Identifying vital nodes in recovering dynamical process of networked system

Vital nodes identification is the problem of identifying most significant in complex networks, which crucial understanding property networks and has applications various fields such as pandemic controlling energy saving. Traditional methods mainly focus on some types centrality indices, have restricted application cases. To improve flexibility process enable simultaneous multiple mining, a deep learning-based vital algorithm proposed this study, where we train influence score each node by using set to approximate rest network via graph convolutional network. Experiments are conducted with generated data justify effectiveness algorithm. The experimental results show that method outperforms traditional ways adaptability accuracy recover dynamical networked system under different classes structure.