Process Mining: Data Science in Action

#ProcessMining #ML #DataMining #DataScience #InternetofEvents #Coursera #Certifications

At a certain public hospital in the US a Machine Learning algorithm was tasked to predict Pneumonia occurrences from the existing data-set of patients. The algorithm did everything right except for one critical mistake - it instructed doctors to send patients with asthma home even though such people are in a high-risk category. Because, in real life, the hospital automatically sent patients with asthma to intensive care, these people were rarely on the ‘required further care’ records in the data-set on which the Machine Learning system was trained. What a scary scenario!

Process Mining hopes to overlay the so-called ‘social system analysis’ over algorithms and help bring the algorithms in sync with the actual process flows such that such blind spots in AI can be addressed. This may also provide the levers for better ethical compliance.

"Process Mining" is one such specialized certification course offered on Coursera by The Eindhoven University of Technology, The Netherlands. To quote the course's chief instructor Prof. dr. ir. Wil van der Aalst:

"Process mining bridges the gap between traditional model-based process analysis (e.g., simulation and other business process management techniques) and data-centric analysis techniques such as machine learning and data mining. Process mining seeks the confrontation between observed behavior and process models."

In other words, Process Mining attempts to provide the right context to the data mining and machine learning exercise for a given business problem.

As illustrated in the image above, there is usually a gap between the 'ideal', documented, process and the actual path that is being practiced. Process Mining helps unearth such deviations. Typically, they are counterproductive and hence they need to be managed for their risk and costs. But Process Mining also identifies deviations that are 'positive'. An easy example will be the case of a doctor in the trauma care unit who deviates from the set processes in order to save the life of a patient in the event of an emergency. Traditional BPM has always struggled to manage such scenarios and reconcile with reality.

Offered since 2013, the course on Coursera is spread over 7 weeks with a intent to balance between mathematical / algorithms theory and practical use of process mining tools. The format has been revised in 2016 where the course moved to the On-demand format, and a final overall quiz exam has been introduced. Each week is graded with a quiz exam where the minimum required passing score to move to the next week is 80%. A timely completion of all the 7 graded modules and passing of the final quiz exam will get you a certificate like this.

Based on the premise of "Internet of Events" this specialization promises to tackle some of the major challenges of the future.