改进的多变元数据可视化方法

Improved Multivariate Data Visualization Method

针对传统多变元可视化方法--星形坐标法(star coordinates,简称SC)降维过程信息损失较为严重、可视化结果无法体现维度分布信息及手动配置维度轴十分繁杂的不足,提出一种改进的星形坐标法(advanced star coordinates,简称ASC),使用沿直径方向的向量作为维度轴,设计维度轴配置策略优化各维度轴之间的夹角及排列顺序,以减小多维信息对象在改进星形坐标系中与在多维坐标系中坐标差别为准则,使用最优化方法实现对用户有意义的降维运算,将多维信息映射到低维可视空间中.实验结果表明,ASC的可视化结果不仅易于理解,而且能够有效提供维度分布信息,有利于用户发掘隐性知识,基于相关度的维配置策略可以大大减轻用户操作负担。

Star coordinates （SC）,a traditional multivariate data visualization technique,loses much information due to oversimple dimension reduction algorithm.And the SC visualization can’t offer the dimension distribution information.Moreover,the manual dimension axis configuration of SC is too complex.To address these problems,the paper proposes the advanced star coordinates （ASC）,which uses the diameter instead of the radius as the dimension axis,designs the dimension configuration strategy to optimize the order and the angle of dimension axes,and projects the multi-dimensional object to low dimension visual space through the dimension reduction algorithm.And the dimension reduction process is meaningful to user and the algorithm uses the minimum of the object coordinates variation between the multi-dimensional coordinates and the advanced star coordinates as criterion.Experimental results show that the dimension reduction algorithm is highly efficient and suitable for the aggregation with a great amount of high-dimensional data.The dimension configuration strategy relieves the user’s operation burden greatly and helps them detect the connotative characteristics of the multidimensional information aggregation quickly and exactly.The visualization is easy to understand and can express the dimension distribution information effectively,which is helpful for user to view the multi-dimensional information and to discover the implicit information in knowledge discovery process.