deviantART - Data Mining V - Global Activity

7 min read

Deviation Actions

parallellogic's avatar
Published:
14.2K Views
.


Status


There was a recent request for a dA data mining series on the languages spoken on dA.  I lack the skills/time to make a bot that can track the languages spoken on individual profiles.  The most pertinent results I can present come from data collection over the past several months from world (so roughly stated, where users log in from represents the languages spoken).  I took the number of comments plus the number of favs that came from each geographic location spread over each day and plotted them as a heat map.  The entire animation represents activity over 24 hours, each frame averages activity over 20 minutes.  For clarity I used a black and white world map for the background and the opacity of the red indicates the level of activity throughout the day.  The video starts and ends at midnight CDT.  The brightest point in the day globally occurs at roughly 4pm CDT while the dimmest (least activity) happens at 4am CDT.  This loosely corroborates the trend seen at the end of this journal parallellogic.deviantart.com/j…

I fudged the data a bit with the colors so you could see it (otherwise it's just a few single pixels that light up in the entire animation) – the data is scaled logarithmically, so the lowest points appear clear as you would expect, but intermediate values are pushed to be brighter.

Video:


Use caution when viewing this video.  The first time I tried to view it at full 1080p I crashed my system, and now every subsequent time I try it at full res my screen dies and then returns a second later with an error saying my video system crashed but recovered successfully.  When I try it on 720p or 360 it plays once but on each subsequent playback I get a single still frame rather than seeing the animation (so I have to refresh the page each time to see it again).  If you're really interested in seeing this, download it and then view it with VLC player or similar where you can loop the animation without worrying about crashing your system in the process :no:

I think the recent XKCD comic summarizes the results well in an anticlimactic manner (specifically the roll over text) xkcd.com/1138/

Briefly stated: dA's interface is strictly English, so you see users predominantly coming from English-speaking regions.  I've included a capture from 4pm CDT – you can compare it against a traditional map to find active regions

Single capture:


This represents the third attempt at processing this data set.  In the past I tried collecting the data in large arrays, but it turns out users don't log in from the middle of the Indian Ocean or any other large body of water, so it's a very empty matrix.  My computer can't handle multiple frames of huge arrays in that fashion so I gave up in the past.  I started fresh with this attempt and used a quadruple nested Hashmap (hour/minute + longitude/latitude + activity level) to hold the data and dump it into the frames for the final animation and that worked pretty well this time around.

Questions and Comments always encouraged :)

deviantART Series
Data Mining I - account type breakdown
Data Mining IIa - point transfers
Data Mining IIb - points followup
Data Mining IIIa - comments per day, number of banned/closed accounts
Data Mining IIIb - official verification of data mining numbers
Data Mining IV  - group watchers





Close Friends & Groups

Close Friends:

:iconwaterbender-1-8-9: :iconberndem-bones: :iconchiisai-kotoba: :iconblakeroats: :iconultimeciaffb: :iconmist-rainbow: :iconthejediknight1: :iconthatkidkale: :icondazza1008: :iconfantasystock: :iconmidnightsoiree:



Groups I have previously or am running:

:iconteentitans: :iconcartoon-obsessions: :iconvarietyclub: :icondevbug: :icondeviantartconcept: :iconfree-davatars: :iconsubscription-center: :iconbadges-of-deviantart:





colored and coded by parallellogic
© 2012 - 2024 parallellogic
Comments40
Join the community to add your comment. Already a deviant? Log In
Pickley's avatar
That's awesome, why not use ~world to get data for all the information? Would be awesome to see that visualised as a heatmap.