2015 was filled with extreme speculations in the US Presidential race. With so many candidates from different backgrounds and varying credentials fighting for the final candidature from their respective parties, this interesting development caught our attention.
It comes as no surprise that sites like Facebook and Twitter became the “War Zones” for the campaigners. Heated opinion exchange and incessant promotions by the supporters became the watermark for this online race. We observed this development during November 2015 on Twitter and tried to establish the course which the elections were taking. Let us begin with examining some visuals from the study.
 |
| Tweetline for the Democrats |
The plot above is what we call a ‘tweetline’ for the Democratic party. It simply displays what proportion of the tweets concerned the party, as opposed to the Republican party - more like ball possession in football.
Here we can clearly see that the Republican party is progressively getting a major share of the tweets. Subsequent investigations showed that the minor peaks were due to statements made by the Democratic party leaders on contentious issues leading to a series of discussions. One immediate observation is that such spikes are very short lived, so it doesn’t really matter in the long run if a party gets a momentary gain in popularity.
 |
| Word Cloud for Donald Trump |
The next visual is the Word Cloud generated for Donald Trump using Machine Learning on the tweets. The algorithm filters out the most relevant words and arranges them as ‘clusters’ representing emotions. The font size is proportional to the strength of the emotions as perceived by the algorithm. Sadness, disgust, and surprise were the major emotions which surfaced up for Donald Trump. But nonetheless, he was the one being talked about the most during the phase. Plots for other candidates like Hillary Clinton were more consistent with their tweetlines. A curious observer could identify the word ‘Cruz’ in the disgust cluster - Ted Cruz happens to be one of his competitors from the same party who doesn’t favour higher taxes on the wealthy, unlike the millionaire Trump.
 |
| Regional Support for the parties |
Let us move on to region-level analysis. The redder the zone, the more the Republican sentiment, and similarly for the Democrats and the yellow color. The map clearly shows the lead for the Republicans in most states. The blank states are the ones where the results weren’t much assertive. The possible reasons could include lack of location data, not clearly differentiable sentiments and lack of meaningful context. The state of Texas currently features Republican majority, as portrayed by the map plot. Arizona seems divided in the opinions.
 |
| MDS of sentiments |
The next plot is the Multidimensional Scaling of the two parties’ sentiments. MDS is a means of visualizing the level of similarity of individual cases of a dataset. The red color represents the Republicans. The plot clearly depicts the rift in the opinions of the two support groups.
A close observation reveals that the Democrat supporters fluctuate more in their stands and have more differences within their community. For the Republicans, despite differences, most major opinions are endorsed by a significant number of people. This could also be identified by noticing more blue colored outliers than the red ones.
A close observation reveals that the Democrat supporters fluctuate more in their stands and have more differences within their community. For the Republicans, despite differences, most major opinions are endorsed by a significant number of people. This could also be identified by noticing more blue colored outliers than the red ones.
The detailed analysis and more information about the co-authors can be found here.
As the clock ticks away, candidates from every party keep getting less in number, and Trump clearly has huge online support. Amongst the many standing for the post, who shall be the last man standing? The dilemma still remains - To Trump or Not To Trump!
