Open Collections

Social influence and its applications : an algorithmic and data mining study

University of British Columbia

Social influence occurs when one's actions are affected by others. If leveraged carefully, social influence can be exploited in many applications like viral marketing (or targeted advertising in general), recommender systems, social network analysis, events detection, experts finding, link prediction, ranking of feeds etc. One of the fundamental problems in this fascinating field is the problem of influence maximization, primarily motivated by the application of viral marketing. The objective is to identify a small set of users in a social network, who when convinced to adopt a product will influence others in the network leading to a large number of adoptions. The vision of our work is to take the algorithmic and data mining aspects of viral marketing out of the lab. We organize ours goals and contributions into four categories: (i) With the ever-increasing scale of online social networks, it is extremely important to develop efficient algorithms for influence maximization. We propose two algorithms -- CELF++ and SIMPATH that significantly improve the scalability. (ii) We remark that previous studies often make unrealistic assumptions and rely on simulations, instead of validating models against real world data. For instance, they assume an arbitrary assignment of influence probabilities in their studies, which focused more on algorithms than on validity with respect to real data. We attack the problem of learning influence probabilities. In another work, we propose a novel data driven approach to influence models and show that it predicts influence diffusion with much better accuracy. (iii) Next, we propose alternative problem formulations -- MINTSS and MINTIME and show interesting theoretical results. These problem formulations capture the problem of deploying viral campaigns on budget and time constraints. In an additional work, we take a fresh perspective on identifying community leaders using a pattern mining approach. (iv) Finally, we examine applications of social influence. We begin with the application of viral marketing. We show that product adoption is not exactly influence. Given this, we develop a product adoption model and study the problem of maximizing product adoption. Furthermore, we propose and investigate a novel problem in recommender systems, for targeted advertising -- RECMAX.

Text

2013-02-19

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/43935

Computer Science

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/4.0/