Health Measurement And Prediction Of The Software Ecosystem In GitHub

In recent years,open source has become an important means to develop core software with high quality and efficiency.With the rapid development of open source software communities and development platforms,the number of available open source software projects has increased rapidly.In a common development platform and environment,open source software projects are interconnected and evolve together,forming software ecosystem.Nowadays,the software ecosystem has become an effective method for building largescale software systems,and has received extensive attention from academia and industry.As the largest open source software community,Git Hub has become an excellent object for studying software ecosystems.Among them,a large number of open source software projects could not continue to be developed and eventually failed,and the software ecosystem formed around open source projects also died out.A health measurement method is needed to assist decision-making.However,there is no universal measurement method to monitor and evaluate the health of software ecosystem.Therfore,this thesis has carried out related research and completed the following tasks:(1)By analyzing the development process of open source software projects,this thesis proposed to consider the health of open source software projects from the three aspects of project roles,project activities and project productivity.Combining the project network topology of the software ecosystem,this thesis constructed an evaluation index to measure the health of the software ecosystem system.(2)Proposed a software ecosystem health measurement method based on principal component analysis,and reveals its dynamic changes in health by tracking the development of open source software projects over time.And got the importance factor of the project according to the network structure of the software ecosystem,which made up for the deficiencies of previous work.(3)Proposed Graph Sage-LSTM,an open source software ecosystem health prediction model based on graph convolutional neural network and long short-term memory neural network,which comprehensively considered the project characteristics in the software ecosystem and the network structure formed by the interaction among projects to predict the health changes of the software ecosystem.(4)To verify the effectiveness of the proposed model,this thesis conducted a series of experiments on the Git Hub open source software ecosystem through the GHTorrent dataset.The experiment results showed the effectiveness of the method.