AY1516 T2 Team AP Methodology
| Project Description | Data | Methodology |
|---|
Contents
Overview
In the table below we outline the algorithms/techniques that we intend to execute for a particular objective.
| Objective | Analytical Approach |
|---|---|
| Network analysis via Degree centrality, Betweenness centrality |
|
| Plan what to publish based on characterisation of audience |
|
Multivariable Regression on tweet content vs Google Trends
With reference to trending topics on a particular day of a tweet, multivariate regression will be performed to relate trending topics to the popularity of a tweet (retweet, likes, etc).
The key variables that we intend to explore are elaborated in the table below:
| Variable | Importance |
|---|---|
|
Retweets |
This measure shows how many times a particular tweet is being shared by followers. We think this is interesting because it highlights the willingness of an individual to share the tweet, increasing the probability that the tweet was interesting. |
|
Url clicks |
This measure shows how many times users actually click on the shortened link shared within a tweet. Given the succinct nature of a tweet, users who click on outgoing links are likely to find the tweet more interesting than other tweets, since clicking on the link would mean interrupting the "flowing" nature while reading the Twitter feed. |
|
Likes |
Compared to Url clicks and Retweets, this measure is the mildest, indicating that the user probably found the tweet interesting, but wasn't compelling enough to share. |
|
Engagement Rate |
A consolidated figure to illustrate how many people who see a particular tweet eventually interact with it (out of the total number of people who saw the tweet), in the following ways/forms:
|
|
Tweet Text |
Although the effectiveness of jokes can be tough to evaluate from a linguistics perspective, our initial approach would be cross referencing the hashtags used in the tweet with Google Trends data (Searches & Events) |
Giving a perspective on the important key variables that affects the popularity of a tweet will aid in the formulation of content that have higher penchant of being a popular tweet
Google Trends Correlation
While planning what content should be created, the team content team usually base it on gut feel, and usually the popular ones are accompanied when it is with regards to a big event in Singapore (eg. GE 2015). We want to incorporate the use of Google Trends to provide a more data-driven approach for SGAG in determining when and what content to publish. By analysing content consumption patterns of previous content around various types of events, future predictions can be forecast and appropriate content can be delivered at the appropriate timings.
