Information Diffusion in Complex Networks: A Novel Proportion-Based Approach

Abstract

When receiving new information, we often are influenced by the opinions of our peers. In my thesis, I present a novel information diffusion graph theoretic model that considers the proportion of peer influence. In this deterministic model, I consider the implications of two-coloured information systems, where each node can only accept one of two colours, for example, political alignment in a two-party system, adopting a new software vs not, or choosing to believe a rumour. At each discrete-time step of our algorithm, it considers the proportional influence of the neighbourhood of a given node to determine what information colour it chooses to accept. The driving idea is ‘if I am influenced by many sources, each of them individually is less influential’. For instance, a node with 100 neighbours receiving influence from 4 is assessed as receiving 4% proportional influence, whereas a node with 5 neighbours receiving influence from 4 neighbours is receiving 80% proportional influence. This model is simulated on three standard graph models of complex networks: Erd˝os-R´eyni-Gilbert Random G(n, p), Barab´asi-Albert Preferential Attachment, and Watts–Strogatz. The thesis details the decision-tree for the deterministic algorithm. Then, simulations on these graph models are run considering the local graph structure of key nodes. Finally we present an extensive Future Work and Open Problems.