7

I've been playing around with Spark and React.js to learn more about how they work and to do something practical with them. So I decided to index the whole DMZ chat into an elasticsearch database, which I can query with pypark (I'll open source all the code).

What I want to do is create some type of dashboard application which shows some completely useless, but funny statistics about the DMZ.

What I need some help with is what type of statistics you guys would like to see:

  1. For instance how many times a day does Simon say donut?
  2. How many times does Avid say "your mum".
  3. Who's the biggest foul mouthed?
  4. Who posts the most images?

Any way if you guys have good ideas, feel free to share them!

2
  • You should probably make 1. How many days Simon does NOT say donut.
    – M'vy
    Aug 24, 2015 at 13:09
  • For any of my suggestions, consider the item being tracked more than the metric being requested. Example: "How many times does Simon say 'pls'" could be a daily rate, an average, or an all-time total. I really haven't figured out exactly which metrics are the most interesting - just which things would be interesting to see covered.
    – Iszi
    Aug 26, 2015 at 14:35

9 Answers 9

2

How many times not having the new profile is mentioned.

I'll edit this post as a come up with ideas.

1
  • That'll eventually hit a limit though. At least in theory. So it won't be so interesting later on.
    – Iszi
    Aug 26, 2015 at 14:27
2

The following assumes there is a reasonable number of hyperlinks that point back to any StackExchange site.

What can we do with a month's worth of StackExchange posts posted in the DMZ? I have no idea, but mining the metadata might give insight such as:

  • Popular sister sites (Links to crypto, serverfault, stackoverflow)
  • Upvotes / Downvotes of linked questions
  • Aggregate tag count of linked Q's
  • Activity / Views

We might discover that we use the chat room as an administrative way to maintain the site (no spam or off topic questions) or if we discuss hot & relevant questions or answers. Or if we're mostly posting jokes, and memes.

Maybe we are in the DMZ to talk about smart security questions during the school year and take off or fight spammers in the summers.

Top votes of linked questions

       Closed  Deleted  votes>5  votes>10   
May     3        1         9        11
June    10       1         1        0
July    12       0         0        0 
Aug     4        3         1        1 
Sept    1        3        19        3

Or we can think about a topic's popularity... besides "count", StackExchange links could highlight certain [tags], such as SSL, encryption, Cloud, or privacy. This would make old forum chats more relevant to the Q&A model of the SE network

TopTags of linked questions

       Tag1              Tag2               Tag3
May   SE/Programming    Bitcoin/Hash        SE/Privacy  
June  SE/Passwords      SE/Mobile           SE/VPN
July  SE/HTTP           SE/Wifi             SE/XSS
Aug   SE/PHP            SE/OpenSSL          SO/Encryption
Sept  SE/VPN            ServerFault/Windows  Crypto/BearStuff

Note that above I referenced tags in other sites.

The only trouble I can guess is that the calculation of top tags would have to be refreshed since the +1 votes change over time.

1

How many times is Simon's mom talked about?

1

How many times is an XKCD comic posted, that's not done by the bot?

1

How many times does Simon say "pls"?

1

Not a stat, but a feature request:

Implement "Shit SE Says", filtered to just the DMZ, somewhere in the application.

Stack Apps Post: https://stackapps.com/questions/3826/shit-se-says-stack-exchange-chat-stars-out-of-context

Application Homepage: http://xn--41a2e.com/shitsesays/

1

I really want to see a history of highly rated +1 comments, linked to source...

... also everything that the "bears" say, those are often interesting conversations (disclaimer, haven't been in the DMZ in years)

Had no idea the API did chat history now

1

Look at the Wolframalpha.com Facebook Analyzer for inspiration.

The post analysis, over time of day, month and other graphs are interesting.

Direct person-Person @ notification count might show peers

1

It would be interesting to see a graph of DMZ activity of a given user, vs Questions, Answers, and comments given over time.

Not sure if this is accessible, but vote-to-close, and review queue date time metadata would be interesting.

e.g. Joe is on StackOverflow or the DMZ in some form from 9am to 9pm. Rarely is he on at 3 am to 4am.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .