Process Mining Clustering Analysis with QPR ProcessAnalyzer 2019.3

QPR ProcessAnalyzer 2019.3 introduces lightning-fast, easy-to-use, built-in Machine Learning and Artificial Intelligence based Clustering Analysis. Clustering is an unsupervised machine learning technique for finding similarities within process mining cases, in order to better understand the differences from process behavior and case attribute point-of-view, as well as detecting problems in process execution and source data quality. Other major improvements in 2019.3 include Calculated Case and Event Attributes to easily augment the process mining data, calculate KPI variables and enable further business relevant analysis. In QPR ProcessAnalyzer 2019.3, we also introduce enhanced grouping for ChartView.

Clustering Analysis

Artificial Intelligence is here and QPR ProcessAnalyzer now assists you in understanding your process mining data! Usage of clustering analysis is super simple - default configuration immediately groups process mining cases into five clusters and shows the five most important characteristics for each cluster! This helps the analyst to find similar cases within a large process mining model and thus easily separate the "apples and oranges" from each other. Clustering analysis utilizes the full case history (ie. the event types visited) as well as business data (ie. the case attributes) in a unique combined way to identify similarities within cases. Check out the example below:



In the above screenshot, we see clustering results for Order-to-Cash model showing five clusters. Cluster 1 contains 24% of all cases and it includes sales orders for Hats. Cluster 2 contains Shirts purchased for Kids. Cluster 3 is sales orders from Women whereas Cluster 4 contains Patricia's sales to Men. Most important characteristic in cases included in Cluster 5 is that they visit the Event Type Purchase Order Created, which means that the order was delivered by supplier and not from company's own stock. All this information is retrieved in a few seconds for any process mining model without setting extra parameters in QPR ProcessAnalyzer's Clustering Analysis!

... and even though this is the introduction of this new Clustering Analysis we already have nice extra goodies. You can super easily re-run the Clustering for any subset that you are interested in more detail. For example, by simply selecting an Event Type from flowchart we get clustering for cases visiting Customer Pick-Up as shown below:




Picture above shows the Clustering Analysis results for the cases that were at least once picked by customer. Largest cluster containing 35% of cases is Women Shoes, 2nd is Men in New York, 3rd Shirts that for some reason have been also sent as a shipment (likely hint that the customer pick-up is not always working so well), 4th cluster contains Mary's deals to Women and last cluster has Returned products which al seem to be for customer group Kids.

QPR Process Mining Clustering Analysis takes process mining to a new level harnessing the power of machine learning in a super easy-to-use and fast format for all analysts and business people. My recommendation is to load your own data and see the results yourself - it only takes a few seconds in QPR ProcessAnalyzer Cloud!

Calculated Case and Event Attributes

All analysts know that when you give a very good answer you typically get a new follow-up question. In process mining this means that once we have built the process mining model and let users to explore the contents we start receiving questions about various KPI values, Conformance checking results and other categorizations of cases. QPR ProcessAnalyzer has always supported ad-hoc analysis as well as ready-made dashboards and KPIs. With 2019.3 we introduce a new easy-to-use way of adding case and event attributes that contain calculated values.

Here is my personal top-10 favorites list of calculated case and event attributes:

1. Calculate the conformance for each process mining case

2. Calculate the effective working hours based on a calendar or each process mining event. Consolidate the results to a case level attribute

3. Currency conversions

4. Calculating re-work

5. On-Time-In-Full - calculated KPI value based on Confirmed Delivery Date and Actual Delivery Dates

6. Maverick Buying - calculated KPI value based on Purchase Order must be created before Invoice Received.

7. Maverick Payment - calculated KPI value based four eyes principle 

8. Source Data quality checks

9. Categorization of case attribute values

10. Finding the starting and ending month for each case

ChartView - Enhanced grouping and 'Show Others'

Process Mining with QPR ProcessAnalyzer deals often very large datasets. QPR Chart View is an excellent tool for creating to-the-point visualizations with a few clicks. One import feature for visualization is the ability to easily show the most important values together with other values grouped and shoed as 'Others'. Now in QPR 2019.3 this is made super easy - check out below:




Above picture shows top five Suppliers WITHOUT showing the other Suppliers at all....




...and the picture above shows top five Suppliers with the 4441 cases from other 10 suppliers shown as the 6th category.


I hope you enjoy QPR ProcessAnalyzer as much as I do and really look forward to hear your comments about Clustering Analysis - it may really open your mind for teaming up with artificial intelligence!!

Best regards,


Written by
Author imageexpand

Teemu Lehto

Doctor of Science (Technology) and Process Mining evangelist active in marketing, sales, consulting, product development and research. Teemu has been involved in 200+ end customer process mining project from order-to-cash, purchase-to-pay, plant maintenance, auditing and service. Teemu is also an active speaker delivering the process mining message as well as writer for several process mining and machine learning scientific articles. Book a meeting with Teemu using the link:

Share Online