WEF16 Davos Twitter Performance Analysis

During the World Economic Forum in Davos 2016 (WEF16) we collected over 480 thousand Tweets with content relating to ‘wef’ or ‘davos’ including popular hashtags and phrases.

Here is a network ‘map’ visualising the traffic and conversation during the conference.

Network Map of Twitter Conversation during WEF16. High resolution version here

In these maps each dot (or node) represents a Twitter User and the lines joining them represent communication between Users by Re Tweets and Mentions. The colours indicate communities of Users that communicate frequently with each other and may therefore share common interests.

Here are two previous articles which explain some details about the methodology for the analysis.

An analysis of Twitter Influencers in the field of Data Science & Big Data
G20 Twitter Communities

The size of each node represents its ‘importance’ in the network using a measure called Page Rank. This post explains a little more about how importance is measured.

By looking at the map we can gain some insight into who important Users were and the way in which they communicated with each other to form communities or Tribes.

It is also possible to rank users by various measures to give an indication of the way in which they performed on Twitter during the conference.

We rank User for four measures: Page Rank, Connectors, Interesting and Overall as follows:

Top Page Rank — the basic Page Rank measure which calculates each Nodes importance based on the importance of it’s connected Nodes and their importance etc.

Top Connectors — Uses a measure called Betweeness to calculate how well connected a User is in the overall network.

Top Interesting — A measure of how important the User is compared to how important the User is in the wider context. For example of Obama or Leo DiCaprio Tweet they will gain a lot of importance in this network because they have millions of Followers. The Interestingness measure corrects for the bias that this creates.

Top Overall — A calculation based on a combination of these (and other) factors to give an overall rank.

Here are the tables.

Gender Equality at WEF16

Let’s examine the topic of gender equality and the people engaging with and promoting this issue at the conference.

Looking at the map and the tables we notice two indications of the importance of this issue.

Firstly, looking at the map and particularly an area at the bottom right, it is noticeable that there are some prominent women Users and that they appear to congregate in a discernible community of Users.

Secondly looking at the Top Overall Users table we can see that there is a large number of women in the top 20 given that only 17% of attendees are women.

So there are some indications that there is a collective voice from influential women making a significant contribution to the overall conversation on Twitter around WEF16.

Communities & Tribes

Analysing the connections using graph theory enables us to discover what the topics are being discussed, who the influential users are, how communities (or Tribes) form around these influential people and which other users are being influenced and/or contributing.

The visualisation of the network and the simple tables are useful but the analysis also provides much more detailed information.

As mentioned above the colours in the maps are representations of the communities of Users which are detected by machine learning algorithms. By identifying how Users group together we can identify Tribes. A Tribe is community with a leader who is the most ‘important’ person in that community. The colours in the maps provide an indication of the communities and Tribes identified but in reality there are many thousands and the colours and maps cannot show all the detail. We also need to examine the information in a more structure way.

Here is a screenshot of the dashboard which presents this information. This is part of a more complex application and it is not the purpose of this post to demonstrate the full capabilities of the application. Just notice the Topics highlighted in green in the centre of the screen.

These topics are identified by a machine learning process which examines the text in the Tweets and the communities of Users in which they appear.

If we just filter the view to contain Topics with the word “women” the view is simplified.

Highlighted on the left is the Tribes of Users each named after the most important person in the list. You can clearly see that there is a large percentage of women leaders.

Highlighted on the right is a list of the hashtags that are frequently used within those Tribes.

In summary, this dashboard is now showing all the Tribes of Users that are identified as discussing the Topic of ‘women’, together with the hashtags being used within these Tribes.

However, you can also see that other topics of discussion are taking place within these Tribes which enables you to see which topics are also important to the Users in the individual Tribes and the wider community.

We can illustrate this on the network map.

Now we have identified that a particular tribe is interested in a certain topic we can look at the Users within the Tribe. Let’s look at Emma Watson’s Tribe:

In this list we can easily identify some Users who are important influencers for the gender equality movement.

Looking at communities of Users (rather than conventional Twitter metrics like Followers, Re Tweets, Hashtags etc) enables a complex subject domain to be efficiently mined for information about the Topics of conversation. This provides a powerful way to find the most Influential people in a given domain, in this case WEF16.

Coming Up Next…

Future posts will examine some

Examining the most important Tweets in the Tribes

A detailed look at the Users who make up a Tribe