Link Prediction - Statistical Pattern Recognition | CMSC 828G, Study notes of Computer Science

Download Link Prediction - Statistical Pattern Recognition | CMSC 828G and more Study notes Computer Science in PDF only on Docsity!

Link PredictionHossam S. ShararaWalaa Eldin M. Moustafa

OutlineOutline „ OverviewOverview „ Link Prediction Variants „ Deterministic Methods„ Deterministic Methods „ Probabilistic Methods „ Supervised Learning Approaches„ Supervised Learning Approaches „ Feature Construction

ApplicationsApplications „ Identifying the structure of a criminal networky g „ Predicting missing links in a criminal network usingincomplete data.

ApplicationsApplications „ Overcoming the data-sparsity problem ing p^ y precommender systems using collaborative filtering(Huang et al, 2005).

ApplicationsApplications „ To analyze users' navigation history to generateTo analyze users navigation history to generatetools that increase navigational efficiency (Zhu2003) „ ie. Predictive server prefetching^ Picture from corbis.com

ApplicationsApplications „ Monitoring and controlling computer viruses thatMonitoring and controlling computer viruses thatuse email as a vector (Lim et al, 2005). Picture from www robocup dePicture from www.robocup.de

Link CompletionLink^ Completion „ ExampleExample „ When a user buys five books online and the name ofone book is corrupted in transfer. „ A link completion algorithm could infer the name ofthe missing book based on the user's name and theh b k^ h^ b^ hother books she bought.

Link CompletionLink^ Completion „ ExampleExample „ Alice, Bob, and a third person attended a meeting. „ Given people’s previous co-occurrences, who is thep p^ p^

, w third person?

Anomalous Link DiscoveryAnomalous^ Link Discovery „ Rattigan et al, 2005.Rattigan^ et al, 2005. „ Link Prediction: Number of Dyads to be evaluatedincreases quadratically.q^ y „ Networks are sparse^ Æ^ extremely few positivecases. „ Focus on discovering surprising links in the existingones. „ Very few common neighbors, or too distant apart.

Link PredictionLink^ Prediction t^ t

t ttii+

tj- tj i <j <k tk

Shortest PathShortest^ Path „ Negated length of shortest path between x and y.Negated length of shortest path between x and y. „ All nodes that share one neighbor will be linked.

Common NeighborsCommon^ Neighbors „ Newman 2001: The probability of scientistsll b i^ i^ i h h^

b^ f^ h collaborating increases with the number of othercollaborators they have in common.

Adamic/AdarAdamic/Adar Adamic et al 2003 Adamic et al 2003 In Links^ Contacts In Links^ Contacts User 1^ User 2? User 1^ User 2? Out Links^ Text Out Links^ Text

Adamic/AdarAdamic/Adar