Custom visualization metrics

ruudvancruchten

Hi all,

For my msc graduation research I'm going to develop a custom visualization/metrics to display on a discovered process (like frequencies and times). The goal is to display the value of the cases on a path/flow. I'm thinking about editing the inductive visual miner or the fuzzy miner for this. What do you think is the best way to approach this, and what miner do you prefer?
Any tips are welcome!
  

JBuijs

My suggested approach would be to create a new visualizer for the Inductive Miner 'object'/result, i.e. your visualizer would be listed in the top right drop-down list.
This way you can develop your plug-in in a separate package, and users can choose which visualizer they prefer.

sjjleemans

(I cannot help with adapting Fuzzy Miner, but I'm the author of the Inductive visual Miner, so I'll sketch what has to be edited)

If the measures you have in mind require alignments, then I think the IvM makes sense, as it already computes alignments.

• first, find out what needs to be computed for your measure. Alignments, animation, performance? In a recent nightly build, press ctrl+c to see the structure of IvM and find out where your computation would fit.
• implement your measure in a ChainLink<A, B>, where A is the input and B is the output of your measure. There are plenty of examples in org.processmining.plugins.inductiveVisualMiner.chain
  
• extend InductiveVisualMinerController, add your computation, and reroute connections to your prerequisites. The controller will automatically call your computation asynchronously when its prerequisites are ready, and handle cancellations.
  
• also in InductiveVisualMinerController, in ChainLink.setOnComplete(), update the visualisation of the model with your computed measures. For an example, please have a look at Cl12FilterNodeSelection.

Hope this helps. If you go for the IvM and need to know more, please let me know!