This year’s EARL conference in London was another huge success. In this blogpost we revisit some of the presentations from Day 1.

In stream 1 the conference started off with a presentation about recommender systems at the Telegraph by Magda Piatkowska. When you think recommender systems Netflix and Amazon immediately come to mind but the problem is quite different for digital news media. Magda explained this very clearly and also showed how the Telegraph tackled the problem. ”Start dumb” is perhaps the best advice we’ve heard at a conference. The final solution was anything but dumb though and we will definitely be checking out context trees in the coming months.

Shiny has been one of the big developments in R in recent years and even at EARL we were impressed how much companies love Shiny. To start with, six talks, out of about fifty, contained ‘Shiny’ in the title, so clearly those focused on Shiny applications or other Shiny related development. But even many of the other talks featured Shiny.
To kick off the Shiny talks in stream 2, Nicky Van Thuyne from Bayer CropScience showed a not so shiny story about how they decided to hold off on Shiny development for a while. Luckily, the features they miss from Shiny are coming soon: the ability to write tests, and coding standards and practices for creating large Shiny apps.

After lunch Emma Liden and her colleague from Aimia showed how they managed to cover the holes in Shiny by adding some custom HTML and JavaScript to visualize timelines dynamically. They were quickly followed by Ciro Montagno from WorldPay, who showed a Shiny app for helping their call centre agents. It was great to see that not only have they managed to increase their performance numbers thanks to the app, the agents that are working with it, love it too.

At the same time in stream 1 it was Alice Daish’s turn to impress. Colleagues who have seen her present before recommended this talk and I was not disappointed. Alice showed how R can benefit established institutions such as museums by again starting simple. No big data, no massive computations in the cloud or complex algorithms. Just a few word clouds and a very convincing data scientist. This is also the first time we have ever seen a sneak peak of a data analysis. Although the analysis of Wi-Fi data wasn’t finished it didn’t stop the audience from asking questions about it and making helpful suggestions for improvement. A sign that this is indeed an interesting topic and one we should keep in mind the next time we visit a museum.

And let’s not forget stream 3 where the variety of talks were perfect showcases for the very broad capabilities of R. Hayfa Mohdzaini from UCEA kicked off by showing how R can be used to automate benchmarking reports, which can save a lot of time. R is also a great tool for data visualisation as was shown in the talk given about heatmaps by Tal Galili of R-Bloggers. While we have made heatmaps in the past using R and were satisfied with the results, during this talk we learned about the vast amount of options there are to improve them (e.g. interactive heatmaps, what to consider when choosing a colour scale or the distance metrics used). And last but not least Michael Spencer of the University of Edinburgh presented his results on modelling the size of catchments taking into account snow. The analysis done concluded that in most cases the regulations actually exceed the needed size for most of the catchments constructed.

Unfortunately, thanks to the parallel streams, one cannot see all the talks but luckily at least some of them were videotaped and we are looking forward to watching the ones we missed. All presentations can be viewed here. We hope you enjoyed this year’s conference as much as we did and hope to see you again next year.