The line and bar charts are c3js indeed, which wraps d3. The mind map graph is done using cytoscape.js.
Data gathering, preparation, initial exploratory analysis and pretty much everything else is done with a combination of Jupyter notebooks, some Python scripting, and a set of Jinja2 templates for the HTML and the city specific JS. The JSON which is loaded to populate the charts is also generated from the same Python scripts. The code base is not necessarily something to be proud of, but it works. We focused on analysis (hence, the default bootstrap theme and nothing fancy).
We're likely to do a making of talk at one or two meetups here in Amsterdam. Hopefully those will be recorded. There's some additional interesting aspects to the collected data not immediately visible in this presentation.
Disclaimer: I did most of the data wrangling and coding on this; also I'm CTO for GoDataDriven, one of the two companies mentioned as producer.
Yeah, I have some ideas about including different data sources. You have to start somewhere, though, and Meetup.com has a nice real-life aspect to it. My best guess would be that Github and HN would add most interesting data from a technology / community popularity and natural language perspective respectively.
Question is: how do you connect a Github repo or a HN post to a specific community? Content based? User based? Topic based? The community detection in the real life Meetup.com network is at the very basis of this and for good reasons, so you'd have to find a way to tie other data to that idea.
Very nice! Despite the super bootstrapped layout, the data itself is really great. And as a result of your work, I'm now seriously considering Amsterdam, haha.
19 comments
[ 2.7 ms ] story [ 57.6 ms ] threadhttp://c3js.org/
Data gathering, preparation, initial exploratory analysis and pretty much everything else is done with a combination of Jupyter notebooks, some Python scripting, and a set of Jinja2 templates for the HTML and the city specific JS. The JSON which is loaded to populate the charts is also generated from the same Python scripts. The code base is not necessarily something to be proud of, but it works. We focused on analysis (hence, the default bootstrap theme and nothing fancy).
We're likely to do a making of talk at one or two meetups here in Amsterdam. Hopefully those will be recorded. There's some additional interesting aspects to the collected data not immediately visible in this presentation.
Disclaimer: I did most of the data wrangling and coding on this; also I'm CTO for GoDataDriven, one of the two companies mentioned as producer.
Question is: how do you connect a Github repo or a HN post to a specific community? Content based? User based? Topic based? The community detection in the real life Meetup.com network is at the very basis of this and for good reasons, so you'd have to find a way to tie other data to that idea.