| Association rules mining is powerful in information processing, which can extract valuable rules from large databases and many automated information systems. However, most of data mining systems support association rules but express them inefficiently, so it isn't easy for the user to extract meaningful rules. The application of visualization techniques in association rules makes data relationship displayed distinctly. According to visualization graphics the user may more expediently participate in the mining process of association rule to realize strategy.Nowadays, the visualization of association rules exists many of common problems such as screen clutter and occlusion. In this paper, a new technique for visualizing association rules based on Java 3D three-dimension coordinates called AR3DV is presented. It adopts X axis to display rules and Z axis display items. At the same time, it adopts Y axis to display support and confidence. The display of AR3DV is realized via the program idea and the program step of Java 3D. AR3DV not only is able to display lots of any-to-many and many dimensions association rules but also overcome problems of screen clutter and occlusion.Furthermore, a visual primitive system of association rules called ARVMiner is realized. The system makes the AR3DV techniques combine with association rules mining based on restriction, which allows the user to view only a particular subset of association rules that contain selected items of interest and display association rules orderly. In this system, firstly the user can choose databases and algorithms. Secondly the user can test and compare the performance of algorithms. Thirdly the user can set the parameter of mining. Finally the process and the result of mining are visual. The system realizes the visualization of mining result and mining process at the same time. End of the paper, different mining examples are used to validate the perfectability and the practicability of primitive system. |
PDF Full Text Request