Consider the following event log. Write the corresponding work-
flow log using the same approach as in Figure 11.7 (page 427). You may use
abbreviations CF, SB, SR, and CC to denote the four tasks in this process.
Draw the dependency graph of the workflow log obtained in
With reference to the event log of the Business Process Intelligence
Challenge 2017 (https://tinyurl.com/bpic2017), does it sometimes happen that a
Case ID Task Label Timestamp
1 Create Fine 19-04-2017 14:00:00
2 Create Fine 19-04-2017 15:00:00
1 Send Bill 19-04-2017 15:05:00
2 Send Bill 19-04-2017 15:07:00
3 Create Fine 20-04-2017 10:00:00
3 Send Bill 20-04-2017 14:00:00
4 Create Fine 21-04-2017 11:00:00
4 Send Bill 21-04-2017 11:10:00
1 Process Payment 24-04-2017 14:30:00
1 Close Case 24-04-2017 14:32:00
2 Send Reminder 19-04-2017 10:00:00
3 Send Reminder 20-05-2017 10:00:00
2 Process Payment 22-05-2017 09:05:00
2 Close Case 22-05-2017 09:06:00
4 Send Reminder 21-05-2017 15:10:00
4 Send Reminder 21-05-2017 17:10:00
4 Process Payment 26-05-2017 14:30:00
4 Close Case 26-05-2017 14:31:00
3 Send Reminder 20-06-2017 10:00:00
3 Send Reminder 20-07-2017 10:00:00
3 Process Payment 25-07-2017 14:00:00
3 Close Case 25-07-2017 14:01:00
loan offer is canceled (meaning that event “O_Cancelled” occurs), but later the
offer is accepted (i.e., “O_Accepted”)? If yes, in what percentage of cases does
this happen?
Consider the workflow log [〈a, b, c, d, e〉, 〈a, b, d, c, e〉, 〈a, c, d,
b, e〉, 〈a, d, c, b, e〉, 〈a, d, b, c, e〉, 〈a, d, c, b, e〉]. Show step-by-step how the α–
algorithm works on this workflow log, and draw the resulting process model.
Consider the log in Exercise 11.20 (page 470). Show step-by-step
how the α-algorithm works on this log, and draw the resulting process model.
Draw a model in BPMN that can produce every possible execution
sequence that includes tasks {a, b, c, d, e}. Discuss the fitness and precision of
this model with respect to the workflow log {〈a, b, c, d, e〉, 〈a, c, b, d, e〉}.
Draw a process model that would have perfect fitness and perfect precision with respectto this workflow log (i.e., a model that can produce exactly this log). Would the
α-algorithm produce this model with perfect fitness and precision?
Using a process mining tool, discover a BPMN process model
from the following event log of an authentication process: http://tinyurl.com/
simpleEventLog (also available in the book’s companion website).
Note: This log is sufficiently simple that it is possible to manually derive the BPMN
process model from a dependency graph.
Using a process mining tool, discover a BPMN process model from the following event log of a telephone repair process: http://tinyurl.com/repairLogs
(also available in the book’s companion website).
Note: This log is more complex and difficult to manually understand using a
dependency graph. Consider using a process mining tool that can discover BPMN
process models (e.g., ProM or Apromore).
Consider there is an AND-split in a process model with two subsequent tasks a and b. What kind of pattern do these tasks show on the timeline chart?
Consider the workflow log that you created for Exercise 11.5 (page 427) from the cases shown in Figure 11.4. Replay these logs in the process model of Figure 11.21 (page 456).
Note down consumed, produced, missing and remaining tokens for each arc, and calculate the fitness measure. Assume that tasks not shown in the process model do not change the distribution of tokens in the replay.
Do the models you discovered in Exercises 11.26 and 11.27 perfectly fit the corresponding event log? If not, describe to what extent and howthe event log differs with respect to the discovered model.
Consider again the health insurance process examined in Exer-cise 11.19 on page 460 (available at http://tinyurl.com/InsuranceLogs) and specif-ically the log containing the 2011 cases (log L1). Based on this log, complete the following tasks using a process mining tool:
1. Extract from the log all the cases containing at least one occurrence of high
insurance check. The filtered log will be called FilteredHigh.
2. Extract from the log all the cases containing at least one occurrence of low
insurance check. The filtered low will be called FilteredLow.
3. Compare the mean cycle time of cases of FilteredHigh and FilteredLow.
4. Describe the main differences between FilteredHigh and FilteredLow in terms of
the frequency and relative order of tasks.
Hint: You may use a log delta analysistool or you may derive a dependency graph for FilteredHigh and another forFilteredLow, and abstract away all infrequent behaviors in the dependency graph(i.e., set the task and the arc abstraction sliders to their minimum)
