一种多维数据的聚类算法及其可视化研究

提出了一种基于主次属性划分的聚类方法和一种新的数据可视化方法.首先,利用数据的主属性和次属性的特征值对数据集进行聚类;然后,采用彩色刺激光谱投影到RGB颜色空间的原理,通过色度学中麦克斯韦的三角平面坐标色度图对各聚类结果进行可视化显示.实验证明了文中方法算法简单、容易实现,可视化结果有利于用户全面地理解数据,为数据的预测、决策起到重要作用.

Clustering for Multi-Dimensional Data and Its Visualization

The paper proposes a property classification based clustering method and a novel data visualization method.First,a data set is clustered with the feature of its prime attributes and secondary attributes;second,under the principle of mapping color stimulus spectrum to RGB color space,every clustering is visualized by the chromaticity diagram of J.C.Maxwell's triangle plane coordinate.The experiments show that the algorithm is simple and easily implemented.The visualization method is helpful for users to view data,and has an important role in prediction and decision.