Social approach for context analysis: Modelling and predicting social network evolution using homophily

Abstract

Understanding the user’s context is important for mobile applications to provide personalized services. Such context is typically based on the user’s own information. In this paper, we show how social network analysis and the study of the individual in a social network can provide meaningful contextual information. According to the phenomenon of homophily, similar users tend to be connected more frequently than dissimilar. We model homophily in social networks over time. Such models strengthen context inference algorithms, which helps determine future status of the user, resulting in prediction accuracy improvements of up to 118% with respect to a naïve classifier.