4th European Conference on Software Architecture
URL with Digital Object Identifier
The exploration of online social networks whose members share mutual recommendations and interactions is a time-dependent and contextual-based process which aims to predict the social status among members, ultimately improving the network's discoverability to achieve societal gain. To address the difficulties associated with the process, this article presents an integrated recommender model whose statements are time-dependent, interaction-aware, and social context-sensitive. The originality of the proposed model stems from the integration of the predictive recommender, social networks, and interaction components. Each model is developed based on: (1) a time history and decay algorithm to consider the decreasing intensity of recommendations among members over time; (2) a predictive aggregating function for improved assessment of recommendations for various social contexts; and, (3) a homophily algorithm to estimate the degree in which a recommender-based contact between similar people occurs to dissimilar people. Details of the framework are described, a recommender search strategy methodology is devised, and a case study is used to demonstrate its capabilities. Possible extensions are then outlined.