Generating A Twitter Ego-Network & Detecting Communities

The rest of this post will highlight some methods of extracting insights from our network graph.Reminder that this is a community analysis of an “ego-network” and therefore, not an objective classification of any given user on Twitter but a classification of the role they serve in my network from my perspective.Visualization’s Analyzing PrinciplesInter-related Position — When we are analyzing the communities, it is important to remember that the exact position of a node (south, north, east, west) does not really matter but it’s relative position to other nodes is no accidental..Although these is some randomness involved in the visualization, it is actually modelled through a mixture of attraction, repulsion, and gravity forces..Therefore, strongly connected nodes will attract each other and non-connected nodes will repel.Intra-related Position — The position of nodes within the territory also matters..Nodes further from the central position and deeper into the community reflect more niche behaviour and information-flow properties.Density — More dense collection of nodes implies stronger interconnectivity..Sparsity implies weaker interconnectivity.Modularity — A measure of clustering quality, how disconnected communities are with each other..“Highly modular networks are characterized by a few highly intraconnected clusters that are loosely interconnected..Networks where users are highly interconnected regardless of cluster affiliation are, therefore, less modular.”Bridge Nodes — Nodes found on the border of communities can sometime act like a link, loosely interconnecting the two communities.Outliers — Nodes far away from the structural territory of their assigned community..Sometimes these are mis-categorizations, sometimes far off bridge nodes.Ego-Network FingerprintCommunity #1-2: Real-life People & Their Real-life FriendsThese two communities generally contain my personal connections or friends (‘followings’) of them that I followed, separated between my dominantly high school connections and ones outside high school..Therefore, it is understandable that they are dense (i.e. strong intraconnections)..I see an interesting spatial representation with my high-school connections put away from the centre and thus away from the rest of my generic network (i.e. in the twitter world, we’ve grown apart, understandably, some of them are not even active)..The ones I interact with more often among this community are closer to my node..Furthermore, stronger friends (or those I know to be at least) also appear more closer to each other, relatively compared to others in the community.. More details

Leave a Reply