基于降维技术的高维数据可视化研究与实现

发布时间：2018-10-05 08:00

【摘要】：数据是人类记录信息的重要形式,而可视化是一种以图形符号等更加直观形象的方式来传达信息的技术。可视化使人类获取知识变得更加高效,它是人类获取信息的重要渠道。随着信息时代的到来,数据爆炸式增长,数据变得越来越复杂,数据维度较高。如何将高维数据可视化并反映数据特征和规律是当今可视化领域的难点和热点问题。本文着眼于利用可视化技术将高维数据可视化,帮助用户发现数据之间的关系,数据与维度之间的关系。本文的主要研究工作如下:(1)提出了一种高维数据可视化方法。由于维度爆炸及可视空间有限,用户很难可视化并探索、分析高维数据。早期的一些工作通过传统的降维方法产生隐式维度,不但损失了一部分信息,更重要的是这些隐式维度很难为用户所理解。因此,本文提出一种高维数据可视化方法,该方法结合用户有限的知识导出符合用户知识的维度,并重新组织数据。然后,利用本文基于散点图矩阵扩展的可视化呈现方法散点饼图矩阵来可视并探索重新组织后的数据。该方法可使用户发现已知数据与未知数据的关系,未知数据与导出维度的关系。实验验证了该方法的有效性。(2)设计并实现了一个高维数据可视化工具。本文通过对现有的可视化工具调研分析发现,目前存在较少的高维数据可视化工具,而且现有的高维数据可视化工具用户探索流程不够完善,不易扩展新的高维数据可视化方法。因此,亟需一个实用的高维数据可视化工具,帮助用户更好地探索、分析高维数据。本文设计并实现了一个高维数据可视化工具,该工具提供一个完整的用户可视探索数据的流程,用户可结合交互,完成对数据的探索,并可保存数据探索结果,供用户分享、查阅。而且,用户可以基于该工具,针对特定的应用扩展新的高维数据可视化方法。
[Abstract]:Data is an important form of human record information, and visualization is a technique to convey information in a more visual way such as graphic symbols. Visualization makes the acquisition of knowledge more efficient, and it is an important channel for human to obtain information. With the arrival of the information age, data explosive growth, data become more and more complex, data dimension is higher. How to visualize the high-dimensional data and reflect the characteristics and laws of the data is a difficult and hot issue in the field of visualization nowadays. This paper focuses on using visualization technology to visualize high-dimensional data to help users discover the relationship between data and dimension. The main work of this paper is as follows: (1) A high dimensional data visualization method is proposed. Due to dimensional explosion and limited visual space, it is difficult for users to visualize and explore high dimensional data. Some of the earlier work generated implicit dimensions through traditional dimensionality reduction methods, which not only lost some information, but also were difficult for users to understand. Therefore, this paper proposes a high-dimensional data visualization method, which combines the limited knowledge of the user to derive the dimension that conforms to the user's knowledge, and reorganizes the data. Then, the scattered pie chart matrix is used to visualize and explore the reorganized data by using the visual representation method based on the expansion of scatter plot matrix in this paper. This method enables users to discover the relationship between the known data and the unknown data, and the relationship between the unknown data and the derived dimension. Experiments show that the method is effective. (2) A high dimensional data visualization tool is designed and implemented. Through the investigation and analysis of the existing visualization tools, it is found that there are few high-dimensional data visualization tools, and the existing high-dimensional data visualization tools user exploration process is not perfect. It is difficult to extend the new high dimensional data visualization method. Therefore, a practical high-dimensional data visualization tool is urgently needed to help users explore and analyze high-dimensional data better. In this paper, a high dimensional data visualization tool is designed and implemented. The tool provides a complete flow of user visual exploration data. The user can combine interaction, complete the data exploration, and save the data exploration results for users to share. Access. Furthermore, based on the tool, users can extend a new high-dimensional data visualization method for specific applications.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【相似文献】