150 likes | 159 Views
Explore how social networks evolve over time through models like preferential attachment, and study network formation strategies at a microscopic level. Discover association rules and predict future edges and nodes in co-authorship networks.
E N D
IEEE Intelligent Systems, Special Issue on Social Learning, 2010
Goal: analyzing how the networks evolve over time • Models of social networks • networks follow power-law degree distribution, • have a small diameter • exhibit small-world structure and community structure • Social networks is a function of time • The Obama network • Various Models have been studied • many tools have been built (example)
Dynamic Models • The preferential attachment model: • assumes that new network nodes have a higher probability of forming links with high-degree nodes • creating a “rich-get-richer” effect Network diameter shrinks over time From: Leskovec et al. KDD’05
Questions raised in this paper • Studying network formation strategies • Microscopic level • Graph Evolution Rules • Association rules • From current network configurations • Predict future edges and nodes • Rules • Old-old • Old-new • New-new
Co-authorship Network • http://www.arnetminer.org/viewperson.do?id=486660&name=Qiang%20Yang • http://academic.research.microsoft.com/VisualExplorer#435931
Customer buys both Customer buys diaper Customer buys beer Basic Concepts: Frequent Patterns and Association Rules (from J. Han) • Itemset X={x1, …, xk} • Find all the rules XYwith min confidence and support • support, s, probability Pr(X,Y) • confidence, c,conditional probability Pr(Y|X). • Let min_support = 50%, min_conf = 50%: • A C (50%, 66.7%) • C A (50%, 100%)
Association Rules on Graphs • First, we need to record time stamps on graphs • Nodes • Edges • A large-degree node (label 3), which at time t is connected to four medium-degree nodes (label 2), at time t +1 will be connected to a fifth node. • The collaboration-rich researcher gets richer.