Evaluation of a Cluster Bundling Technique for Parallel Coordinates Plot Based on Distribution Feature

In order to solve problems caused by cluster separation during bundling,such as layout and cluster center deviation,we proposed an evaluation method on cluster bundling from distribution attributes.Firstly,we developed the measure gauges consisted of 5 elements,such as correlation,cluster center,and dispersion.Secondly,we designed the visual mapping method of each distribution attribute and analysed the correlation between the assessed value and actual value to support the gauge.Finally,we evaluated the cluster bundling presentation in several formats.Through several evaluation experiments,we verified the better way to mapping the data distribution feature in parallel coordinates,and the influence of dimension reordering and parameter selection in visualization.