search

bupaverse / processanimateR

Token replay animation for process maps created with processmapR by using SVG animations (SMIL) and the htmlwidget package.
https://bupaverse.github.io/processanimateR/
Other
66 stars 11 forks source link

Support heuristicsmineR Causal nets #24

Open fmannhardt opened 4 years ago

fmannhardt commented 4 years ago

processanimateR already supports interpreting the DiagrammeR object returned by heuristicsmineR::render_causal_net; however, no animation is rendered.

To support causal nets we would need to change the lifecycle of tokens from being case-based to being based on individual edges, as multiple tokens may travel for the same case.

mkrasmus commented 4 years ago

Thanks @fmannhardt,

I'm only just beginning to learn about process mining/maps in general, and reading now a bit of Dr Weijters work re Flexible Heuristics Miner. I will take a look at heuristicsmineR but was hoping you might let me know of where to look (e.g., PM4PY) if the following use case does not apply:

I have an eventlog showing the intake, decision activities and outputs of various 'cases'. The edges of my created process map aptly demonstrates the pathways from intake, through activities, to the outputs. The edges are thickened as per relative frequencies, and performance is shown in median days. I am happy with this overview of the system, but I would like to present to others a per-case predicted pathway. Given a set of Xi ... Xm predictors, could I then present a per-case predicted pathway as represented by percentage chance estimates for each pathway, edges thickened appropriately?

Ideally, users will be choosing cases within a Shiny environment and process maps will be rendered on a per-case basis.

Does this fit heuristicsmineR or anything else?

Thanks again Michael

mkrasmus commented 4 years ago

Hi @fmannhardt

I think I've kind of figured this out.. filtering on predictors may help with creating process maps per case. Probably overly crude method but will try this out. Any other suggestions though would be much appreciated.