To prevent spam users, you can only post on this forum after registration, which is by invitation. If you want to post on the forum, please send me a mail (h DOT m DOT w DOT verbeek AT tue DOT nl) and I'll send you an invitation in return for an account.

Markov trace clustering using event logs generated by myself

Hello everyone,
  I built a petri net using pipe and generate some event logs,then input them in the prom,using markov trace clustering plug in,but the event log is too simple to get a useful clustering dimensition.cause it only have two attributes as follows.
  If I want to clustering traces by trace attributes,for example,put case 1,3,5 in a cluster,case 2,4,6,7 in another cluster,how can I change my event log?add some attributes?
 Thank you in advance.


  • kfgz087kfgz087 Posts: 61
    I have asked my colleague to anwser your question. Please wait for a while.
    Guangming Li
    PhD student in Process Mining at Eindhoven University of Technology
  • TangYahuiTangYahui Posts: 59
    Thank you very much! I will be waiting for the response.
  • bhompesbhompes Posts: 10
    edited May 2018
    Dear Yahui,

    I understand you want to cluster the traces of an event log you created yourself. Without any data attributes, the only dimension that can be used to cluster upon is of course the control-flow dimension. In this tool, currently, the data attributes are primarily used, and no control-flow ordering is taken into account, only frequency or occurrence of activity names. If you want, you can create your own dimensions by extending the tool. If you don't want to extend the tool or to cluster on control-flow you have to add attributes at the case or event level.

    Note that the one involving clustering over time is not the one you should use for the purpose of your question. Also, you cannot set the number of clusters that you want to have output in the Markov clustering algorithm. The whole point of this algorithm is namely that it tells you how many clusters there are.

    Note also that I've already answered several of your and other people's threads on how to use the plug-in before, for example in:
    Perhaps a tool such as trace alignment by J.C. Bose is a better alternative?

    Kind regards,

    Post edited by bhompes on
    Bart Hompes - Eindhoven University of Technology
  • TangYahuiTangYahui Posts: 59
    Dear Bart, Thank you again for you response. I will add a attribute in my csv event log,for some cases set this attribute to "a",some set to "b",others set to"c" or something.And try your plug in again ,use this attribute as cluster dimension,if it works,I will tell you the result. I also know that for output a better cluster result,the number of cluster can not be set at first,but generate by the algorithm itself. Kind regards, Tangyahui
  • TangYahuiTangYahui Posts: 59
    Dear Bart,

     Though add an attribute in my csv file,and set different value on different traces,I got the clusters I want.

     Thank you again for your response,I know you are very busy.And your responses are very helpful for me.

    kind regards,

Sign In or Register to comment.