting? Is that what the
weather people do?

Forecasting: principles and practice Background 7

Poll: How proficient are you in using R?

1 Guru: The R core team come to me for advice.

2 Expert: I have written several packages on CRAN.
3 Skilled: I use it regularly and it is an important part of
my job.
4 Comfortable: I use it often and am comfortable with
the tool.
5 User: I use it sometimes, but I am often searching
around for the right function.
6 Learner: I have used it a few times.
7 Beginner: I’ve managed to download and install it.
8 Unknown: Why are you speaking like a pirate?

Forecasting: principles and practice Background 8

Install required packages

install.packages("fpp2", dependencies=TRUE)

Forecasting: principles and practice Background 9

Approximate outline
Day Topic Chapter
1.1 Time series graphics 2
1.2 The forecaster’s toolbox 3
1.3 Seasonality and trends 6
1.4 Exponential smoothing 7
2.1 ETS state space models 7
2.2 Transformations 3
2.3 Stationarity and differencing 8
2.4 Non-seasonal ARIMA models 8
2.5 Seasonal ARIMA models 8
3.1 Time series cross-validation 3
3.2 Dynamic regression 9
3.3 Hierarchical forecasting 10
3.4 Advanced models 11
Forecasting: principles and practice Background 10
Outline

1 Background

2 Case studies

Forecasting: principles and practice Case studies 11

CASE STUDY 1: Paperware company
Problem: Want forecasts of each of
hundreds of items. Series can be stationary,
trended or seasonal. They currently have a
large forecasting program written in-house
but it doesn’t seem to produce sensible
forecasts. They want me to tell them what is
wrong and fix it.

Additional information
Program written in COBOL making numerical calculations limited.
It is not possible to do any optimisation.
Their programmer has little experience in numerical computing.
They employ no statisticians and want the program to produce
forecasts automatically.
Forecasting: principles and practice Case studies 12
CASE STUDY 1: Paperware company
Methods currently used

A 12 month average
C 6 month average
E straight line regression over last 12 months
G straight line regression over last 6 months
H average slope between last year’s and this
year’s values.
(Equivalent to differencing at lag 12 and taking
mean.)
I Same as H except over 6 months.
K I couldn’t understand the explanation.
Forecasting: principles and practice Case studies 13
CASE STUDY 2: PBS

Forecasting: principles and practice Case studies 14

CASE STUDY 2: PBS

The Pharmaceutical Benefits Scheme (PBS) is the

Australian government drugs subsidy scheme.

Many drugs bought from pharmacies are subsidised

to allow more equitable access to modern drugs.
The cost to government is determined by the number
and types of drugs purchased. Currently nearly 1% of
GDP.
The total cost is budgeted based on forecasts of drug
usage.

Forecasting: principles and practice Case studies 15

CASE STUDY 2: PBS

The Pharmaceutical Benefits Scheme (PBS) is the

Australian government drugs subsidy scheme.

Many drugs bought from pharmacies are subsidised

to allow more equitable access to modern drugs.
The cost to government is determined by the number
and types of drugs purchased. Currently nearly 1% of
GDP.
The total cost is budgeted based on forecasts of drug
usage.

Forecasting: principles and practice Case studies 15

CASE STUDY 2: PBS

Forecasting: principles and practice Case studies 16

CASE STUDY 2: PBS
In 2001: $4.5 billion budget, under-forecasted by
$800~million.
Thousands of products. Seasonal demand.
Subject to covert marketing, volatile products,
uncontrollable expenditure.
Although monthly data available for 10 years, data
are aggregated to annual values, and only the first
three years are used in estimating the forecasts.
All forecasts being done with the FORECAST function
in MS-Excel!
Problem: How to do the forecasting better?

Forecasting: principles and practice Case studies 17

CASE STUDY 2: PBS
In 2001: $4.5 billion budget, under-forecasted by
$800~million.
Thousands of products. Seasonal demand.
Subject to covert marketing, volatile products,
uncontrollable expenditure.
Although monthly data available for 10 years, data
are aggregated to annual values, and only the first
three years are used in estimating the forecasts.
All forecasts being done with the FORECAST function
in MS-Excel!
Problem: How to do the forecasting better?

Forecasting: principles and practice Case studies 17

CASE STUDY 3: Airline

Forecasting: principles and practice Case studies 18

CASE STUDY 3: Airline
Economy class passengers: Melbourne−Sydney

30

20
Thousands

10

1988 1989 1990 1991 1992 1993

Year

Forecasting: principles and practice Case studies 19

CASE STUDY 3: Airline
Economy class passengers: Melbourne−Sydney

30

20
Thousands

10

Not the real data! Or

is it?
0

1988 1989 1990 1991 1992 1993

Year

Forecasting: principles and practice Case studies 20

CASE STUDY 3: 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.

Forecasting: principles and practice Case studies 21