Improving Link Prediction in Dynamic Co-authorship Social Networks

Abstract

The current study is based on a type of social cooperation networks called co-authorship network which its related activities revolve around authors’ citations and their mutual cooperation in researches. Cooperation networks are established to better achieve consistent goals or common goals, and interactions are supported by computers in these networks. Cooperation social networks’ order and system focus on their structure, behavior and dynamicity. The current study aims at improving coauthorship social network links prediction by the use of RapidMiner data mining software. To do this, supervised learning algorithms were used. In comparison to the similar cases, it was concluded that link prediction accuracy is 98.63% and the error rate of the method has decreased in comparition with all other methods in this search. The average of improvement in error rate is 47%