The Design And Implementation Of Visual Analytic System For Graph-Structured Data

In computer science, graphs are now widely used for data modeling in application domains for which identifying relationship patterns, rules, and anomalies is useful. These domains include the social networks, the Semantic Web, bibliographical networks, and knowledge graphs, among many others. Graphs are abstract data structures that model structural relationships among objects. Mining the hidden structure or model from the graph-structured data is the basis for solving many problems. Visual analytic methods are effective approaches to assist data mining in such area, which can integrate the visual capacity of human and computing power of the machine. Therefore, the visual analytic methods through interactive user interfaces are useful and helpful approaches for data analysis. This paper is based on the visual exploration model for researching and designing a prototype system for displaying the graph-structured data. The visual layouts and interaction methods is based on the visual analytical model, which used to assist users to find knowledge from data or models. To improve the adaptability and scalability of the visual analytical system, the component-driven programming model is used in the system design. Therefore, the main contributions are summarized as follows:(1) The interactive methods and layouts in the visual analytical system are designed for the graph-structured data. The interactive methods are designed to assist users to manipulate the visual results in the screen, and the advanced visual exploration is available via these interactive methods. The graph layouts are designed to visual the common used graph patterns. User also can add their own visual graph layouts to the visual analytical system by defining a new layout component.(2) The architecture of the visual analytical system for graph-structured data is designed to provide flexible, reusable, scene-oriented analysis capability. The architecture that designed in this paper follows the main principal of the separation of front-end and back-end which decoupled the system in physical layer. And followed the principal of component-driven programming and unidirectional data flow to design the front-end for visual components decoupling and high engineering.(3) The optimization plan for the system architecture under the massive data is proposed in this paper to dress the challenges under big data processing and visualization. In order to effectively deal with large-scale data analysis, this paper presents a visual analysis system based on a distributed architecture model program, allows the use of a distributed NoSQL database systems scale storage layer, and allows the visualization and analysis system and other third-party Big Data processing tools integrated, so as to enhance the system of mass data processing and analysis capabilities to provide data to support the final visualization.(4) The prototype system is implemented under the real-world requirements for visual analysis, and the experimental study of the analysis capabilities of the visual analytical system in large data sets, build flexibility, system responsiveness. In the implementation section, the visual analytical system for visualization the user behaviors are implemented under massive web log data sets. Through a set of comparative experiments to verify the advantages of the system in building flexibility for multi-user analysis capabilities, visualization capability of rapid response.