Looking back to our first Leuven Smart City Hackathon

We've been organising Data Science Leuven for about 7 months now, and the response we often get from community members is that we are able to keep quite a steady pace, organising one meetup every month. Well, the secret behind that, is very simple. From the beginning, we decided organising a meetup should be low effort. We physically get together every few months, we have a Google Drive and a Slack channel for centralised communication and, most of all, we make the meetups very simple: 3 speakers, every speaker gets 20 min + 10min questions. It should be data-related. Absolutely no sales pitches. After the talks, we go for drinks at STUK Leuven, which I find as least as important as the talks themselves. Done. No theme nights, no complex procedures, no fancy setups, no corporate stakeholders to please. Nobody expects a perfectly organised event. But it also has a very pleasant side-effect is that people feel it's truly community driven, and we get a lot of participation from the audience, proposing new talks, inviting their friends to a next meetups and just overall making sure everybody is having a fun night.

This weekend we wanted to take that engagement with the community one step further by organising a Leuven Smart City hackathon. I'll be honest, I underestimated all the work that was involved in organising such a hackathon. And I only took care of the data aspects. Linde Vloeberghs had to move mountains to get everybody on the same page, and making sure everything was ready for this 2 day event. So, how did it go?

Friday Kickoff

We got together on Friday night so we got the chance to get to know each other and to introduce the challenges. There were 3 topics:

1. The data science track: can we build insights on the data that was made available? That can be from creating visualisations, building complex models, or just finding a unique insight that is useful for the citizens of Leuven.
2. The ideation track: how do we see Leuven in 2030, what should we work on next? How can we improve the health and lives of citizens, what about the circular economy, can we do something about polluting traffic?
3. The boring track: It is important to get the basics right. While Leuven already offers a cool dashboarding tool, https://leuven.incijfers.be/, there is no Open Data Platform yet. Can we kickstart this for them?

For the data science tracks, there were 3 main data sources to work from:

1. Leuven City Data: Every building, every tree, every bench, every playground and much more was all made available by the city of Leuven.
2. Airquality data: Low-cost sensors from leuvenair.be and a bit more fancier sensors from Option measure Particulate Matter in several places in Leuven
3. Parking data: Provided by Smart Flanders, we got a real-time and historical overview of the occupation of the major underground parking lots in Leuven: Ladeauze, De Bond, Center, ...

Saturday: Let's get to work!

With team names such as "We Kan", "Team Oisschot", "2 PHDs and a Hentenaar", "Team Ukkel", "Team Frederik" and "Team AAAAAARRRRRRRRR", everybody was all set to start hacking away. One thing we noticed from the beginning, is that it wasn't trivial to work with the datasets and everybody basically struggled the entire morning just to parse the data. Some recurring issues were "omg, what's this crazy datetime format", "hmm, working with a 1GB CSV file is not super fast on my laptop", "THAT JSON complexity" and "I don't really understand what these fields actually mean". All good reasons why we should have an Open Data Platform with plenty of metadata so people can get a quick understanding of the data that is available.

During the afternoon, energy was rising again because the first insights were coming back. Still, 1 day for a hackathon is very short. So there was quite some pressure within the different teams to work on all the ideas they had.

Yeah yeah, get to the results please

It's pretty impressive what people can deliver in just 8 hours, as long as they have an exciting challenge to work on and you get out of their way so they actually focus on the problem at hand.

First and foremost, team We Kan delivered an Open Data Platform using CKAN and Azure Cloud. It is hosted here. It goes without saying that it's still experimental but it shows that innovation doesn't have to be expensive or take 6 months to deliver something.

Next up, we had some pretty neat visualisations from the Leuven City data. Below you see the historical buildings of Leuven that are still existing today.

Of course, not every building was registered 100+ years ago, and there were definitely buildings on the outskirts of Leuven 100 years ago, but most of them probably didn't survive. The visualisation does give you an idea of the expansion of Leuven over time.

Istvan was able to plot all the trees managed by the city of Leuven. Every color is a different sort of tree.

When it comes to air pollution data, the most impressive analysis was definitely coming from Team AAAAARRRRR, who built non-linear models to understand the impact of humidity and temperature on the particulate matter levels.

But other teams had some very nice insights as well. Team Frederik looked for a correlation between air pollution and traffic data, but that was quite ambitious as not a lot of traffic data is available yet for Leuven, at least not in Open Data.

"2 PhDs and a Hentenaar" also had mixed results. One of their presentations wouldn't even load on the screen, which was "symbolic for their day". But they did find the optimal times for athletes to do their run in Leuven: between noon and 6pm, the air pollution is usually lowest.

In our team, we tried to hunt for the impact of wood stoves on air pollution, and while we did see some sensors which measured significantly more particulate matter than others, we couldn't pinpoint it to wood stoves.

Finally we also looked at the Wifi counters, which were in the Option dataset. There were only 4 locations in Leuven, but you could see that some locations saw a lot more Wifi Devices (= visitors) at night than during the day. Guess where in Leuven that was the case :-) We also found that Ladeuze parking was often fully booked, especially on Fridays and Saturdays. No surprises there, but it should help the City to steer traffic away from Ladeuze towards other parkings. Or even better, towards the edge of the city, and using public transport or bikes to get into the city center. Every graph on the right represents a day of the week, and every line is the number of free spaces in Ladeuze for a given day in the last 8 months or so. You see on day 4 and 5 (Friday and Saturday), the number of available spots reaches 0 quite often.

Finally, Linde hosted the ideation track together with Leuven 2030 and they came back with a million ideas. I'm expecting a blog from them to structure their thoughts and findings :-)

What's next?

It was great fun organising and doing the hackathon. There's definitely going to be another one. But we'll probably need some time to organise it all. Big thanks to our sponsors for helping us out and supporting us. And big thanks to Linde Vloeberghs for doing most of the work. We really hope we can get enough traction, also within the city of Leuven to do it again. So, feel free to share this with anybody you know that might be interested in our results. Get in touch through [email protected] or contact Linde or me directly.