Process Mining II

25 important questions on Process Mining II

What is conformance checking?

Comparing behaviour of the model with the behaviour of the log.

What are local diagnostics?

Pinpoint in the log or the model where it is incorrect.

What are global conformance measures?

Scores how well your model fits with your log. This is between the event log and the model.
  • Higher grades + faster learning
  • Never study anything twice
  • 100% sure, 100% understanding
Discover Study Smart

What has process mining to do with audits?

Process mining can help to detect fraud, malpractice, risks, and inefficiencies.

Can events in a business process also be evaluated while the process is running?

Yes

What happens if A fires? (there are two output places)

You consume one token, two tokens are produced in the output places p1 and p2.

What if the model can generate the trace?

Than the fitness of the model is good.

What is the name for fitness in conformance checking?

Token-based fitness (f)

How do you calculate the token-based fitness (f)?

First replay the traces in the discovered model.
There could occur a fitness problem when there are one or more missing tokens during replay (mi), or tokens are remaining after replay (rj). Then you can calculate the ftrace (= 0.5 ( 1- (m/c)) + 0.5 (1 - (r/p))).

We do not have to remember the formula.

What is the name for precision in conformance checking?

Advanced Behavioural Appropriateness (a'b).

What is the idea of advanced behavioural aprropriateness?

Compare the behaviour in the model with the behaviour in the event log.

You check for all possibilities which relationship it has and put it in a table. Then you do a calculation.

What two types of relationships does advanced behavioural appropriateness compare? Explain them.

- Follows relationship: Always follows, Never follows, Sometimes follows
- Precedes relationship: Always preceders, Never precedes, Sometimes precedes

What is artificial negative event-based conformance checking?

The availability of negative events (activities that were denied to happen) can measure traditional evaluation mesures (like recall, precision, F-measure). However, the availability of negative events in real-life logs is unlikely. Therefore put them in event logs.

How does artificial negative event-based conformance checking work?

At each positive event, it is checked whether another event appearing in the log could also occur, given its prefix. Therefore, it is tested whether in the log there exists a similar prefix sequence with this other event occurring at the same position of the positive event. When this is not the case, a negative event can be injected into the log at the position of the positive event

Then there will be a table with traces and it shows the trace and below each letter in the trace the other possible letters are mentioned too.

Do there also exist methods for measuring the generalization and simplicity?

There exist generalization methods. A simplicity metric are: look at duplicates, the number of events.

How do you go from a process to attribute names like activities, and times?

A process consists of cases. A case consists of events (one event relates to one case). Events can have attributes. Attributes are activities, time, cost, resources, etc..

Note: cases can also have attributes.

What is the difference between conformance checking and enhancement?

In enhancement we aim to produce a new model.

What are the enchancement techniques?

Model repair and model extension
There are two types of extension techniques: decision mining and organizational mining.

What is decision mining?

It aims to find rules that explain routing choices in terms of the characteristics of a case. Decision points are XOR-splits. To do this you can apply classification techniques (like decision trees). This can provide additional insights about what factors affect the execution of a business process.

What is the resource-activity matrix? And how do you make it?

A matrix used in organizational mining. It gives the mean number of times a resource performs an activity per case.

You have a table with on the columns case id's, activities, types, time, and resources. Then for each activity the mean number of times it is performed by a certain resource is added to a new table (the matrix).

What is the handover of work matrix?

It is a table with the number of times work is handed over from one resource to another.

What are requirements for event logs?

Events need to be referred to an activity name and a process instance
Timestamps, resource information, additional data elements, case ID's are also needed
Sometimes scoping and correlation (this can be hard)

There are different event types and different activity instances. What means it when they overlap?

Examples form activity instances:
- Schedule-assign-start-complete
- Schedule-assign-reassign-start-suspend-resume-complete
- Start-suspend-resume-suspend-abort activity
- Start-complete
- Complete

The last one happens the most.    

Note: there can be overlapping activity instances in event logs, then you do not know which start and end point is related to each other.

Give some examples of data quality issues.

Missing case IDs. Imprecision timestamps. Granularity of events. Missing events. etc.

Mention some process mining tools.

ProM, Disco, Perceptive Process Mining, Celonis Discovery, etc.

The question on the page originate from the summary of the following study material:

  • A unique study and practice tool
  • Never study anything twice again
  • Get the grades you hope for
  • 100% sure, 100% understanding
Remember faster, study better. Scientifically proven.
Trustpilot Logo