The main visualization challenge of the project is to display multiple process mining results at the same time, in an integrated way. The size of the physical screen is the main limiting factor when it comes to displaying multiple windows. Therefore, we apply several solutions to cope with this issue. First of all, we create a new frame, detachable from the main frame, and use it to place all process mining results. Thus, should two screens be available, the table of visualization can be placed on one screen, while the plugin results can be displayed on the second screen. On this new frame, windows are organized next to each other, in an easy-to-identify way. Even though such a frame layout is already enough for the visualization of the plugin results, we decided to make some changes as it was lacking the desired flexibility. Hence, replacing the windows with dockable ones to allow moving them around is one of the most important visualization features that is supported in the project. A large part of the dockable functionality is taken from the DockingFrames 1.1.2 2 and adjusted for the project needs.
In the following, we explain the framework of the windows, with details related to the layout of the windows frame. Then, we give a list with the frame functionality items. Finally, we show the result visualization obtained using the PROCUBEplugin.
Figure 5.8: Dockables functionality. Panels are wrapped into dockables. Dockables are put onto stations which lay on the main-frame. As such, dockables can be moved to different stations. Figure 5.8, taken from [47], shows the
framework based on which dockable windows are created. Dockables are not stand-alone windows. They require the support of a main window (the Main-Frame). The main window is most of the times a JFrame. As long as this frame is visible, so are the rest of the compo- nents on it. In case of non-dockable panels, they are just directly connected to the main frame. Consequently, the main frame can con- sist of several panels, with different data dis- played on them. To support floating panels, however, an additional layer is added between the panels and the main frame. The compo- nents of this layer are the so-called Stations. Among their purposes is also to allow the user to drag & drop panels and to minimize or max- imize windows. A central controller is used to wire all the objects of the framework together.
It manages the way elements look and their position in the frame and it monitors all the occurring changes within windows. Further, each panel is wrapped into a dockable. Dockables are the final components and they are the ones that actually offer the floating behaviour.
To display dockables in a certain layout, a Grid component is used. The matrix of the grid gives an organized way of displaying windows in the screen. For our project, the matrix of the grid component corresponds with the matrix of the table of visualization. That is, the plugin results for different cells are shown in the same order with the one used to display the cells in the visualization table.
In the view of the above approach, the following visualization capabilities are supported: • Default layout with all the dockables normalized. Normalized dockables are placed on the
main visualization frame, in the way cells are displayed in the visualization table.
• Dockables can be maximized. A maximized dockable takes all the space it can, most of the time, by covering other dockables.
• Dockables can be minimized. Minimized dockables are not visible right away. They can be restored to a normal state, by pressing again the minimization button.
• Dockables can be extended. Once extended, dockables have their own window, independent of the main visualization frame. This functionality is very useful as it allows, for example, moving windows with plugin results on different screens.
• By the drag & drop operation, dockables can be placed on any part of the screen. For example, by dragging one dockable on the place of another one, these two are swapped with each other.
• When multiple plugin results are available for the same visualization cell, each result win- dow is a new tab in a tabbed pane. That makes it easy to quickly identify plugin results corresponding to the same visualization cell.
• Unnecessary windows can always be closed.
Figure 5.9: Visualization of plugin results in the PROCUBEtool. Each plugin result is displayed in a dockable window and can be part of a tabbed pane.
Figure 5.9 shows several windows with plugin results. Two Log Dialogs, a Fuzzy Miner, two Heuristics Miners and a Social Network form the visualization results. Multiple tabs can be distinguished since multiple plugin results exist for the same visualization cell. All the windows are dockable. After undocking a window, the rest of the windows are automatically rearranged in the screen.
Chapter 6
Case Study and Benchmarking
In the previous chapter, the implementation of the process cube was described as a combination of external technology (Palo, MySQL, ProM) and newly-introduced process-cube-related features. Further, we continue with an evaluation of the functionality of different event logs and an assess- ment of the PROCUBEsystem performance. The results presented in the chapter are based on the event data of an artificial digital photo copier event log and on event data of a Dutch municipality event log.