M. Cebrian, E. Frias-Martinez

Abstract

Word-of-mouth communication has been shown to play a key role in a variety of environments such as viral marketing and churn prediction. A family of algorithms, generally known as information spreading algorithms has been developed to model such pervasive behavior. Although these algorithms have produced good results, in general, they do not consider that the social network reconstructed to model the environment of an individual is limited by the information available. In this paper we study how the missing information (in the form of missing nodes and/or missing links) affects the spread of information in the wellknown Dasgupta et al. (2008) algorithm. The results indicate that the error made grows logarithmically with the amount of information (links, nodes or both) unknown