Research On Relationship Of Academic Cooperation And Implementation Of Cooperation Relationship Prediction System

The academic cooperation relationship means an attribute of the cooperation relationship between scientists in diverse research fields. With the rapid development of science and technology, the interdisciplinary research is blooming, as the relationships between different research fields have become increasingly close. However, it is common that there are uncertainty and randomicity when researchers find their co-authors, which leads to inefficiency in academic research. It is meaningful to dig into the current co-authorship network, find the intrinsic property and establish a platform that can provide proper co-authors for researchers.Academic cooperation network is actually co-authorship network, which is composed of co-authors and its relationships. Co-authorship network is made of homogeneous co-authorship network and heterogeneous co-authorship network. Every node has the same property and behavior characteristic in homogeneous co-authorship network, relationship among their nodes is simple. But it is different in heterogeneous co-authorship network. The main focus of this paper is finding intrinsic property in current co-authorship network and predicting the co-authorship of future works. Firstly, this paper research current co-authorship network. Secondly, the found property is described in mathematical models and is used for predicting the co-authorship of future works. Then, the validity is evaluated by experiments. Finally, the software engineering approach is used for realizing the co-authorship prediction model.We conduct a deep research to solve the problems above. First some basic conceptions of social network and the definition of co-author network are presented. We analyze the properties of co-author network and the development process and demerit of the approach. We provide an approach to measure edge weight with the consideration of time-delay and impact factor. Through experimental comparison, the effect of two improved approaches is better than former ones.Another prediction model is presented in homogeneous networks, which aims to calculate the probabilities of collaboration between non-collaborated authors. We analyze the different relations between the trust degree of collaboration and the degree of collaboration, the similarity of authors and that of research fields, the trust degree of collaboration and the possibility of recommendation. Under different restrictions of collaboration relations, we are able to calculate the possibilities of collaboration.A model is introduced to predict possible collaboration relations in heterogeneous networks, the aims of which are to find collaborators with the highest possibilities. The five different approaches of the meta path are defined and the relationships between cooperation probability and the meta path are presented in the feather space. Our model is validated with real data. Through experimental comparison, the system can find the best co-author. Finally, we implement cooperation relationship prediction system.