Automated process improvement uses process mining techniques to recommend optimizations to business processes. It can suggest changes to tasks, control flow, decisions, and resource allocation based on event log analysis. Process mining discovers predictive models and simulates the effects of different changes to identify sets of improvements that optimize given performance metrics. Key challenges include scaling to real processes, estimating impacts on multiple metrics, and usability of change recommendations.

1. Automated Process
Improvement
Status, Challenges & Perspectives
Marlon Dumas
Professor of Information Systems @ University of Tartu
Co-founder @ Apromore
Keynote, BPMDS + EMMSAD’2020, 8 June 2020

2. Process Mining 1.0
/
event log
discovered
process model
Automated Process
Discovery
Conformance
Checking
Variant Analysis
Differences
& diagnostics
Performance
Mining
business rules / process model
Enhanced
process model
event log’
Dumas et al. Fundamentals of Business Process
Management, 2nd edition, Springer 2018

3. Prescriptive Analytics
Predictive Analytics
Diagnostic Analytics
Descriptive Analytics
The Evolution of Process Mining
3
Process Mining 1.0
Automated Process Discovery
& Analysis
Process Mining 2.0
Predictive Process Monitoring
Automated Process Improvement

4. Operational
Level
Predictive Process
Monitoring
Predicting future states, outcomes,
or properties of a process instance
or group of process instances
Prescriptive process
monitoring
Recommending actions on the
basis of predictions to maximize a
performance indicator
Tactical
Level
Automated Process
Improvement
Robotic Process Mining
Search-Based Process Optimization
Process Mining 2.0
4

5. Predictive Process Monitoring
• What is the next activity for this case?
• When is this next activity going to take place?
• How long is this case still going to take until it is finished?
• What is the outcome of this case?
• Is the compensation going to be paid? Or rejected?
5

6. Predictive Process
Monitoring
Event
stream
Predictive
models
Detailed predictive dashboard
Alarm-based prescriptive dashboard
Aggregate predictive dashboards
Event
logDatabase
Enterprise
System
6

7. Event log
Classifier /
Regressor
/
Outcome predictionAttributes
Traces
7
Event log
Structured
predictor
Next activity /
Future path prediction
Attributes
Traces
Performance measure
prediction
Predictive Process Monitoring:
General Approach

8. Predictive Process Monitoring Approaches
Teinemaa et al. Outcome-Oriented Predictive Process Monitoring: Review and Benchmark. TKDD 13(2):17:1-17:57, 2019.
Verenich et al. Survey and Cross-benchmark Comparison of Remaining Time Prediction Methods in Business Process Monitoring.
TIST 10(4), 2019.
Tax et al. An Interdisciplinary Comparison of Sequence Modeling Methods for Next-Element Prediction. Software and Systems
Modeling, 2020, to appear.

9. Explaining
predictions
Helping users
understand the
causes of predicted
outcomes
Turning
predictions
into actions
Prescriptive
process monitoring
Challenges in Predictive Process Monitoring
10

10. Prescriptive process monitoring
Event log
(completed
traces)
Predictive
model(s)
Running
trace
Appl
y
P( )
Prediction Alarm/
no alarmPolicy
+/- -
- +
Cost model
Teinemaa et al. “Alarm-Based Prescriptive Process Monitoring”. Proceedings of BPM Forum’2018

11. Search-Based
Process Optimizer
Domain Knowledge
IoT, Web& social sensing
streams
Automated Process Improvement
EnterpriseSystem

12. Example: Improvement Opportunities
1
Officer
Clerk
Clerk Officer
Officer
Clerk
Skip credit history
check when customer has
previous loans with bank
Allocate an additional clerk
on Monday-Tuesdays, one
less officer on Fridays
This task can be
automated with an RPA
script
For consumer loans,
check credit history
before income
If loan-to-annual-
income ratio > 1.5,
allocate a senior officer
If credit rating is C or D,
do not wait for appeal

13. Given
• one or more event logs recording the
execution of one or more processes
• one or more performance measures that
we seek to maximize/minimize
• a process model, decision rules and
resource allocation rules
• a set of allowed changes to the process
model and associated rules
Find
• One or all set(s) of Pareto-optimal
changes to the process model and rules.
Automated Process Improvement

14. Task
• Automate individual tasks or groups of tasks
• Recommend best practices for task execution
Control-flow
• Task elimination/addition
• Task merging/splitting
• Task re-ordering, parallelization
Decision (data)
• Add / delete decision points
• Refine / enhance decision rules
Resource
• Re-allocate resources
• Refine / enhance resource allocation policies
Automated Process Improvement
Types of Changes

15. Automated Process Improvement
16
16
Execution data
Executable routine
specifications
Robotic Process
Mining
(Task Automation)
Decision Rule
Optimization
Flow Optimization
Optimized process
model
Resource
Optimization
Decision rules
Optimized resource
allocation policies
Optimized decision
rules

16. 31
Automatable Task
Example

17. Starting Point: UI log
18
V. Leno, A. Polyvyanyy, M. La Rosa, M. Dumas and F. Maria Maggi. Action logger: Enabling process mining
for robotic process automation. In Proceedings of Demonstration Track at BPM 2019, 124–128, 2019

18. Robotic Process Mining:
Synthesis of RPA Scripts for Task Automation
19

19. [copy to clipboard]
A task is automatable if every step in the
task can be deterministically executed
based on input data, or data produced by
previous actions
[select cell C1]
[select cell C2]
[edit cell C2]
20
Automatable Task

20. 32
Synthesis of RPA Scripts as a
Transformation Problem
V. Leno et al. Automated Discovery of Data Transformations for Robotic Process Automation.
In AAAI-20 Workshop on Intelligent Process Automation, New York, USA, January 2020

21. 322
Synthesis of RPA Scripts for Task Automation:
Approach
V. Leno et al. Automated Discovery of Data Transformations for Robotic Process Automation.
In AAAI-20 Workshop on Intelligent Process Automation, New York, USA, January 2020

22. 323
Preprocessing
Filter out redundant actions
 Control-flow redundancy (e.g. double copying without pasting)
 Data-flow redundancy (e.g. double editing of text field with replacement)
Identify candidate routines (repetitive sequences of actions)
 Sequential pattern mining  candidate routines
 Clustering similar or related routines

23. 324
Extracting examples from candidate routines
For each candidate routine trace:
 Collect the values of all read cells/fields (Inputs)
 Collect the latest values of all modified cells/fields (Outputs)
 Create input-output transformation example (Inputs, Outputs)
Inputs = [“Albert”, “Rauf”,
“11/04/1986”, “+61 043 512
4834”, “arauf@gmail.com”,
“Germany”, “99 Beacon Rd,
Port Melbourne, VIC 3207,
Australia”]
Outputs = [“Albert Rauf”, “11-04-
1986”, “Germany”, “043-512-
4834”, “arauf@gmail.com”, “99
Beacon Rd”, “Port Melbourne”,
“VIC”, “3207”, “Australia”]

24. 325
Transformation discovery
FOOFAH – transformation discovery by example
 Program synthesis as a search problem in a state space graph
 Heuristic search approach based on A* algorithm
 Cost function is the amount of manipulations
 Deals with string and table manipulations
+61 039 689 9324
+61 035 341 2938
+61 079 149 3015
+61 039 689 9324
+61 035 341 2938
+61 079 149 3015
039 689 9324
035 341 2938
079 149 3015
+61 039 689 9324
+61 035 341 2938
+61 079 149 3015
039 689 9324
035 341 2938
079 149 3015
039 689 9324
035 341 2938
079 149 3015
split_first(0, ‘
‘)
split(0, ‘ ‘)
drop(0, ‘ ‘)
drop(0, ‘ ‘) join(0, ‘ ‘)
join(0, ‘ ‘)
Input Output

25. 326
Transformation discovery
FOOFAH – transformation discovery by example
 Program synthesis as a search problem in state space graph
 Heuristic search approach based on A* algorithm
 Cost function is an amount of manipulations
 Deals with string and table manipulations
+61 039 689 9324
+61 035 341 2938
+61 079 149 3015
+61 039 689
9324
+61 035 341
2938
+61 079 149
3015
039 689 9324
035 341 2938
079 149 3015
split_first(0, ‘ ‘)
split(0, ‘ ‘)
drop(0, ‘ ‘)
drop(0, ‘
‘)
join(0, ‘ ‘) join(0, ‘ ‘)
Input Output
+61 039 689 9324
+61 035 341 2938
+61 079 149 3015
039 689 9324
035 341 2938
079 149 3015
039 689 9324
035 341 2938
079 149 3015

26. 327
Foofah-Based Transformation Synthesis:
Limitations
Does not scale up to real examples when applied directly
Discovers complex transformations (low readability)
Optimizations
 Synthesize one transformation per output field and use UI log to discover input-to-output
data flows
 Discover patterns in the input values and discover one transformation per input pattern

27. Robotic Process Mining
28
 Extracting candidate routines from noisy UI logs
 Handling heterogeneous occurrences of candidate routines
• Many ways of performing the same routine
 Discover transformations where the output fields are not (always) derived from fields
that are explicitly accessed (e.g. using screenshot processing and/or eye tracking?)
 Discover conditional transformations
• The transformation steps to be performed depend on conditions in the inputs
 Handling complex data types, e.g. copying a purchase order consisting of multiple line items
Open Challenges

28. Automated Process Improvement
29
29
Execution data
Executable routine
specifications
Robotic Process
Mining
(Task Automation)
Decision Rule
Optimization
Flow Optimization
Optimized process
model
Resource
Optimization
Decision rules
Optimized resource
allocation policies
Optimized decision
rules

29. Search-Based Process Optimization
Discover
Process
Model
Metaheuristics
Optimizer
(e.g. Genetic,
Hill Climbing)
Candidate
Changeset
Evaluator
Candidate
Changeset
Generator
New Pareto
front
Event log
Candidate
Change-
sets
Discover
Simulation
Model
Simulation
Model
As-Is Process
Model
Current
Pareto front
Business
Process
Simulator
Allowed
Changes

30. 331
Automated Discovery of Simulation Models
from Event Logs
Camargo et al. Automated Discovery of Simulation Models for Event Logs, Decision Support Systems, to appear, 2020
https://github.com/AdaptiveBProcess/Simod

31. • Exploring search spaces of process changes is challenging due to combinatorial
explosion and the fact that process changes may have non-additive effects
Scalability
• Simulation allows us to estimate changes in time and cost metrics, but what about
quality metrics (e.g. defect rates)?
Estimating the Effect of Changes
• How can analysts conveniently capture the allowed space of changes and their
associated costs?
• How to help analysts to navigate through the discovered change-sets and their trade-
offs?
• How can we make the analyst trust the change recommendations made by an
automated system?
Usability
Search-Based Process Optimization: Challenges
32

32. https://sep.cs.ut.ee/Main/PIX
The Process Improvement Explorer (PIX)