Abstract

Online social networks (OSNs) are becoming increasingly prevalent in human life. However, the factors that influence human behavior in OSNs are not well understood. Prior research assumes nodes in OSNs to be static and unchanged over time. Unfortunately, existing methods do not sufficiently consider important phenomena in collective social behavior. This research developed a new theory-driven framework for modeling temporal OSNs. Three models were built based on the framework to represent the effects that may possibly shape temporal online social networks. The random network model (RNM) represents randomness in the interaction among networks nodes and serves as a baseline model. Two window-based models, namely Average Aggregation Model (AAM) and Exponential Aggregation Model (EAM), represent respectively the herd effect, and the recency and primacy effects in human social behavior. To evaluate the models, we examined 42 OSNs of the GitHub software development community that committed a total of 5,499,611 events during Jan. 25-30, 2017. The results show that both EAM and AAM achieved superior performance compared with RNM. The research makes three contributions: (1) developing a new theory-driven framework for characterizing online social behavior, (2) developing and validating three models for simulating temporal OSNs using social and cognitive theories, and (3) providing empirical findings of simulating GitHub OSNs and user behavior.

Share

COinS
 

A Theory-Driven Framework for Modeling Temporal Online Social Networks of GitHub

Online social networks (OSNs) are becoming increasingly prevalent in human life. However, the factors that influence human behavior in OSNs are not well understood. Prior research assumes nodes in OSNs to be static and unchanged over time. Unfortunately, existing methods do not sufficiently consider important phenomena in collective social behavior. This research developed a new theory-driven framework for modeling temporal OSNs. Three models were built based on the framework to represent the effects that may possibly shape temporal online social networks. The random network model (RNM) represents randomness in the interaction among networks nodes and serves as a baseline model. Two window-based models, namely Average Aggregation Model (AAM) and Exponential Aggregation Model (EAM), represent respectively the herd effect, and the recency and primacy effects in human social behavior. To evaluate the models, we examined 42 OSNs of the GitHub software development community that committed a total of 5,499,611 events during Jan. 25-30, 2017. The results show that both EAM and AAM achieved superior performance compared with RNM. The research makes three contributions: (1) developing a new theory-driven framework for characterizing online social behavior, (2) developing and validating three models for simulating temporal OSNs using social and cognitive theories, and (3) providing empirical findings of simulating GitHub OSNs and user behavior.