Learning data science by doing it

University of Leeds students have delivered innovative new transport modelling ideas to an external client, Derbyshire County Council, in place of traditional written examinations for a taught module.

As a result of this work, the Council is now entering an on-going relationship with the module and bidding for new research funds with Leeds. This success shows how taught modules can be used to deliver pilot studies to prime research bids.

Students from the new Mathematical Transport Modelling MSc in the Institute for Transport Studies at Leeds studied ‘Data Science for Transport’ with module manager Dr Charles Fox and course leader Dr Richard Connors.

The new module was designed from conception in collaboration with industry stakeholders, keen to get their potential future employees trained in relevant skills.

Derbyshire County Council’s Transport Data and Analytics team were especially interested in this idea and worked very closely with the University of Leeds to develop the syllabus, including opening up their historical databases of vehicle monitoring all over Derbyshire for use by our students.

Instead of abstract exercises each week and a written exam at the end, the entire module was built around a real project for the Council, with each week’s work adding software chosen to practically illustrate the rigorous theory presented in class but also forming part of a large-scale data analysis tool.

University Academic Fellow Charles Fox said of the project:

“While based on some quite profound philosophical and mathematical ideas, Data Science is at heart an applied subject and real-world Data Scientists spend the bulk of their time making things work for real, including task like programming, data cleaning, and fixing dates and times as well as the more glamour versions of the subject found in most books such as ‘deep learning’ and ‘big data.’

A common complaint from industry is that many students only learn the theory, so for this module, we chose to show the whole process from client consulting to fixing database glitches. We did this by throwing the students in at the deep end, and taking the risk of trusting them with Leeds’ reputation and brand to work from the start on a real project with a real client.

It’s only been possible to do this thanks to Leeds’ University Academic Fellow excellent training programme and its sessions on innovative teaching which gave me the confidence to try something a bit different.”

Students presenting to Derbyshire Council.

Students presenting to Derbyshire Council.

The students worked closely with the Council including by phone meetings, email discussions, and visiting their site.

A final presentation was made by the students to the head of the Council’s team, who is responsible for all the transport analysis in the county.

The students showed the council how to make new use of their old data from Bluetooth and flow sensors around their network, to make new inferences about the specific routes being taken between origins and destinations, and to infer likely flows on parts of the road network which don’t have any sensors on them. This is important to a rural county like Derbyshire which has many small roads on which it is not cost-efficient to install expensive sensors.

An example flow map showing inferred traffic levels is shown here, for an area around Chesterfield:

Inferred traffic levels in Chesterfield.