Modeling "meeting events"

[Deleted User]

I am working on a problem in which the events are meetings. The meetings
happen at a given place, starting time, duration, and has two or more
attendees. The are certain periodic meetings (weekly, monthly, etc) with
the same participants, usually in the same location and time.

Being
new to this community, I am interested in any modeling recommendations
for this type of problem. I understand that there is no way to represent
geospatial information, so I assume that the meeting names need to
include the location.

The objective is to detect the recurring meetings, or the sequence of meetings of a given individual, etc.

Is this a typical process mining problem? Are there papers that would suggest ways to model this?

There is a lot of noise in the data (random meetings). Are there algorithms that are better or worse for this type of problem.

Thanks in advance for any suggestions.

JBuijs

Dear 'Tdarr2201',

I'm not sure I understand your problem correctly.
You talk about modelling, are you trying to model meetings in a process model?
Or do you have historic meeting data on which you want to apply process mining?

If you want to apply process mining, what is the data structure? What information is available?
I think that in your situation the determination of a 'case' is crucial. A meeting itself can not be a case since it then would have only 1 event.
Another approach would be an event log with only 1 case for which all meetings are events.
Do you have an idea for another case candidate?

When you have determined the case you should determine how you would name your events. Ideally you would specify multiple classifiers (See the XES standard definition) which allows you to change your notion of when 2 events should be considered equal.

Let me hear what you think and what your open issues are.

[Deleted User]

The meetings are the historical data from which I want to apply process mining. The data structure is something like data from a calendar program: the data includes the meetings, where the meeting was held, the starting time, duration and participants. There are regularly scheduled meetings, plus a lot of noise ("random meetings"). A better analogue is probably a record of meetings at public places such as coffee shops (some people hold regular get togethers, some just show up randomly for a coffee).

As I am new to this area, I think that there is one case and all the meetings are events in the case.

My open issues are probably pretty basic questions: what are the factors for determining a case? What are the factors for naming events? What is the purpose of a classifier?

Is the first step to review the XES standard definition? I am looking for more insight than syntactic descriptions of a standard and something more along the lines of modeling design principles / best practices, etc.

-Thanks-

JBuijs

Dear tdarr2201,

I think that there are two resources that you might find interesting:
1) indeed, the XES standard proposal, which can be found at http://www.xes-standard.org/_media/xes/xes_standard_proposal.pdf. This will provide you with a basic idea of what the XES standard stores and in which structure.
2) My master's thesis (PDF, 10 MB) is about extracting event logs from data bases. In my thesis, more specifically chapter 3, I explain what you should consider when determining your case.

I hope these documents help you further.
If you have any questions after reading these documents, please post them here!!

Anne

Hi Tim, 

Perhaps you find it useful to read our blog post on how to interpret the log data for process mining purposes. It answers questions such as 'What is a case?', 'What is an activity'? and so on. 

Here is the link: http://fluxicon.com/blog/2010/09/how-to-get-started/

Best, 

Anne