Eric Verbeek

Eric Verbeek

Eric is a scientific programmer at the Mathematics and Computer Science department. He is the custodian of the process mining framework ProM.

Activity

  • Hi,

    Thanks for letting us know. Indeed, the processmining.org site got overhauled a while back, as a result of which this link did not work anymore. I've replaced it now with a link (https://research.tue.nl/en/publications/pmsup2sup-a-process-mining-project-methodology) that does work. Note that following the link will bring you to a page where the broken...

  • The links have now been fixed.

    Thanks for mentioning it, the ProM website got transferred but we forgot to update the links here.

  • The PALIA suite is still available through the link https://paliasuite.webs.upv.es/indexEN.php?panel=paliasuite.spherademo. However, PALIA suite requires Internet Explorer (IE). You can configure Edge such that it will start the URL above in IE mode. In Edge, go to the Default Browser tab in the settings, select Allow for the IE mode, and add the URL to the...

  • If you can only find the Conformance Checking of DPN plugin, my guess is that you loaded the PNML file as a DataPetriNet using the "PNML Petri Net with Data files" importer. It would be better to load the PNML file using the "PNML Petri net files" importer.
    The regular replayers cannot handle a Data Petri net, they need a regular Petri net.

  • ProM Lite cannot import general Excel files (like .xls or .xlsx files). But if you export data from Excel to a CSV file, ProM lite can import it.

  • ProM Lite can open XES log files and also CSV log files, but an imported CSV log file first needs to be converted to an event log. For this conversion, the user needs to specify which columns contain the relevant attributes (like the case id, timestamp, etc.).

  • Hi,

    Indeed, this is what we typically encounter. The process owner typically has an idea of how the process is going, but this may not be exactly how the process is going. Using the data, we can show that in some cases the process deviates from what the process owner has in mind. Possibly, there may be good reasons for such deviations, but the process owner...

  • Hi,

    A trace attribute is a piece of data that is associated to a case rather than to a single event for that case (event attributes). A possible example for a trace attribute could be the name of the case owner, like the name of the patient that is being treated, or the name of the person that orders something.

    As an alternative, we could add a trace...

  • Hi,

    We assume that the data is exported from the HIS system to an event log file, preferably using the (compressed) XES file format. In this event log, we have a trace for every case in the system, where every trace corresponds to a sequence of events that occurred for that case. Both a trace and an event can have data attributes, like the name of the...

  • For some event logs, the other miners may just result in what we call spaghetti models. For such event logs, the fuzzy miner may produce a more sensible result.

  • Yes, the inductive miner allows for loops. However, it only allows for loops with a single point of entry and a single point of exit.

  • Dear Surya,

    Can you open a command prompt, go to the folder where you have installed ProM Lite (1.3, I assume), and then run ProM Lite by starting the ProMLite13.bat in that command prompt. This may provide us with more information on what's going on.

    The first (splash) screen shows for about 2 minutes. The idea is that ProM is up and running by then....

  • @PedroPinedoBorobio I only now understand that you are referring to the alignments shown around 8:26 in the video. Indeed, you can swap the log move (e,>>) with the model move (>>,f), yielding a different correct alignment with the same costs. In fact, this is often used to reduce the search space for alignments, as any of these similar alignments would...

  • Hello Pedro,

    Yes, the order matters. The order in the trace should not change: If we abstract from the >> symbols, it should always read abcdeg, as this is the trace (a sequence) that we need to align in the model. The order in the model is the sequence in which the transitions must fire.

    Your alignment is correct, but it is not optimal. The first...

  • Hello Pedro,

    You can download CPN IDE from https://cpnide.org, use that to create a Petri net, save it as a CPN file, and then import that file into ProM.

    CPN IDE can be used to create colored Petri nets, but as long as you do not specify any colors for the places or any expressions on the arcs, things should work out fine.

    I would be interested to...

  • Hi,

    The alpha miner will give you an accurate process model provided that a number of assumptions hold, including the following:
    - No noise in the event log.
    - If two activities can directly follow each other in the process, then in some trace they will indeed directly follow each other.
    - The process model should not contain constructs that mix...

  • ProM Lite contains a plugin called "Analyze with Woflan" that you can use to check whether a Petri net is (1) a workflow net and, if so, whether it is (2) sound. If it is not a sound workflow net, the plugin will report diagnostic information on the issues.

  • @PedroNavarro ProM Lite 1.3 will work to some extent on MacOS, but some plugins rely on a 32-bit library. As from Catalina (10.14), support for 32-bit libraries has been dropped in MacOS. Please be aware of this.
    Examples of plugins that will not work are the Hybrid ILP Miner and the Inductive Visual Miner.

  • Dear Abdullah,

    I'm assuming here that you are interested in the routes taken by the phone owners through the building, and want to see whether there are, say, congestions in the building. If you want to discover other things, it would be useful to know what you want to get out of the data. What do you want to learn from the data?

    As you are collecting...

  • Hello Abdullah,

    What would be your notion of a case? From your description, it seems that some people (wearing sensors) are walking around. Is every person then a case? How can I seen this in the log? For example, which sensors belong to the same case?

    When you know what your notion of case is, you can then think of what the activities are. The...

  • Dear Abdullah,

    Okay, so now you have a couple of activities. However, the importer will have problems with the walk activities, as it uses white spaces to separate fields. It would be better to enclose the activity names in quotes, like "walk staraight from point 4 to point 5", or just to replace all white spaces with _, like...

  • Dear Abdullah,

    What is the process to discover from this event log? The log only has one activity: stop. The Uid can serve as case ID, but if there is only a single activity, there is no real process to discover.

    Kind regards,
    Eric.

  • Hi Mirley,

    If you select the department attribute as the attribute to use while mining, then you will get a flow that shows how cases flow between departments.

    Some miners allow you to select any attribute as the attribute to use as the activity. For such miners, you can simply select the department activity and go. Other miners may require a classifier...

  • This may have something to do with the security settings of your browser. Some browser may simply refuse to download a document using HTTP. Instead, they require HTTPS. Please try the link below if all else fails:

    https://www.promtools.org/prom6/downloads/FutureLearn%20-%20Process%20mining%20with%20ProM%20-%20Event%20logs.zip

    Kind regards,
    Eric.

  • Hello @PriyaNaguine ,

    The transition 'd' has two input arcs: One from the place following 'a', and a second from the place following 'b'. As a result, both places need a token for 'd' to be enabled. However, this is impossible if we start with only one token in the initial (leftmost) place. Therefore, 'd' can never fire, hence it is dead.

    Note that if...

  • Dear @LimesS ,

    No, unfortunately not. We're still looking for a 64 bit version of the lpsolve library for MacOS.

    Any plugin that relies on lpsolve (like many replayers do) will fail on recent versions of MacOS. As per Catalina, MacOS has dropped support for 32 bit libraries, and on MacOS we only have a 32 bit library for lpsolve.

    Kind regards,
    Eric.

  • Hi,

    In abstract terms, the Event Class relates to the object in your model. If your model is a process model, then it either relates to an activity or to a step (like start or complete) of an activity. This allows you to position the event in your model. Note that the model could also be an organizational model, in which case the event class typically...

  • Dare @KevinMcKenzie ,

    If you need CSV files, you can go to the website of the Task Force on Process Mining, where they have a page containing sample event logs: https://www.tf-pm.org/resources/logs. This also contains CSV logs, just select the csv tag at the right and then select an event log. You will then be redirected to a download page for the CSV file...

  • Hi @EagleZhang ,

    Please check out Chapter 10 of the book "Process Mining: Data Science in Action" by Wil van der Aalst (ISBN 978-3-662-49851-4) for the refined framework.

    Kind regards, Eric

  • This depends a lot on the real event log at hand, and the real process that underlies this event log (that generated this event log).

    The inductive miner is a very good miner provided that some requirements hold, like:
    - The process needs to be nicely structured.
    - The process should not include multiple activities with the same label.
    If these...

  • Hi Nick,

    Please note that the visualization of the mining result is not a concern for the miner itself. The settings/filters you refer to would be the settings/filter for the visualizer, not for the miner (as it has run).

    Perhaps it may help to read the publications multi-perspective process mining from Felix Mannhardt, see...

  • @EagleZhang

    The 80% percentage usually works well. That's why we use it as a default value. But I agree that there may be cases where it may not work.

    Kind regards,
    Eric.

  • @MaxAlainBienvenuMBELLEMBELLE

    This may just have something to do with the fact that the links still uses the http protocol rather than the https protocol. Could you please use the link https://www.promtools.org/prom6/downloads/FutureLearn%20-%20Process%20mining%20with%20ProM%20-%20Event%20logs.zip instead and see whether that works for you?

    Kind...

  • Thanks for the info. That is more or less what I expected. The plugins that do not need that library will work, but the plugins that do need that library will fail. Unfortunately, some important plugins (like the replay plugins) do need this library :-(.

  • Hi,

    Are you using a version of Java that was already installed on the computer? Some Java versions have this exception, like 8u131.

    If possible, please use the version of ProM Lite 1.3 that comes with the 64-bit JRE8. This version (8u161) should not have this issue.

    Kind regards,
    Eric.

  • Dear Camaal,

    You mean you also got it working on MacOS? My guess is that ProM Lite 1.3 may start on MacOS, but some of the plugins will just not work as the 32-bit library cannot be loaded properly. As an example the Inductive Visual Miner may start and show the discovered model, but it will not show an animation as it fails to replay the log on the model....

  • Hi,

    I propose to do the following. Open the ProM.ini file with your favorite text editor, and find the line with PROM_USER_FOLDER. Should be around line 38. The change this line into the following:

    PROM_USER_FOLDER = .

    Note the dot at the end, that one's crucial. Then start ProM Lite 1.3, which will again install all packages again. If you then exit...

  • Dear Gilbert,

    ProM Lite 1.3 will not work properly on MacOS /unless/ the MacOS predates Catalina. As from Catalina, MacOS has dropped support for 32-bit libraries, and some of the relevant plugins rely on such a library on MacOS.

    If possible, please use a Windows machine to run ProM Lite 1.3, and download the installer that contains a 64-bit JRE8....

  • Hi,

    If you are using Windows, than it would be best to download the "ProM Lite 1.3 with 64-bit JRE8" from https://www.promtools.org/doku.php. This installer will install ProM lite 1.3 with a JRE8. This will prevent issues with the JRE, as we know it will work with this JRE.

    Kind regards,
    Eric.

  • Hi,

    Which OS are you using, and which version of ProM lite 1.3 did you download if your OS is Windows (with or without the JRE)?

    This could be a Java-related issue if you are using some JRE that was already installed on your machine. If so, you could try to download and use a different JRE.

    Kind regards,
    Eric.

  • Dear Camaal,

    Indeed, ProM will only work properly on a Mac if you are using a MacOS that predates Catalina. As from Catalina, MacOS has dropped support for 32-bit libraries, and on Mac we need a 32-bit library.

    On Windows, there should be no problem.

    Kind regards,
    Eric.

  • Hi,

    That's unfortunate. Can you start ProM Lite 1.3 using the .bat (or. sh) file, and then try again? I guess this will fail as well, but for me the texts written by ProM in the shell are of interest. In this text we may find an explanation what goes wrong.

    If you start ProM Lite 1.3 using the .exe file, these texts will not show, and will get lost....

  • Dear @PrashanthiRatnakumar ,

    If you started ProM Lite 1.3, go to the Action tab (middle tab), and then type "induc" in the search field, then three actions (plugins) should be shown:
    1. Mine Petri net with Inductive Miner,
    2. Mine process tree with Inductive Miner, and
    3. Mine with Inductive Visual Miner.
    If ProM only shows the last one, please make...

  • Hi,

    There are five colors involved, ranging from bad (red) to good (light yellow). Each colors spans 20% of the range, that is 20% between the minimal value and the maximal value. As a result, everything from 80% of the maximal value to the maximal value is colored red.

    Kind regards,
    Eric.