Friday, September 16, 2016

Trump vs Hillary on Twitter.

WIKIleaks dumps massive new DNC leak on Sept 13th, 2016 (see one video on Youtube). It is obvious that both campaigns are using public media to gain more support. Here, I have something interesting to reveal too.

What I did are,

1) Using two group of hashtags, which are assumed to support Hillary or Trump respectively (not 100% accurate, but should be with high accuracy rate),
  1. Support Hillary: #votehillary2016 #VoteHillary #NeverTrump #IAmWithHer #WeAreWithHer #NoTrump #TrumpLies #StopTrump #DumpTrump #TrumpUnfit
  2. Support Trump: #MakeAmericaGreatAgain #HillarysBigotry #CrookedHillary #Hillary4Prison #NeverHillary #TrumpTrain #VoteTrump #LockHerUp #WakeUpAmerica #TrumpsArmy
 2) Writing an app to listen to Twitter events that will be triggered when any above hashtag is posted on Twitter.

3) Using descriptive statistics methods to do some analysis. Possibly, we can do more advanced analysis on the data. I am looking for Opinion mining training data for Twitter or other social media. Please let me know if you happen to have it and will share it with me.

My app has been continuously running and collecting data for more than two weeks now. I analyzed the first week's data and put some result here. It looks like both Trump supporters and Hillary supporters have been working hard on Twitter and some automation post applications are used to post tweets on Twitter. It also looks like Trump side know more about how to use social media auto app than Hillary side.

So, I suspect that Hillary really get great support from high-tech firm CEOs? (see Here's Which American CEOs Are Supporting Hillary Clinton and Here's Why Your Facebook Timeline Is Suddenly Full of Hillary Clinton Supporters)

Now, let's see some data and pictures. 

Let's see in one week, each candidate has how many supporters on the twitter.

Obviously, Trump supporters work much harder than Hillary's supporters on Twitters. Data are collected from Sept. 5 (Monday) to Sept. 11 (Sunday). You can also see the peak time of tweets post are on Thursday and Saturday. I will continuously observe if the is a really behavior pattern. 

Let's see what time in a day has the most tweets screaming on the twitter.

It is interesting to see, during one day, if  people tweets at certain time of a day. For example, lunch time? dinner time? Below picture shows a clear pattern of user behavior. it looks like people like to post tweets at UTC time 5pm (1pm EST) and 24pm (8pm EST).
This pattern could be useful for both Hillary and Trump campaigns to effectively post on social medial like twitter. And, in the fact, I think they do it already. I will show evidence in the following analysis.  

How different language speaking people support Hillary or Trump?

We filter out those languages that has less than 200 post in a week and get below picture.

It is clear that Hillary has more support only from es (Spanish) and "unknown". On the other hand, Trump get more support from people who speaks British English (en-gb), French (fr), German (de) and all the rest. People, who is interested in finding out all languages listed, can check language code by click here.

What kind devices/application each side use mostly?

After investigating the language people use, we can have a look at what application/devices users use to post tweet. The interesting phenomena is that Twitter Web Client contributes the most tweets.

Since there are too many different application and devices are used to post on twitter, I just plot two different data as below. I sorted largest number of tweets for each application/device. Not surprisingly, Hillary side and Trump side have different top 20 devices that post tweets. So, I plot two different pictures, one is for Hillary side's top 20 and the other is for Trump side's top 20. You can compare listed applications and see what is shown in Trump's side not in Hillary side. It will be interesting to dig more with those application. 

For Hillary's top 20 applications,

For Trump's top 20 applications

Let's list out top used devices/app for Trump and Hillary respective. Then, you can align them and carefully compare both sides. Clearly, Trump side attracts more auto application to tweets. On the Hillary side, one stands out application is "Monkey Thank U". So, who are they? What do they post on twitter? You can explore them with data I share at the bottom of this post.

Hillary Trump
Application Num
1 Twitter Web Client 181518
2 Twitter for iPhone 161013
3 Twitter for Android 151083
4 Twitter for iPad 62717
5 Monkey Thank u 10782
6 Mobile Web (M5) 10512
7 TweetDeck 9591
8 RoundTeam 5058
9 Annie Green 1.0 4708
10 diana aquaviva 1.0 4161
11 4040
12 Hootsuite 2839
13 Twitter for Windows Phone 2717
14 2679
15 prohiggins1.0 2678
16 Twitter for Windows 2476
17 DropOuthillary1 2252
18 DeviationStand 1.0 2239
19 Daniel Addison 1.0 2232
20 zerosum 1.0 2206
Application Num
1 Twitter Web Client 995545
2 Twitter for iPhone 922009
3 Twitter for Android 830775
4 Twitter for iPad 282343
5 Mobile Web (M5) 72408
6 61257
7 IFTTT 32233
8 Tweet Jukebox 26551
9 Twitter for Windows 16821
10 TweetDeck 13780
11 RoundTeam 11398
12 PinkPoniesGreat 10120
13 Mobile Web (M2) 9350
14 Hootsuite 8373
15 StopMadness 7841
16 Twitter for Windows Phone 7800
17 thebestappsever 7644
18 oneoftheapps 7437
19 rollingtwitter 6585
20 WhytePantherTest1 6337

Let's explore a little bit what both sides' robot post on the twitter.

No comments:

Post a Comment