Project Prediction Research Based On Modeling Of Scientific Cooperation Networks

An important system for funding scientific research in China is the National Natural Science Foundation(NSFC),which aims to promote the advancement of science and technology in the country by supporting basic and frontier science.NSF not only provides financial support to scientific researchers,but also helps them to gain more opportunities for collaboration.This paper analyzes the information contained in the fund projects based on the data of funded projects and related dissertations of the Natural Science Foundation from1989-2017(the latter retrieves the data of dissertations associated with the fund projects from the Web of Science database),proposes a research cooperation network model based on the evolution of the time-series supernetwork,and constructs the relevant features in the fund cooperation network and the dissertation cooperation network based on the The main factors affecting the fund projects are identified by the prediction model of fund projects.In general,the main contents of this paper include the following three aspects:(i)Empirical analysis of fund project data.In this paper,547,193 fund project data and 1,622,572 dissertation data were collected using NSF project data and its associated dissertation data.The data were first pre-processed and integrated,and then the integrated available data(175,324 data in total)were analyzed,including the changes in the number of funds and the amount of fund grants,the relationship between the change in the number of occupational age and the number of dissertations,the trend of the number of funds and other related factors,and the fund-dissertation bipartite graph network.(ii)Modeling the evolution of scientific cooperation network.In this paper,we propose a research cooperation network model based on the mechanism of temporal hypergraph evolution,which can better simulate the real data and has a high threshold value in the process of network propagation.In this study,we construct a research cooperation network model based on the temporal hypergraph evolution mechanism through an in-depth study of the network temporality characteristics in the research cooperation network structure and adding the memory effect and recession effect of nodes in the process of modeling.The memory effect indicates that nodes will prefer nodes that have been connected in the past when choosing cooperative relationships.The recession effect indicates that nodes are influenced by the memory effect,and the connection relationship closer to the current time step generates a greater weight.Also in this paper,the number of neighbors selected by nodes is improved so that the number of neighbors selected by nodes at different time steps is not always exactly the same.(iii)A prediction model of fund projects based on research collaboration networks.In this paper,we construct a fund collaboration network and a dissertation collaboration network through the collaboration relationships in fund data and dissertation data,respectively,and then explore the features of their network topology and extract the features related to the number of funds from them.Finally,a fund project prediction model is constructed using a machine learning approach,and the features input to the model include features from the empirical analysis part and network features contained in the cooperation network,and the prediction results are analyzed and validated,and the importance of features in different disciplines is explored.