A Centrality Estimation Method Based on Hidden Markov Model in Social Delay Tolerant Networks

Complex network analysis method has recently been proposed to solve the problem of contact prediction in Delay Tolerant Networks (DTNs). Centrality Estimation remains as an important issue in such scenarios. Existing schemes such as single window and cumulative window centrality estimation, however, can not predict the node contact capability accurately due to the fact that the messages always have a specific lifetime associated with them. In this paper we proposes a new centrality estimation method based on simplified Hidden Markov Model (HMM) to address this challenge. The historical and current centrality information is used to compute the comparative centrality of two encountering nodes before the expiration of a message. Experimental results based on real traces show that our approach outperforms the existing schemes in terms of estimation accuracy, leading to significant improvement on delivery efficiency.