Abstract: According to the requirements of visual analysis of high dimensional data, this paper proposes a radius coordinate visualization method that can analyze high dimensional data in the reduced space by maximum likelihood estimation of intrinsic dimension so as to apply a few attributes in radius coordinate visualization. The radius coordinate visualization can reveal interesting relations between classes and features, integrate various machine learning methods to classify dataset in optimal projection that obtained from different variable arrangement. Experimental results applied on the six datasets in UCI database show good performance of accuracy and visualization.

Key words: visualization, radius coordinate, high dimensional data, estimation of intrinsic dimension

摘要： 针对模式分类算法不直观的问题，提出一种基于径向坐标可视化分析高维数据的方法。由最大似然原理估计高维数据的本征维数，用较少的变量结合径向坐标可视化方法对高维数据进行可视化降维分析。在径向坐标中揭示高维数据集中类别和特征间的关系，寻找基于不同特征排列顺序的最优映射，并结合多种机器学习方法对数据集进行分类。应用于UCI数据库中的6个数据集的结果表明，该方法具有较好的可视化和分类效果。

关键词: 可视化, 径向坐标, 高维数据, 本征维数估计

CLC Number: 

• TP18