Authors

Akshay Patil

Type

Text

Type

Dissertation

Advisor

Gao, Jie | Skiena, Steven | van de Rijt, Arnout | Liu, Juan.

Date

2015-08-01

Keywords

Computer science

Department

Department of Computer Science.

Language

en_US

Source

This work is sponsored by the Stony Brook University Graduate School in compliance with the requirements for completion of degree.

Identifier

http://hdl.handle.net/11401/77305

Publisher

The Graduate School, Stony Brook University: Stony Brook, NY.

Format

application/pdf

Abstract

As the complexity of online activities increases, social network structures have come to play an increasingly important role in the experience and effectiveness of an individual's online life. These structures exhibit a high degree of dynamism. Also, online platforms have provided us with unprecedented opportunities to study behavior and dynamics of these network structures. Understanding whether/why/when a person will behave in a certain manner can be important in a number of social domains. In our work, we model these network dynamics and design accurate prediction algorithms for these behavioral models. The main challenges that we address in this dissertation are 1) the ability to predict imminent departure events and the probable adverse impact of these events, 2) understanding the processes that drive group growth and stability, and 3) implications of social influence on opinion and relationship formation. Our work offers interesting insights on factors that cause and affect these network events. The methods proposed in our study use a diverse set of features that help us in building richer predictive models that result in more accurate predictions. Another aspect of our research deals with the innovative use of spectral graph theory concepts in unifying activity information of people across different social platforms. We also contribute in the development of a large-scale news and blog analysis engine that provides ready access to a wealth of interesting statistics on millions of people, places, and things across a number of interesting web corpora. The work we present has a wide range of applications: helping spot malicious behavior, forecasting group stability, predicting churn, recommending better content, deanonymizing network identities, and detecting trends in news data. Our techniques have been evaluated and validated on several large-scale, real-world datasets that span different domains. | 177 pages

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.