A social network analysis of the 400k tweets about Lima COP 20

With more than 410 000 tweets in 15 days, the Conference of the Parties of Lima, or « COP 20 », was a success on twitter. Last year, when I analysed the discussions on twitter about the « COP 19 » in Bonn, I found only 15 000 tweets. The twitter analysis software Visibrain shows that there were 104 000 twitter accounts responsible for these 410 000 tweets, wich means that each account sent around 4 tweets – 70 % of them being retweets. The co-mentions file extracted from Visibrain, which will allow us to map the twitter accounts network depending on their respective mentions, corresponds to 363 000 tweets and 100 000 twitter accounts – almost all of them !

01_Total_COP20

The accounts with the most mentions, and the tweets with the most RT, correspond to the organizers of the COP 20, respectively @UN_ClimateTalks for the United Nations and @LimaCop20 for Peru, but also to celebrities defending the environment such as @LeoDiCaprio, and NGO like @WWF. The following social network analysis, made using the free software Gephi, shows only the 30 accounts with the most important number of mentions. The bigger the node is, the more mentions a twitter account got from / gave to other accounts (for example, 21 780 mentions for @LimaCop20, 86365 for @Greenpeace).

02_COP20_top_30_mentions

The twitter accounts who receive the most mentions and the most active mentioning accounts

An analysis and filtering of the 50 twitter accounts that received the more important number of mentions enable us to add to the list of influential accounts some politicians. For example, we can find @JohnKerry and @ElizabethMay, but also @LaurentFabius and @RoyalSegolene for France – the country that will organize the next Conference of the Parties in 2015.

We can also see that @Greenpeace benefits to the influence of other of its representatives, like @kuminaidoo, international director of the NGO, and very active on twitter.

03_COP20_weightin_TOP50

The 50 most active accounts (twitter accounts that mentions the most the other accounts) allow us to see more peruvian accounts such as @UeenPeru, @LarutaDelClima … Even if there are not a lot of mentions of these twitter accounts by other twitter users, they chose to compensate this lack of visibility by being extremely active and by mentioning other twitter accounts – mainly each other’s.

04_COP20_weightout_TOP50

The analysis of the 104 000 twitter accounts that sent tweets about the COP 20, using the Louvain algorithm, demonstrates that there was a large diversity of groups active in this discussion. However, the first 8th groups correspond to 70 % of all the twitter accounts.

We will see for each of these 8 communities what are the 30 twitter accounts with the most important number of mentions. The bigger the node is, the more mentions a twitter account got from the others. The links corresponds to the mentions between the different accounts.

First community: the United Nations network

The most important community of the COP 20 network corresponds to the twitter account mentioning and mentioned by the official accounts of the United Nations active in the Conference of the Parties: @UN_Climatetalks, @UN, @Unep …

This network seems to be particularly unbalanced, as the first most mentioned account got a large part of the twitter mentions.

05_COP20_1er_reseau_TOP30

Second community: the local organizers of the Conference of the Parties

This results seems largely due to the fact that 25 % of the 410 000 tweets are in Spanish, thus most probably emitted by Peruvians. The fact that 40 000 tweets are geo-localized in Peru, as much as for the USA, confirms this hypothesis.

This community is more unbalanced than the first one: @LimaCop20, @ManuelPulgarVidal (minister of the environment of Peru and president of the COP 20) and @Cfigueres (Executive Secretary of the UN Framework Convention on Climate Change) got most of the mentions. We can also see a lot of links between the twitter accounts, which signifies that the accounts with few mentions mentioned a lot the more influent ones – in order to give better visibility to tweets from Peru.

06_COP20_2eme_reseau_TOP30

Third community: the NGOs rally behind Greenpeace

The NGO’s network seems more balanced than the first two, and mixes environmental celebrity activists such as @LeoDiCaprio with NGOs : @Oxfam, @Greenpeace, @Ecowatch …

We can see that @Greenpeace network is extremely active on twitter, with @Greenpeace100re, @GreenpeaceUSA, @Gpph (Greenpeace Philippines), @Kuminaidoo, @Greenpeacearg, @Greenpeacecl …

There are a few exchanges between these twitter accounts, suggesting that the most important part of their mentions come from other communities.

07_COP20_3eme_reseau_TOP30

Fourth community: Policy advocates & specialized media

This network seems to correspond to policy advocates and specialized media from various backgrounds sharing one thing in common: they exchange a lot with each others – and they like @JohnKerry, as about 50 % of them are American. I suspect that most of them are recurrent participants to the Conference of the Parties.

In my social network analysis of the 2013 COP 19 in Bonn, the twitter accounts of @Yebsano, @Duycks and Rttcclimatenews where already detected by Gephi as members of the same community : https://cartorezo.files.wordpress.com/2014/04/04_adp2014_influenceurs.pdf

Also, @350 and its communication director @agent350 have a particuliar interest for climatic refugees.

08_COP20_4eme_reseau_TOP30

Fifth community: Specialized media and news network on climate change

09_COP20_5eme_reseau_TOP30

 @EcoInternet, @ElClimate, @El_EcoNewsfeed, @wbclimatechange are information networks where people discuss about climate change and environmental issues. We can see that people discussing on #COP20 seem to prefer information from this sources rather than from @reuters or from @bloombergnews.

It is interesting to note that @MAC_Europa, the twitter account of Miguel Arias Cañete, EU Climate Action and Energy Commissioner, is a part of this community.

Sixth community: Political Activists

The network seems to correspond to political activist demanding more democracy in environmental policies.

10_COP20_6eme_reseau_TOP30

Seventh community: the wildlife lovers

The @WWF clearly leads this network, with the help of @wwfeu, @wwfnews, @wwf_uk, @climatewwf …

11_COP20_7eme_reseau_TOP30

The last and not the least influent network: the frenchies!

France being the next country to organize a Conference of the Parties, it is only logical that French is the third most used language in the COP 20 tweets (corresponding to 6 % of it).

For the moment, it’s mostly NGO’s (@RACFrance, @Attac_fr, @Oxfam_fr) and Green MP (@Denis_Baupin, @M_orphelin …) who are involved in the discussion – with the French Environment Minister, @RoyalSegolene, and the French Foreign Affairs Minister, @LaurentFabius.

12_COP20_8eme_reseau_TOP30

Other kinds of influence: centrality in the network and connection to influent people

Gephi allows us to make other calculation of influence than the number of mentions for and from each twitter accounts on the map. Two of them are particularly interesting.

Betweeness Centrality measures the connectivity of a node (a twitter account) to all the other twitter accounts of the network. In our case, an important betweeness centrality value signifies that a twitter account is mentioned and mentions other accounts from various communities, enabling it to reach a very heterogeneous audience.

The 50 accounts with the most important betweeness centrality confirm that @LimaCop20, @Un_Climatetalks and @Greenpeace are indeed weel connected not only in their respective communities, but to all the other networks within the exchanges about the COP 20. The separate network of COP environmental activists is also well connected, through @yebsano and @duycks.

13_COP20_betweeness_TOP50

Centrality Eigenvector measures the connectivity to other well-connected node. In our case, it means that a high value corresponds to a twitter account exchanging with the most influent accounts. It can be useful to detect specific people who seem to have little influence but are in fact able to propagate their message to influent people on twitter.

14_COP20_eigenvector_TOP50

Advertisements

4 réflexions sur “A social network analysis of the 400k tweets about Lima COP 20

  1. Thanks for this–very interesting analysis. We did something similar at the UNFCCC: http://climatetalkslive.org/. You can navigate back to the COP 20 dates to see each day’s activities.

    I wanted to ask what source terms you used to discover and track these accounts? Did you start with a taxonomy of terms and hashtags?

    • Thank you for your interest ! I decided to limit the search to all tweets using #COP20 and #LimaCOP20 hashtags, as there were a lot of them. Otherwise I would probably have used broader terms. I used the software Visibrain in order to access to all the tweets via Gnip.

      ClimateTalkLives is very impressive, do you think a search feature to track specific keywords will be added ?

  2. Pingback: Un análisis de los 400 mil tweets sobre el Lima COP 20 | Turismo en Perú | Intiways

  3. Pingback: Social network analysis of the 4.6 millions #COP21 tweets | DATA VISUALIZATION & SOCIAL NETWORK ANALYSIS

Laisser un commentaire

Entrez vos coordonnées ci-dessous ou cliquez sur une icône pour vous connecter:

Logo WordPress.com

Vous commentez à l'aide de votre compte WordPress.com. Déconnexion / Changer )

Image Twitter

Vous commentez à l'aide de votre compte Twitter. Déconnexion / Changer )

Photo Facebook

Vous commentez à l'aide de votre compte Facebook. Déconnexion / Changer )

Photo Google+

Vous commentez à l'aide de votre compte Google+. Déconnexion / Changer )

Connexion à %s