Recent years have witnessed an explosive growth of various social media sites such as online social networks, blogs, microblogs, social news websites and virtual social worlds. These platforms give billions of people unprecedented access and opportunities to create and share different kinds of content. With mass production and consumption of information, people frequently exchange information with each other, as well as exchange their thoughts, opinions and feelings with each other. As such, social influence research is becoming increasingly important. Social influence is defined as change in an individual’s thoughts, feelings, attitudes, and behaviors that results from interaction with another individual or a group. Social influence can be exploited in many important applications including viral marketing, information diffusion analysis, recommender systems, expert finding, link prediction, and growth hacking, among others. Apart from those applications which mainly happened online, many important applications in the physical world can also benefit greatly from social influence research. Such applications include public health behavior promotion, prevention and control of infectious diseases, among others. In this dissertation, we have studied social influence in online social networks from the following three aspects: 1) how to measure social influence among users in online social networks from the perspective of nonlinear dynamic systems; 2) how to simultaneously model social influence among users and interactions among messages in information diffusion; 3) how to optimize the propagation of influence in the context of competing cascades under time critical conditions. The main contributions of this dissertation are as follows. 1. A novel approach to measure social influence among users in online social networks based on nonlinear dynamic system analysis. Existing methods for social influence measurement in online social networks fail to take advantage of nonlinear dynamic information in user behavior patterns. Existing work also requires explicit causal knowledge and assumptions concerning specific user interaction model. Inspired by recent advances in ecology research, we proposed to take the perspective of nonlinear dynamic systems and adopted the Convergent Cross Mapping approach to measure social influence among users in social media. This approach is able to deal with nonlinear influence relation and it does not require any explicit causal ...
修改评论