Research On Multi-dimensional Data Visualization Methods

Multi-dimensional data visualization technology is an effective tool to handle high dimensional data. Visualization technology maps high dimensional data into an intuitive graphical information to display. Users can use visualization results to analyze high dimensional data and find useful hidden information. With the development of science and technology, the scale and complexity of high dimensional data are increasing. Existing multi-dimensional visualization technologies have been unable to meet user’s demand in the display of high dimension data. Thus, research on multi-dimensional data visualization method has important significance.In this thesis, we mainly analyze and improve the disadvantages of two multi-dimension data visualization techniques which are named parallel coordinates visualization and scatter plot visualization. We propose improved visualization methods in order to enhance the results of high dimension data visualization. The main content of this thesis is described as follows:(1)The paper researches multi-dimensional data visualization techniques. And the basic principles, disadvantages and advantages of parallel coordinates visualization and scatter plot visualization are introduced and studied in detail.(2) A parallel coordinates visualization method based on dimension reordering and clustering algorithm is presented. In this method, a new dimension reordering technique is proposed to analyze the relationship between dimensions and optimize the dimension order. It combines with clustering algorithm. Then we show the results by using parallel coordinates visualization. In addition, we also modify label display to overcome the shortcoming of original visualization method which label contains less information. This method uses the amount of similarity and dimension name as a new way to display. At last, to enhance interaction of visualization results, we apply zoom technology in new visualization method. The experimental results show that this visualization method can better display data information, and make the dimension information better meet user’s needs. The improvement of label and interactive visualization provides more convenience for users to analyze data.(3)The paper presents a fast scatter plot visualization method which is based on sampling. We use sampling technique to select representative data, which is used to reduce dimension and data clustering. And we use scatter plot to visualize. This method proposes multi-view collaborative visualization to overcome the disadvantage of the single scatter plot visualization. Finally, a new reordering algorithm is proposed to optimize the relationship between two dimensions. The experimental results show that this method can shorten visualization time and improve the efficiency of visualization. And it can show data from multiple angles and it can analyze the relationship between dimensions. Visualization results are more effective to help users understand data.