大数据可视化的挑战与最新进展

发布时间：2018-05-02 20:30

本文选题：大数据 + 可视化　；参考：《计算机应用》2017年07期

【摘要】：大数据的来临增强了可视化的重要性。可视化分析挖掘人类对于信息的认知能力与优势,将人、机有机融合,借助人机交互高效洞悉大数据背后的信息与规律,是大数据分析的重要方法。针对大数据数据量大、维度高、多来源、多形态等特点论述了大规模数据、流数据、非结构和异构数据的可视化方法。首先讨论了大规模数据的可视化技术:1)采用分而治之的原则将大问题分解成较小的任务并采用并行处理的方式解决以提高处理的速度;2)通过聚合、采样、多分辨表示的方法进行数据约简;3)针对高维数据选择若干个视图,在多个角度下生成不同的可视化结果。然后针对监控型、叠加型两类流数据探讨了流数据的可视化过程。最后阐述了非结构化数据以及异构性数据的可视化技术。总之,可视化能够克服计算机自动化分析方法的劣势与不足,整合计算机的分析能力和人们对信息的感知能力,有效地洞悉大数据背后的信息与智慧,但其理论研究成果也非常有限,同时面临着数据规模大、动态变化、维度高、多源异构等方面的挑战,这些也逐渐成为今后的大数据可视化研究的热点与方向。
[Abstract]:Big data's arrival enhanced the importance of visualization. Visual analysis is an important method to explore the cognitive ability and advantage of human being for information, to combine human and machine organically, and to understand the information and law behind big data with the help of human-computer interaction. Aiming at big data's characteristics of large amount of data, high dimension, multi-source and multi-form, this paper discusses the visualization methods of large-scale data, stream data, unstructured data and heterogeneous data. In this paper, we first discuss the visualization technique of large-scale data: 1) decomposing large problems into smaller tasks using the principle of divide-and-conquer and using parallel processing to improve the speed of processing. The method of multi-resolution representation is used for data reduction. (3) several views are selected for high-dimensional data, and different visualization results are generated from multiple angles. Then, the visualization process of stream data is discussed for two kinds of stream data: monitor type and superposition type. Finally, the visualization technology of unstructured data and heterogeneous data is introduced. In short, visualization can overcome the disadvantages and shortcomings of computer automated analysis methods, integrate the analytical ability of computers and people's perception of information, and effectively discern the information and wisdom behind big data. However, its theoretical research results are also very limited. At the same time, it faces the challenges of large scale of data, dynamic change, high dimension, multi-source heterogeneity and so on. These also become the focus and direction of the visualization research of big data in the future.
【作者单位】：宁波工程学院电子与信息工程学院;计算机辅助设计与图形学国家重点实验室(浙江大学);甘肃农业大学信息科学技术学院;
【基金】：国家自然科学基金资助项目(61422211)~~
【分类号】：TP311.13

【参考文献】