Let&#39;s unleash the latest and greatest tools of network theory on the glamorous world of Eurovision. &nbsp;Everyone knows that there are strong bonds in the competition, but who are they between? Let&#39;s find out…<br />
&nbsp;<br />
The most well known network communities these days are social networks where groups of friends define communities. These communities in turn may or may not be connected to one another by &#39;weaker&#39; ties. For example, our working together at Capgemini gives our friends second degree connections to all our work colleagues. Our current work colleagues are also second degree connected to people at all the organisations that we&rsquo;ve worked at before. There are some <a href="http://socilab.com/">great mapping tools</a> available to help you map your own networks &ndash; maybe you&rsquo;ve tried some?<br />
&nbsp;<br />
It is quite straightforward to represent the Eurovision voting in a network with countries linked by the votes they make for one another. Here at FIO we considered all the votes between 2000 and 2014 to see how strong the voting links between countries are.<br />
&nbsp;<br />
We used the <a href="http://www.mapequation.org/">infoMap algorithm</a> – a community detection algorithm. It uses programming rules and a simple optimisation procedure to partition a network into sets of non-overlapping communities. If no communities are found the network can be considered as one big community<br />
&nbsp;<br />
We were mildly disappointed when infoMap reported that there isn&rsquo;t a community structure in Eurovision. This can&#39;t be, we thought &ndash; everyone knows that there are countries which always vote for one another. So we decided to investigate further. Help arrived in the form of a method developed by some people in our network (<a href="http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.00179… freely available</a>). This is a way to discover any statistically significant relationships in a network. For Eurovision we tested for statistical significance against the hypothesis that countries assign votes to each other at random. Lo and behold, using this method, we found a network with non-overlapping communities &ndash; as expected given what we know about strong voting relationships.<br />
&nbsp;<br />
The nice thing about this is that the statistically validated links define a well connected system. This is in contrast to other systems we&#39;ve analysed (i.e. banks) where networks are often many disconnected components.<br />
&nbsp;<br />
Anyway &ndash; if you&rsquo;re waiting for the results &hellip; we see 9 communities that are fairly well connected between them. Here is a table of community composition ordered by size:<br />
1. Ukraine, Azerbaijan, Russia, Georgia, Moldova, Belarus, Armenia, Poland, Czech Republic.<br />
2. Turkey, Bosnia and Herzegovina, Macedonia, Albania, Serbia, Croatia, Slovenia, Montenegro.<br />
3. Iceland, Denmark, Sweden, Norway, Finland.<br />
4. Greece, Cyprus, Bulgaria.<br />
5. Romania, Hungary, Italy, San Marino.<br />
6. Spain, Portugal, France, Germany, Israel, Andorra, Monaco, Austria,Switzerland.<br />
7. Lithuania, Estonia, Latvia, Slovakia.<br />
8. Ireland, United Kingdom, Malta.<br />
9. Belgium, Netherlands.<br />
&nbsp;<br />
<br />
<iframe width='100%' height='500px' frameBorder='0' src='https://a.tiles.mapbox.com/v4/sleuthy.m6l6ob5g/attribution,zoompan,zoomw…
<br />
Although we&rsquo;re not politicians here at FIO, these start to make some sort of sense as they seem to map to some sort of cultural/linguistic/geographical structure of Europe. The majority of the clusters are easily interpretable….<br />