lpetrich
Contributor
Network of Thrones
and some others:
Game of Nodes: A Social Network Analysis of Game of Thrones | gameofnodes
A song of nodes and edges – Network analysis in Game of Thrones – Data Skunkworks
Network analysis of Game of Thrones family ties
Another Game of Thrones network analysis - this time with tidygraph and ggraph
[1707.05213] Balance of thrones: a network study on "Game of Thrones" that unveils predictable popularity of the story
Andrew Beveridge and Jie Shan, authors analyzed "A Storm of Swords", the third book in the series, because "the main narrative has matured, with the
characters scattered geographically and enmeshed in " counting a connection between characters if their names or nicknames are within 15 words of each other. A connection does not mean that the characters are friends, only that they interact with each other. With their count of connections, they then looked for which characters interact the most with other characters, and what cliques of characters there are.
They found seven main cliques: "Lannisters and King’s Landing, Robb’s army, Bran and friends, Arya and companions, Jon Snow and the far North, Stannis’s forces, and Daerenys and the exotic people of Essos." They used a measure of cliquishness that is essentially the departure from randomness of the connections, where the randomness is estimated from how many connections each individual has. The authors used Louvain's method of finding cliques, a fast method for finding the division into cliques that maximizes the total cliquishness.
The King's Landing clique had 37% of the network, and it in turn could be divided into four subcliques.
Then the authors used various measures of centrality. A simple one is "degree", how many characters are connected to each one. The most connected character is Tyrion of King's Landing. Then "weighted degree", weighting by how many connections between pairs of characters. Tyrion won there also, and the others do not get much change.
The next measures take into account more indirect connections, by weighting more connected characters more. The "eigenvector" approach gives every character an importance that is the weighted sum of the importance of each character that this character is connected to. It gives a boost to Robert and Stannis, because their relatively few connections are to prominent characters. Bran, Jon, and Daenerys score lower because they are more isolated from the main action. Then Google's PageRank algorithm, which spreads each character's importance among its connected ones. PageRank gives a big bump to Daenerys.
The next one is closeness, how near each character is to other characters. Every character has about the same distance from the others, except for more distance Daenerys. Then betweenness, how likely a character is to be in a path that connects other characters. Tyrion is beaten by Jon and Robert here. Daenerys is next, behind those three.
and some others:
Game of Nodes: A Social Network Analysis of Game of Thrones | gameofnodes
A song of nodes and edges – Network analysis in Game of Thrones – Data Skunkworks
Network analysis of Game of Thrones family ties
Another Game of Thrones network analysis - this time with tidygraph and ggraph
[1707.05213] Balance of thrones: a network study on "Game of Thrones" that unveils predictable popularity of the story
Andrew Beveridge and Jie Shan, authors analyzed "A Storm of Swords", the third book in the series, because "the main narrative has matured, with the
characters scattered geographically and enmeshed in " counting a connection between characters if their names or nicknames are within 15 words of each other. A connection does not mean that the characters are friends, only that they interact with each other. With their count of connections, they then looked for which characters interact the most with other characters, and what cliques of characters there are.
They found seven main cliques: "Lannisters and King’s Landing, Robb’s army, Bran and friends, Arya and companions, Jon Snow and the far North, Stannis’s forces, and Daerenys and the exotic people of Essos." They used a measure of cliquishness that is essentially the departure from randomness of the connections, where the randomness is estimated from how many connections each individual has. The authors used Louvain's method of finding cliques, a fast method for finding the division into cliques that maximizes the total cliquishness.
The King's Landing clique had 37% of the network, and it in turn could be divided into four subcliques.
Then the authors used various measures of centrality. A simple one is "degree", how many characters are connected to each one. The most connected character is Tyrion of King's Landing. Then "weighted degree", weighting by how many connections between pairs of characters. Tyrion won there also, and the others do not get much change.
The next measures take into account more indirect connections, by weighting more connected characters more. The "eigenvector" approach gives every character an importance that is the weighted sum of the importance of each character that this character is connected to. It gives a boost to Robert and Stannis, because their relatively few connections are to prominent characters. Bran, Jon, and Daenerys score lower because they are more isolated from the main action. Then Google's PageRank algorithm, which spreads each character's importance among its connected ones. PageRank gives a big bump to Daenerys.
The next one is closeness, how near each character is to other characters. Every character has about the same distance from the others, except for more distance Daenerys. Then betweenness, how likely a character is to be in a path that connects other characters. Tyrion is beaten by Jon and Robert here. Daenerys is next, behind those three.