Message Boards Message Boards

Instagram Viral Hashtag Analysis

Posted 11 years ago
Some hashtags on the social image-sharing network Instagram have gone somewhat viral. The two that I use here are "#selfiesunday," which involves taking a photo of oneself on a Sunday and "#throwbackthursday" or "#tbt," which involves posting a picture of oneself from their past on a Thursday. Both these hashtags involve days of the week, and variations can be made such as "#selfietuesday" and "#throwbackwednesday." I took statistics from the Instagram search and made bar graphs for each of these phenomena covering all days of the week and "#tbt"-type acronyms.

The "#selfiesunday" analysis was quite straightforward. The code used follows:
selfie = {5562997 + 41069 + 37999 + 19635 + 7752 + 3262,
  138536 + 3290, 101829, 67855, 72681, 104998, 759679 + 11878};
BarChart[selfie,
ChartLabels -> {"Sunday", "Monday", "Tuesday", "Wednesday",
   "Thursday", "Friday", "Saturday"}, ScalingFunctions -> "Log"]
The variable "selfie" contains post counts for different hashtags for days of the week. Multiple addends account for variations in the hashtag such as "#selfiesundays" and "#selfiesundayyy." I accounted for all of these variations down to 5000 posts. The graph had to be plotted on a log scale because the original hashtag was much more popular than the variations, and a linear plot would have diminished the variations to an unreadable level. The resulting graph follows:

The "#throwbackthursday" graph was more complex for two reasons. One, I had to include both full hashtag names, such as "#throwbackthursday" and sometimes more-used acronyms, like "#tbt." Two, some acronyms, such as "#tbt," can represent two days of the week, Tuesday and Thursday in this case. The other scenario is "#tbs" for Sunday and Saturday. Also, I cannot account for alternate meanings of the acronym. I solved the first problem simply with a multi-series graph. For the second I used a weighted RandomChoice. The RandomChoice was weighted with the post counts of each of the two possible days, and I Table'ed it over the post count of the acronym. I then Count'ed the results and used them as the post counts for the acronyms. The code and graph follow:
 throwback = {{158687 + 5432, 119419 + 5096, 342816 + 10122,
    82899, 34538975 + 1222803 + 252837 + 157518, 173771 + 6478,
    130427}, {169790, 144941, 1293641, 105580, 132526223, 241935,
    134165}};
 Count[
  Table[RandomChoice[{352938, 36172133} -> {tues, 
      thurs}], {133819864}], tues];
 Count[
  Table[RandomChoice[{164119, 130427} -> {sun, sat}], {303955}], sat];
BarChart[throwback,
ChartLabels -> {{"Full Name", "Acronym"}, {"SUN", "MON", "TUE",
    "WED", "THU", "FRI", "SAT"}}, ScalingFunctions -> "Log"]

More information is available upon request.
POSTED BY: Jesse Friedman
2 Replies
Hi, Jesse, this seems very interesting, thanks for sharing! But I have a few questions.
  • Where did you get the data?
  • Is this a single account data or all Instagram?
  • On Tuesday there was only #selfie tag?
  • Did you account for #selfie variations like #selfies, #selfiee, #selfietime, etc. ?
It would be great if you could tell us more details about this research!

Thanks!
Sam
POSTED BY: Sam Carrettie
Thanks for your interest, Sam! To answer your questions,
  1. The instagram app for iOS has limited hashtag data. If, say, you search for "selfie" in hashtag mode but don't execute the search, the app presents that hashtag with post count as well as other hashtags including that word. Therefore, post counts were inputted manually. Instagram has an API, however, so I might be able to make the program self-updating.
  2. The data is for all posts on Instagram from nonprivate accounts.
  3. I'm afraid I don't understand the question!
  4. I accounted for all variations that had more than 5000 posts. For example, #selfiesundayy has about 42000 posts right now, but #selfiesundayz has only 1800. Just to be clear, I'm not analyzing hastags like #selfie, only weekday-influenced variations, such as #selfiethursday and #selfiesaturday.
I hope this helps!
POSTED BY: Jesse Friedman
Reply to this discussion
Community posts can be styled and formatted using the Markdown syntax.
Reply Preview
Attachments
Remove
or Discard

Group Abstract Group Abstract