IT

supports the design, analysis, execution and monitoring of business processes based on explicit process models. E.g. Comunda, Flowable, Bizagi Studio

- List of tasks assigned to a person or role

- Management of the workflows of the workers

- A connector to another application running in a separate environment (Service Interface)

- Run-tim engine which controls the execution

User Inclusion

Decision Support

Service Integration (Delegate Java code or Connector REST / SOAP)

1) What happend?

2) What will happen?

3) Why did it happen?

4) What is the best that can happen?

Data Science and Process Science

1) Make it possible to reconstruct and analyse business processes based od digital traces in IT systems (EVENTLOGS)

2) Make it possible to implicit and otherwise hidden process knowledge contained in data and thus make it tangible and transportable

3) is data-driven and process-centric

Case ID

TImestamp

Activity

Attributes

Events

Instances

Discover a process model based on the event log, under the condition not be "overfitting" and "underfitting". 

(Underfitting: exclude so many case variants, that the process doens't fit to realisty)

(Overfitting: Everything is considerd form the eventlogs).

Existing process model is compared with an event log of the same process, aiming to indicate disagreement between the model and event log

reveal performance problems (e.g. unimely completion of a case, missed deadline, tardiness, as well as recurring quality problems) and annotate the process model with frequency and time information

Inout are multiple event logs (e.g. from the same process perfrrmed at different locations) and aims to identify execution gaps

Use of ML techniques to predict compliance and performance problems

Turn diagnostics of the current state of a process into improvements actions (e.g. case waiting time is too long -> assing additional resources, detect repetivive work which can be automated using Robotic Process Automation). 

Control-flow perspektive (order of acitivities)

Organizational perspektive (information resources hidden in the log, actors)

case perspective (properties of cases)

time perspective (timing and frequency)

decision perspective

Case ID

Activity

Timestamp

Correlation (events need to be related)

timestamps (need to be ordered per case)

snapshots (different lifetime)

scoping (which tables to incorporate)

granularity 

Play-In: Example behavior (event logs) is taken as input, and the goal is to construct the model (e.g. Alpha Algorithm using PETRI NET)

Play-Out: given a model, it is possible to generalte behavior (generated logs)

Replay: Uses an event log and a process model as input

Avg. Time customer spend waiting for a seat

avg. time to clean the plates

avg. time until food can be served

Path discovery in the process

As we can relate the time perspective to the resources

Supervised and Unsupervised. 

Data Mining results may be both descriptive and predictive. 

Decision trees, association rules, and regression functions say somethings about the data set used to learn the model. 

They can also be used to make predictions for new instances. 

Response Variables - dependent

Predicotr Variables - independent

Type of response variable: Cliassification (categorical variables9 and Regression (numerical)

Decision Tree

k-Means Custering

Clustering 

Pattern discovery