Abstract: In the light of misrecognition which caused by over-stretched and over-deviation of cluster curves during edge bundling in parallel coordinates plot, we proposed a force-directed skeleton-based bundling method（FDSBB） for Clustering data. Firstly, a optimized force-directed algorithm was used to form the layout of skeleton for each cluster. Secondly, the size of each cluster was taken into consideration to adjust the repel force between clusters, so the offsets of all clusters were transferred to the minority. Finally, we developed poly-skeleton to bundle the cluster in a band-form, which mapping the distribution attributes of clustering data（such as the value of the cluster center, correlation of clustering data, etc） more objectively. Experimental results show that FDSBB method could apparently improve the visual performance of clustering data in parallel coordinates.