Why does Alpha algorithm return better fitness than Heuristic and Genetic Miner ?

Shital

Hi,

I trying to work on the first set of http://data.3tu.nl/repository/uuid:bd8fcc48-5bf3-480e-8775-d79d6c700e90

Using ProM 6.4.1, I created the Process models for Alpha, Heuristic and Genetic. Why is it that the fitness of Alpha is better than the other two. Please find my findings below.
Regards,
Shital

JBuijs

Dear Shital,

The Alpha miner is 'lucky' in this example. See also the paper we wrote that uses this event log
https://scholar.google.nl/citations?view_op=view_citation&hl=en&user=hQjcD-0AAAAJ&citation_for_view=hQjcD-0AAAAJ:Tyk-4Ss8FVUC
or, if you really want, my PhD thesis
http://repository.tue.nl/780920 (Chapter 6 if I recall correctly)

In general the Alpha miner is not robust to incompleteness and/or noise. The other algorithms try to handle these, but as you have seen have other issues. Note that the perfect discovery algorithm does not exist (yet)

Shital

Thanks Joos