User-guided discovery of declarative process models
Summary (2 min read)
Introduction
- These problems can be solved by discovering declarative models.
- Logs provide detailed information about systems and human behaviour.
- Instead of explicitly specifying all the allowed sequences of events in a business process, the possible ordering of events is implicitly specified with constraints, i.e., rules that must be followed during execution.
- To solve this problem the authors propose to apply the truncated semantics introduced in [19] to discover significant DECLARE constraints also from truncated process instances.
II. PRELIMINARIES
- DECLARE is a declarative language proposed by Pesic and Van der Aalst in [8] [9].
- Templates are abstract entities that define parameterised classes of properties, and constraints are their concrete instantiations.
- The co-existence(A,B) template specifies that if one of the events A or B occurs, the other one should also occur.
- Finally, the succession(A,B) template requires that both response and precedence relations hold between the events A and B. Templates alternate response, alternate precedence and alternate succession strengthen the above templates by specifying that events must alternate without repetitions of these events in between.
III. APPROACH
- The starting point of their work was a case study concerned with the monitoring of vessel behaviour in the domain of maritime safety and security.
- The first step of their discovery algorithm is to generate a DECLARE model Dcandidates consisting of candidate DECLARE constraints.
- At the end of the checking phase, a filtered LTL model L including the remaining LTL rules is available.
- If PoE = 50% the discovered constraints will only involve 50% of the event classes in the log (the most frequent ones).
- The DECLARE Miner generates a DECLARE model object by using the algorithm described in subsection III-A.
IV. ADVANCED MINING TECHNIQUES
- The authors introduce two advanced techniques to support the discovery of DECLARE models: the truncated semantics for LTL formulas and the LTL vacuity detection.
- Often the available logs are extracted from larger logs and the process instances are prefixes of larger process instances.
- In a truncated path the truth value of an LTL formula can be non-definitive (i.e., temporarily violated or temporarily satisfied).
- To address this problem, [19] introduces a strong semantics and a weak semantics for LTL formulas where a formula is evaluated to false and true respectively if its truth value is non-definitive.
- A typical example of a vacuously satisfied constraint is given by the formula “every request is eventually acknowledged” in a process instance that does not contain requests.
V. CASE STUDY
- The authors present the results of the DECLARE mining phase of a larger case study on the monitoring of vessel behaviour.
- In the considered log, each process instance corresponds to a specific vessel.
- V shows the settings for an experiment aimed at discovering chain response constraints for vessel type passenger ship with a PoI of 100% (i.e., constraints satisfied by all instances).
- TABLE VII shows the results for this experiment.
- The authors specify the percentage of nonvacuously satisfied instances, i.e., interesting witnesses, and also the PoI parameter, i.e., the percentage of process instances where the constraint is (vacuously or non-vacuously) satisfied.
VI. CONCLUSION
- The authors have introduced a novel approach to discover declarative models from logs that allows users to guide the discovery process towards specific properties.
- Moreover, the authors have shown how results on truncated semantics can be used to obtain significant results in the case that only partial logs are available.
- The authors have also applied results on vacuity detection to identify, for each discovered constraint, the percentage of interesting witnesses, i.e., process instances where the constraint is non-trivially valid.
- Also, given a constraint and a process instance where it is violated, the level of “healthiness” of the process instance can be evaluated based on the number of violations.
Did you find this useful? Give us your feedback
Citations
147 citations
126 citations
113 citations
99 citations
89 citations
References
10,863 citations
1,953 citations
1,593 citations
1,111 citations
952 citations
Related Papers (5)
Frequently Asked Questions (12)
Q2. What have the authors stated for future works in "User-guided discovery of declarative process models" ?
In the near future, the authors want to extend their approach by providing users with the possibility to discover strongly, neutrally or weakly satisfied constraints depending on the level of reliability or flexibility they need in the discovery process. Given a constraint and a process instance where it is nonvacuously satisfied, it could also be useful to provide further information about how many times the constraint has been “ activated ” in the process instance ( counting the number of violations for ¬witness ( ϕ ) ). Also, given a constraint and a process instance where it is violated, the level of “ healthiness ” of the process instance can be evaluated based on the number of violations.
Q3. What is the parameter used to avoid the discovery of less-relevant constraints?
The parameter Percentage of Events (PoE) can be used to avoid the discovery of less-relevant constraints referring to event classes which rarely occur in the log.
Q4. How many candidate constraints are generated and checked?
For instance, for a log including 30 event classes and a single template with 4 parameters, 810.000 candidate constraints are generated and checked.
Q5. What is the proposed approach for the discovery of DECLARE models?
Their proposed approach for the discovery of DECLARE models is used in the first phase of the case study to build a declarative reference model (representing the normal behaviour of vessels) starting from historical logs.
Q6. What is the sub-log associated to a tanker?
The sub-log associated to vessel type tanker/unknown cargo type D is composed of an alternation of event under way using engine followed by moored or at anchor.
Q7. What is the first step of the discovery algorithm?
The first step of their discovery algorithm is to generate a DECLARE model Dcandidates consisting of candidate DECLARE constraints.
Q8. What is the percentage of events that will be used to generate the candidate constraints?
For instance, if PoE = 50% the discovered constraints will only involve 50% of the event classes in the log (the most frequent ones).
Q9. What are the two advanced techniques to support the discovery of DECLARE models?
In this section, the authors introduce two advanced techniques to support the discovery of DECLARE models: the truncated semantics for LTL formulas and the LTL vacuity detection.
Q10. What is the definition of the existence template?
The existence templates involve only one event (unary relationship) and define the cardinality or the position of an event in a process instance.
Q11. What is the definition of the succession(A,B) template?
the succession(A,B) template requires that both response and precedence relations hold between the events A and B.Templates alternate response, alternate precedence and alternate succession strengthen the above templates by specifying that events must alternate without repetitions of these events in between.
Q12. What is the LTL constraint generation phase?
The DECLARE Miner generates a DECLARE model object by using the algorithm described in subsection III-A.The LTL constraints generation phase is supported by the DECLARE2LTL plug-in.