Open Source Ecosystem Identification And Evolution Based On Graph Convolutional Network

The evolutionary analysis of open-source ecosystem has been a hot research problem in software development and academic research.Identifying open-source ecosystems from open-source communities is a prerequisite for the evolutionary analysis of open-source ecosystems.With the rapid development of open-source communities,it is difficult to accurately identify open-source ecosystems in the increasingly complex open-source project networks using traditional identification methods.Evolutionary analysis based on the identification results of open-source ecosystems can discover the changes in the process of its development and promote the benign development of open-source ecosystems.Therefore,this thesis aims at the identification and evolution of open-source ecosystems and conducts relevant exploratory work.In this thesis,we follow the process of open-source project network construction,open-source ecosystem identification and open-source ecosystem evolution analysis.Firstly,we propose a project relevance calculation method that integrates project semantic information and developer behavior information,and construct a more comprehensive open-source project relevance network with projects as nodes.Then we design an open-source ecosystem identification model based on graph convolutional network,and use the powerful characterization ability of graph convolutional network to integrate node features into the identification target,and realize the accurate identification of ecosystem in open-source project network.Finally,this thesis defines an open-source ecosystem evolution stage model,designs multi-dimensional indicators from two levels of open-source projects and developers,and completes the analysis and research on the evolution characteristics of each stage of the open-source ecosystem.The experimental results show that the open-source ecosystem identification method based on graph convolutional network can effectively identify the open-source ecosystem from the complex and huge open-source project network,and the overall performance is greatly improved compared with the traditional community identification algorithm.By analyzing the evolution of the identified open-source ecosystem with multiple indicators,we obtain the change pattern of each indicator in different evolutionary stages of the open-source ecosystem,which provides a theoretical basis for promoting the benign development of the open-source ecosystem.