| In the real experience world,there are many complex systems in which the exact relationships between individuals are incomplete or difficult to observe.However,each individual in the system has corresponding time series data that can reflect the connections between them.We refer to such systems as hidden edge systems,such as the World Wide Web,social network and Stock fund network are relatively common hidden edge networks.Like other complex systems in the real world,hidden edge systems also have a very clear community structure.The goal of this thesis is to detect the community structure in hidden edge systems,so as to have a deeper understanding of the internal operating mechanism of hidden edge systems and effectively predict and control them.Therefore,this thesis first proposes a time series analysis method based on ADTW distance on the basis of dynamic time warping distance,in order to represent the relationships between individuals in the system.Then,combined with some observable relationships in the hidden edge system,a network model of the hidden edge system is constructed.Finally,some optimization is made on the Louvain algorithm,and an optimized Louvain community detection algorithm is proposed to detect the hidden edge network and obtain the network community structure of the hidden edge system.This thesis selects the Shanghai 180 Index constituent stocks as the experimental subjects for research.Based on the above methods,the network community structure obtained through modeling and community detection can better reflect and explain the real situation.Moreover,the network community structure obtained through validation in the test set can also predict the development and change trends of individuals in the system in the next period of time.Through theoretical analysis and experimental verification,the method proposed in this thesis can accurately reflect the true situation of hidden edge systems,thereby effectively explaining and predicting hidden edge systems. |
