Study of Geo-Social Networks, Social Cascades and Applications - PDF document

Study of Geo-Social Networks, Social Cascades and Applications� Cecilia Mascolo Computer Laboratory, University of Cambridge Joint work (mainly) with: Salvatore Scellato Location, location, location.� Plethora of new services: increasingly important, And social networks. excitingly new. 2

Growing levels of popularity, More and more importance. millions of users and the attention of media and investors. Information, social What is the effect of geography over social structures structure and space. And information flows?

Probability of friendship Distance matters. decreases with distance. Interesting questions... and potential applications What ’ s the spatial structure • of these online social networks? • Can we discriminate between users according to their attitude towards long- range ties? How is information • spreading across space over social links? • Can we design applications exploiting location information in social networks? Flickr: Oberazzi

Geographic structure of location-based online social networks Geographic Social Network • Given a graph G=(N,K) and the geographic location of the nodes: • Place all nodes in a 2D metric space adopting great-circle 1,120 km distance on the Earth. • Assign a weight to each edge equal to the geographic distance between the two 1,070 km nodes. 210 km

Geo-social measures • How close are the neighbors of a given node to the node itself? Node locality User A • How spatially inter-connected are the neighbors of a given node? Geographic clustering coefficient User B User C User D Geographic Properties

Geo-social Metrics Geographic spreading of information on location-based online social networks

URLs spread on Twitter Interaction between tweets, YouTube videos and Content and drive Web traffic. Delivery Networks. Geo Social Cascades • We studied user locality and geo clustering but how are the geographical properties of the users participating in an information cascade? We define two measures: • the geodiversity is the geometric mean of the distances between all users in the tree • the georange is the geometric mean of the distances between each user and the root of the tree

Geographic social cascade spreading • We have collected more than 3 millions geo-tagged tweets containing URLs of about 1 million YouTube videos. • Around 10% of social cascade steps cover less than 1 km , with more than 30% shorter than 1,000 km. • the final properties of a cascade can be estimated even from the very first users involved in the initial stages. Geosocially inspired system design • Location information extracted from URL-based social Location Country Servers Location Country Servers cascades is used to improve Washington USA 552 Frankfurt Germany 314 cache replacement strategies Los Angeles USA 523 London UK 300 for multimedia files in a CDN. New York USA 438 Amsterdam Netherlands 199 Chicago USA 374 Tokyo Japan 126 San Jose USA 372 Toronto Canada 121 Dallas USA 195 Paris France 120 • Geographic locality of online Seattle USA 151 Hong Kong Hong Kong 83 Atlanta USA 111 Changi Singapore 53 social interactions can be Miami USA 111 Sydney Australia 1 exploited to do pre-fetching of Phoenix USA 3 Web content, caching of normal HTTP traffic, datacenter design and storage partitioning.

Results of our study Cache hits • Effect of power-law popularity : even small cache sizes achieve high levels of cache hits • Effect of cache size : the larger the better but with a plateau (diminishing returns). • Effect of geographic weights : both Performance gain (geocascade) Geosocial and Geocascade popularity weights provide performance gains. • Effect of workload size: geographic weights see higher performance gains with higher workloads. Improving link prediction systems for location- based online social networks

The importance of Place-Friends • Location-based user activity has a great potential to improve link prediction systems. • We have collected four monthly snapshots of data containing user profiles, friends list and check-ins. • We found that about 30% of new links are added among “ place-friends ” , or users who check-in at the same places. Design of a link prediction system We make two observations: • new links overwhelmingly appear between people who share a friend • places with lower entropy foster more social ties among people going there Our design proposal builds on two key choices: • reducing the prediction space by focusing on friends-of-friends and place-friends only • including check-in information in the prediction system itself.

Conclusions • We are studying social networks with location-based information • We have investigated their structure with two new geo-social metrics which take into account both social connections and geographic distance: node locality and geographic clustering coefficient. We have highlighted differences between purely location-based social network services and other online social communities. • We have presented two possible applications of these models • information propagation over space on Twitter can be used to improve planetary-scale CDNs. • user check-ins at places provide invaluable information to be exploited in link prediction systems for geographical online social networks. References • Exploiting Place Features in Link Prediction on Location-based Social Networks . Salvatore Scellato, Anastasios Noulas, Cecilia Mascolo. In Proceedings of 17th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD 2011). San Diego, USA. August 2011. • Socio-spatial Properties of Online Location-based Social Networks � Salvatore Scellato, Anastasios Noulas, Ranaud Lambiotte and Cecilia Mascolo. In Proceedings of Fifth International AAAI Conference on Weblogs and Social Media (ICWSM 2011). Barcelona, Spain, July 2011. • Track Globally, Deliver Locally: Improving Content Delivery Networks by Tracking Geographic Social Cascades � Salvatore Scellato, Cecilia Mascolo, Mirco Musolesi, Jon Crowcroft.� In Proceedings of 20th International World Wide Web Conference (WWW 2011). Hyderabad, India. March 2011. • Distance Matters: Geo-social Metrics for Online Social Networks � Salvatore Scellato, Cecilia Mascolo, Mirco Musolesi, Vito Latora.� In Proceedings of the 3rd Workshop on Online Social Networks (WOSN2010). Co- located with USENIX2010. Boston, MA. June 2010.

Thanks!� Questions? Cecilia Mascolo http://www.cl.cam.ac.uk/users/cm542/