Abstract
Influence maximization is the problem of finding a small subset of seed nodes (individuals in reality) in a social network that maximizes the spread of influence under certain influence cascade models. Most prior algorithms for influence maximization are designed for Online Social Networks (OSNs) and require centralized computation. Directly deploying the above algorithms in distributed Mobile Social Networks (MSNs) will overwhelm the networks due to an enormous number of messages. In this thesis, therefore, we design a new cross-layer strategy to jointly examine the MSN and Mobile Ad Hoc Networks (MANETs) to facilitate efficient seed selection. We extract a subset of nodes as agents to represent nearby friends during the distributed computation. Specifically, we formulate a new optimization problem, named Agent Selection Problem (ASP), to minimize the message overhead transmitted in the MANET. We prove that ASP is NP-Hard and design an effective distributed algorithm. We use real and synthetic datasets for simulation. And the simulation results manifest that the message overhead can be significantly reduced compared with the existing approaches.