The Infosys Labs research blog tracks trends in technology with a focus on applied research in Information and Communication Technology (ICT)

« Can BPM support you in your Compliance Challenge? | Main | Distributed System Security Contd.. »

Process Mining - Existing Methods and Challenges

Process Mining has been gaining lots of attention from business analysts lately and is considered one big breakthrough in Business Process Management Paradigm. Process mining has been there as one research item for a long time, I wonder what has made it popular all of a sudden now. Is it because process mining has been presented as commercial tools and consultant’s aid by some companies? Academics are active in field of process mining for quite some time now and have created some useful open source tools as well.

Most of the process mining or process discovery tools provide lots of features of mining process which includes, mining of exception routes as well as most common routes. These tools provide an excellent start for purpose of process elicitation. The usual method of process elicitation is to conduct workshops and interviews with process owner and users. This method not only takes lots of time to elicit the process, the elicit process may be incorrect as it may be vision of people of how process should work rather than how process is actually working and lastly the elicited process may not have all exception paths. Other than that process mining tools can provide lots of information to process consultants just by looking at the mined process, this information includes the average time to complete the task, probability of taking one route as compared to other, identify potential process bottlenecks by identifying repetitive loops. These tools may also provide information on which role or user usually performs certain activities.

All this is good but lets analyze the input which is needed by these tools. Most of these tools expect input as lots of sequences of correlated activities as they executed in the system. The format would be like this, I would provide 3 sequences A -> B - > D, A -> C -> D, A -> B -> D. The output of this process may be A->(B or C)->D. This may also indicate that there is more probability of choosing path B after A is completed as compared to C.

The current challenge is how to get this activity sequences as input to the tool. One option is to create these activity sequences using audit logs. If the application stores the audit log of how an order or request is processes as input, these mining tools have capability of importing the available data as input to the tool. These activity sequences are available in ERP tools, workflow systems and of course in BPM tools. In my view all these kind of applications are already process aware. The real challenge is to extract processes from applications which are not process aware, for example from COBOL application which interacts with a web based application.

Is there a way to mine the process , even partially, from any set of existing applications?


I disagree that the activity sequences are available in ERP systems. The ERP vendors may describe what the workflows "should be" but this is not the necessarily how they "actually" are. Capturing audit trails from ERP systems can be difficult as the ERPs rarely use process ids or case numbers.

There are ways around this but it's not easy. I have just done this for BPCS. Mining processes is an art as well as a science. You have to look at different types of audit trails and this is often limited by the operating system as well as the code.

Nice article which explains the idea of process mining very clearly and briefly.

To answer your question at the end of the article:
For the open-source tool ProM (see the website) there exists a tool called the 'ProM Import Framework' which can convert source data to the MXML file structure. This MXML file can then be read by ProM to perform process mining. The ProM Import Framework works with plug-ins which enables you to write a custom converter to convert your data to the MXML file format.
For my graduation project (which starts at September 1st) I'm trying to create a more general tool which does not require any programming. Any input, requests or comments are greatly appreciated!!!

Good luck with process mining and data pre-processing!

I have been observing the most common method of process elicitation. This method is about conducting workshops, interviewing people, observing people at work to elicit the as-is process.
Please share your experience of using any method or tool for proces elicitation.

We use three types of sensors to collect data for process mining, namely ETL for systems, barcode/RFID for real world objects, or structured interviews for people. This info is transformed into a data cube for process analytics, of which one is process mining.

Thanks For Your valuable posting,this is for wonderful sharing,I would like to see more information from your side.

Great Article..It was very informative..I need more details from your side..include some tips..

Post a comment

(If you haven't left a comment here before, you may need to be approved by the site owner before your comment will appear. Until then, it won't appear on the entry. Thanks for waiting.)

Please key in the two words you see in the box to validate your identity as an authentic user and reduce spam.

Subscribe to this blog's feed

Follow us on