Research On Key Nodes Identification Method Of Developer Network Based On Human Dynamic Feature Perception

The key nodes play a vital role in the network,having important influence in the structure and function of the network.For different networks,the key nodes play different roles.For example,the developer network that this thesis mainly studies,in this network,influential developers and projects often play an significant role in driving software development efficiency and quality,and the research trends and results of famous developers can be learned and imitated by others more quickly.Although many scholars have proposed corresponding solutions for nodes identification from different perspectives,there are still some shortcomings.Firstly,the research objects of the existing work are mainly single-layer networks and multilayer singlepart networks,but with the increasing richness of network functions and the gradual improvement of information collection ability,the network types can also be described by single-layer bipartite networks or multilayer bipartite networks.Secondly,few researchers have considered the effect of nodes' continuous interaction over time on node ranking in human dynamics.For the above problems,the work done in this thesis is as follows:(1)Single temporal feature.In human dynamics,the interaction behavior between nodes mainly shows two characteristics in terms of temporal,namely,Burstiness and Memory.Firstly,this thesis constructs a developer-project bipartite network based on the two types of nodes:developers and projects;Secondly,incorporates the temporal characteristics between nodes into the topology of the network,and propose a Temporal-Weighted Bipartite Network model(TWBN);Thirdly,this model is combined with Bi Rank algorithm based on iterative optimization,proposing Burst Bi Rank algorithm and Memory Bi Rank algorithm according to the Burstiness and Memory characteristics respectively;Finally,the two algorithms are used to identify influential developers and important project nodes in the developer network.(2)Temporal feature fusion.For the fusion process of multiple temporal feature between nodes,firstly,this thesis seeks the best proportion of each feature according to the Kendall correlation coefficient,and converts the fused feature value into the weight of the connection between the developers and the projects;Secondly,constructs the corresponding Feature Fusion Weighted Developer-project Bipartite network model(FFWDB),combined with the existing algorithm Bi Rank,proposed the BMBi Rank(Burstiness-Memory Bi Rank)algorithm;Finally,this algorithm is used to identify key nodes in the network and compare it with all benchmark algorithms.(3)The data set used in the experiment of this thesis are all real data screened out on Git Hub.On the basis of this data set,a large number of comparative experiments were conducted between the proposed algorithms and all the benchmark algorithms.Which not only verified the effectiveness and reliability of the algorithms in this thesis to identify key nodes in the network,but also shown that the Burstiness is a better than Memory in measuring the temporal behavior characteristics of nodes.