DCM 2019

Discrete choice modelling course 

May 13-15, 2019

Source: Michael van Eggermond

What is choice modelling?

From deciding to take a taxi or public transport, buying this or that item, choosing a destination of our next holidays or of our next visit of a public park, to what we have for lunch, we all make a series of choices every day. To understand choice making is therefore highly relevant in many fields of research and practice.

Choice modelling is an applied, quantitative method to study how choices are made. Insights form such econometric models can be used to derive elasticities with respect to price and other qualitative attributes, but also to quantify market shares for future products and services, such as high speed rail or telecommunication products.

Scope  

The course will provide an introduction to the key concepts of choice modelling, from basic to complex models. In hands-on exercises, the participant will learn how to use the state-of-the-art Apollo software which is a free package using the open source statistical software R to develop and estimate choice models and how the results are interpreted introducing the concepts of willingness to pay and elasticities.

Furthermore, the course also covers how choice modelling techniques are applied for forecasting. On the third day, the course will also introduce the participant into the art and science of machine learning, in the context of discrete choice modelling (DCM). We will explore various classification models as a complementary mode of analysis to DCM. Machine learning approaches may be used to uncover non-linearities in responses, as well as identify important variables and interactions, or to identify clusters and other structural features present in data.

The course will be relevant to individuals from all fields in which consumer choices are relevant, such as for example: marketing, economics, transport, environmental science, public health, planning and logistics. This variety of applications will also be reflected in the practical examples taught in the course.
 

Audience

This introductory course is intended for practitioners, academics, and managers in government and industry alike. It assumes no prior experience with discrete choice modelling but also those with some exposure will find it useful. Prior knowledge of basic statistical concepts and R will help understanding the concepts taught, but advanced training is not necessary.  

Dates & Program

The course will be held from May 13 to May 15, and starts with a two-day primer in discrete choice theory, followed by one day on machine learning.

Day 1

Morning

  • Introduction to choice modelling & data requirements
  • The Multinomial Logit model and estimation
  • Practical: Data handling and estimation of Multinomial Logit models in Apollo

Afternoon

  • Analysis and interpretation of results (e.g. WTP)
  • Practical: Further estimation examples in Apollo
     

Day 2

Morning

  • Nested Logit and other GEV models
  • Practical: GEV model estimation in Apollo

Afternoon

  • Random coefficients models
  • Practical: random coefficients models in Apollo
     

Day 3

Morning

  • Machine learning vs. discrete choice modelling
  • Introducing the caret package in R
  • Popular classification models

Afternoon

  • Data conditioning for classification
  • Avoiding over-fitting
  • Interpreting results from machine learning versus those from DCM

 

Venue

The course will be held at the Future Cities Laboratory, located in University Town, the new campus of the National University of Singapore.  

Future Cities Laboratory
Future Cities Laboratory

Presenters

Dr Stephane Hess

Dr Stephane Hess

The course will be taught by Dr Stephane Hess, Professor of Choice Modelling and Director of the Choice Modelling Center at the University of Leeds, Honorary Professor of Choice Modelling at the University of Sydney, and Honorary Professor of modelling behaviour in Africa at the University of Cape Town.

Professor Hess is an experienced educator having lectured choice modelling at different universities and conducted several similar courses to this one around the world in recent years. His area of work is the analysis of human decision using advanced discrete choice models, and he is active in the fields of transport, health and environmental economics and has published his research widely. He is also the founding editor in chief of the Journal of Choice Modelling and the founder and steering committee chair of the International Choice Modelling Conference

 

Dr Pieter Fourie

Dr Pieter Fourie, project leader of the Engaging Mobility project at FCL, is experienced in applying so-called machine learning methods in transportation analysis and will introduce these methods as a complementary mode of analysis to DCM.

Dr Michael van Eggermond

Dr Michael van Eggermond, a Senior Researcher at FCL and experienced choice modeller. He will guide the students in the various practical sessions.

Fees

The fees are 1,500 SGD for a 3 day discrete choice modeling and machine learning course. Discounts are available for participants from research institutes, educational institutes government agencies. The following discount scheme applies for academics and agencies:

Agencies 10% (1,350 SGD)

Academics 20% (1,200 SGD)
 

Registration

To register, please fill our registration form available at external page https://bit.ly/2VAyXML by April 27, 2019.

More information

For further information or special request such preferred areas of application being covered in lectures and practical examples, please contact Michael van Eggermond at eggermond@ivt.baug.ethz.ch or +65 9048 6808.

This flyer is available as a pdf external page here.