You are viewing a single comment's thread from:
RE: eSteem - Full Year 2017 Statistics
Thank you for the contribution. It has been approved.
Hi @minitature-tiger, great work again!
Just out of curiosity: How did you filter the locations? Is it just a check if the exact string is in a dictionary of known countries, or did you go deeper like substring parsing, or detecting region names within a country etc?
You can contact us on Discord.
[utopian-moderator]
Thanks @crokkon!
It was very manual
(deep shame).
I put all the locations into a pivot table, sorted by count, and then went through the list from the top. I think the first one was Indonesia, so I searched all the other strings for Indonesia, and summed the count if there was a hit (so Berlin, Germany would be a hit for Germany for example).
And then I went down the list, making sure not to double count, and grouping cities into the right country etc (i.e. Berlin by itself in with Germany). Very manual!
I'm not a big fan of the location analyses, just because they rely on user input, and not enough people enter the details for it to be scientific. So I didn't want to spend too long developing something more complex. But for eSteem it's one of their aims to reach developing countries so I thought it was important to see whether that was the case. There's enough data for it to be interesting and indicative but not enough to prove any firm conclusions.
I'd be interested if @good-karma and the team have better data themselves, from it being a mobile app, and whether that correlates to the data above or if it's wide of the mark.
No, not "deep shame", rather the opposite. Kudos for digging through the data by hand! :)
Free text fields are hard to analyze because users can put anything there and I guess your approach still provides the most reliable results.
Adding a GPS location to the the
json_metadata
could actually make a nice (OPTIONAL) feature for reliable location reporting...WOW @miniature-tiger, as a newbie, I was curious to see Indonesia (Where I live), rate so high on the list and well above everyone else! At first I thought it was targeting posts to my location, a bit like the other (unnamed) social portals do with their ads, but it appears there a lot of Indonesian folk out there with time on their hands to contribute.