Research On Visual Analysis Method Of Temporal Information Networks Based On Multi-level And Multi-granularity Characteristic

Network space has become an invisible battleground in the modern information warfare and has played a vital role in the transmission of information and intelligence.Intuitively and vividly representing information network help commanders gain a deeper understanding of the network while making further analysis.As the main carrier of data transmission,information network are multi-level,multi-granularity and time-varying.With the continuous development of information technology,network scale has enlarged rapidly and network structure has become more complicated.However,inter relations within network layers and the temporal property have increased the difficulties in understanding and analyzing information network.As a result,it presents challenges for the network visualization technology.This research focuses on complex temporal information network from the perspective of visual analysis on vital nodes,multi-granularity structure organization,dynamic layout algorithm and interactive visual analytics methods.The main goal is to enhance users' cognitive competence and analytical ability for the information network.The main contributions are the follows:Firstly,a multi-level based vital nodes visual analysis method is proposed.Base on the characteristics of multi-layer structure and inter relations within network layers,a multi-layer network model is firstly built from the perspective of network structure and interdependent relations.An integrated importance index is then proposed considering both structure and business properties and a visual analysis strategy is designed that could adapt to index changes.A DOI-based vital nodes associative analysis method is proposed to combine node importance and DOI degree while supporting interactive analysis for filtering and querying demands within large scale network according to different tasks.Secondly,a content-based multi-granularity organization method is presented for the purpose of helping users to understand network structure rapidly and intuitively.A content-based clustering algorithm is firstly proposed that locates kernel communities as visual abstraction with consideration of structural and attribute properties which could represent the main structures accurately and providing guidance for the exploration.Based on the achieved kernel communities,a details on demand visual analysis strategy is designed.This provides global context and local details according to users' interests,which enable the discovery of hidden patterns in networks without priori knowledge.Thirdly,a multi-constraint dynamic layout accelerated algorithm is proposed in consideration to both node property and network structure.Two constraint parameters,an evolution parameter and an energy parameter,are introduced to achieve an aesthetic layout while maintaining a coherent mental map.To cater to the need for real time display requirement,a multilevel spreading algorithm is proposed base on the multi-granularity network structure to further accelerate the layout algorithm efficiency in large network scale.Fourthly,a multiview-based interactive visual analytics strategy is proposed according to the visual analysis requirements and analysis task in information network application scenario.A synthetic method is proposed for the visualization of the multi-variate dataset.The proposed method reduces visual clutters and allows interactive analysis through a tightly coupled exploration strategy.In order to reveal the network structure and interdependent relations within network layers,a 2.5D method is proposed so as to satisfy readability of the network structural features and demonstrates the relations in the network.Further interaction strategy allows discovery and analysis of vital nodes in the interdependent network.Finally,a prototype system of temporal information networks based on multi-level and multi-granularity characteristic named InfoVS is designed and implemented.This gives experimental supports to the visualization and visual analytics technologies proposed in this thesis.Some parts of this system have been implemented in related projects.