Detecting Influential Nodes Incrementally and Evolutionarily in Online Social Networks

Detecting influential nodes and understanding their evolution patterns are very important for information diffusion in online social networks. Although some work has been done in literature, it is still not clear that: (1) how to measure the influential degree of nodes for information diffusion, and (2) how influential nodes evolve during the diffusion process. To address the two challenges, we identify an incremental approach to measuring users' influential degrees, detecting local and global influential nodes, and analyzing their evolution patterns, for which we propose three methods to partition time window. The three methods are the uniform time window, the non-uniform time window, and the uniform retweets number window, respectively. We apply our model on real data set in Sina weibo and conduct extensive analyses, from which we gain several interesting findings. We also validate the effects of our method, by comparing the influence spread with our detected influential nodes as seeds, to other seed selection algorithms, which shows that our work has better performance.