CN107391715A - A kind of process model mining method of free single cycle structure - Google Patents

A kind of process model mining method of free single cycle structure Download PDF

Info

Publication number
CN107391715A
CN107391715A CN201710635987.8A CN201710635987A CN107391715A CN 107391715 A CN107391715 A CN 107391715A CN 201710635987 A CN201710635987 A CN 201710635987A CN 107391715 A CN107391715 A CN 107391715A
Authority
CN
China
Prior art keywords
activity
mark
single cycle
free single
movable
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201710635987.8A
Other languages
Chinese (zh)
Inventor
杜玉越
贺朝阳
王路
张福新
李鹏
刘伟
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shandong University of Science and Technology
Original Assignee
Shandong University of Science and Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shandong University of Science and Technology filed Critical Shandong University of Science and Technology
Priority to CN201710635987.8A priority Critical patent/CN107391715A/en
Publication of CN107391715A publication Critical patent/CN107391715A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/24—Querying
    • G06F16/245—Query processing
    • G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
    • G06F16/2465—Query processing support for facilitating data mining operations in structured databases
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0633—Workflow analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/20—Administration of product repair or maintenance
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2216/00—Indexing scheme relating to additional aspects of information retrieval not explicitly covered by G06F16/00 and subgroups
    • G06F2216/03—Data mining

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • General Physics & Mathematics (AREA)
  • Strategic Management (AREA)
  • Operations Research (AREA)
  • Databases & Information Systems (AREA)
  • General Business, Economics & Management (AREA)
  • Quality & Reliability (AREA)
  • Marketing (AREA)
  • Tourism & Hospitality (AREA)
  • Fuzzy Systems (AREA)
  • Game Theory and Decision Science (AREA)
  • Educational Administration (AREA)
  • Development Economics (AREA)
  • Mathematical Physics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Software Systems (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The invention discloses a kind of process model mining method of free single cycle structure.The process model mining method is to solve the problems, such as the excavation of the short circulation in parallel construction, and on the basis of α algorithms, α is proposed for freely single-cycle structurefslAlgorithm.The algorithm has redefined the completeness of the daily record comprising loop structure, and new cyclic order relation is with the addition of on the basis of basic activity orbution.By pre-processing daily record, the activity repeated in daily record is extracted, lists the neighbouring relations of repetition activity, therefrom finds loop structure present in daily record, carrys out excacation stream process model in a manner of the later stage adds loop structure.Finally, the present invention verifies α by the instance analysis to certain computer maintenance companyfslThe validity and correctness that algorithm excavates.

Description

A kind of process model mining method of free single cycle structure
Technical field
The present invention relates to a kind of process model mining method of free single cycle structure.
Background technology
Process model mining is built bridge between the knowledge in data mining and process model building and two fields of analysis, from event day Knowledge is extracted in will, and then finds, monitor and improve real process.The scene for having three types process model mining has used event day Will, respectively process are found, consistency detection and process strengthen.Process is the discovery that using the Web log mining not comprising prior information Obtain model；Consistency detection is whether the process model for comparing daily record actual conditions and excavating coincide, or mould known to comparison Whether type and the situation of its caused daily record coincide；Process enhancing is for an already present process, is produced using real process Raw daily record makes improvements or extended.In process model mining, the completeness of daily record is vital.Completeness and noise It is similar, the difference is that very few this of data that noise represents in completeness expression daily record to contain containing excessive data in daily record is asked Topic.Lacking the non-complete event log of special daily record behavior will have an impact to the real work of mining algorithm, cause Correct model can not be excavated.Noise and incompleteness describe the problem of quality on event log, and on model The measurement of quality can not be simply judged, it is necessary to be described from different dimensions.Process model has degree of fitting, accuracy, general Change degree and succinctly spend this four main measure dimensions.Mark in the degree of fitting expression event log of model is in process model Repeatability, its behavior for allowing to be reflected in event log of the high specification of a model of degree of fitting occur.The accuracy of model determines Whether model allows the behavior beyond daily record to occur, and the accuracy of model is high to illustrate that it does not allow the behavior beyond excessive daily record to send out It is raw.The extensive degree representative model of model is not limited solely to the behavior in daily record, it is allowed to more extra behaviors.Succinctly spend Specification of a model is can to explain behavior and simplest model seen in daily record.At present, domestic and foreign scholars propose many mistakes Journey mining algorithm, such as：
(VAN DER AALST W, WEIJTERS T, the MARUSTER L.Workflow mining of document 1: Discovering process models from event logs[J].IEEE Transactions on Knowledge and Data Engineering,2004,16(9):1128-1142) propose the α algorithms of excacation flow model and demonstrate The version that can accurately excavate.From when putting into practice mining process model in daily record, work corresponding to many actual operation flows Make labyrinth in flow model be present, such as non-free selection structure (non-free-choices), invisible task (invisible tasks), duplication of name task (duplicate tasks) and loop structure different in size (loop).α algorithms By the excacation for analyzing the dependence implementation model in daily record between activity.The definition of these dependences is based on local Daily record behavior, it can not be handled for the labyrinth such as short circulation and non-free selection structure.Document 2 (DE MEDEIROS A, VAN DER AALST W,WEIJTERS T.Workflow Mining:Current Status and Future Directions [J].Lecture Notes in Computer Science,2003,2888(9):389--406) the α proposed+Algorithm extends α algorithms, excavated for the short circulation that length is 1 and 2.Document 3 (WEN L, VAN DER AALST W, WANG J, et al.Mining process models with non-free-choice constructs[J].Data Mining and Knowledge Discovery,2007,15(2):Non-free selection structure can be handled by 145-180) proposing on this basis α++Algorithm.Document 4 ([hears vertical process model mining algorithm research [D] the Beijing of outstanding based on Workflow net：Tsing-Hua University, 2007) The time is needed to propose β algorithms using tasks carrying, for solving situations such as short circulation and how concurrent complex branches.Document 5 (cares for Spring qin, often making friends, the such as Tao Qian can solve process model mining algorithm [J] computer integrated manufacturing systems of Various Complex task, 2009,15(11):2193-2198) cycle task and iterative task in daily record are found using heuristic rule, it is proposed that τ is calculated Method.Document 6 (VAN DER AALST W, DE MEDEIROS A, WEIJTERS A.Genetic process mining [C] // International Conference on Application and Theory of Petri Nets.Springer Berlin Heidelberg,2005:48-69) and (DE MEDEIROS A, WEIJTERS A, the VAN DER AALST of document 7 W.Genetic process mining:an experimental evaluation[J].Data Mining and Knowledge Discovery,2007,14(2):Genetic algorithm 245-304) is applied to process model mining, solved non-free Select the Mining Problems of structure, invisible task and duplication of name task dispatching labyrinth.Excavate, based on shape in region based on language The region of state is excavated and ILP mining algorithms have a good guarantee to the degree of fitting of model.Document 8 (LEEMANS S, FAHLAND D,VAN DER AALST W.Discovering block-structured process models from event logs-a constructive approach[C]//International Conference on Applications and Theory of Petri Nets and Concurrency.Springer Berlin Heidelberg,2013:311-329) Excavation for block structure process model proposes IM algorithms, and the model for excavating to obtain has high degree of fitting.Do not examining In the case of considering duplication of name activity, for the event log of loop structure be present, the work that frequency of occurrence is more than 2 is certainly existed It is dynamic；For α serial algorithms, when loop structure and other structures are in the different branch of parallel relation, between the activity of definition Dependence exist limitation；When not occurring continuously for the activity in the same loop structure in daily record, α serial algorithms will The loop structure being nested in parallel construction can not correctly be found.
The content of the invention
It is an object of the invention to propose a kind of process model mining method of free single cycle structure, process model mining method master To be realized by pre-processing daily record, main target of excavating is to excavate to be present in parallel junction in the circulation Complete Log redefined Free single cycle structure in structure, this method are equally applicable to the event log in the absence of parallel construction.
To achieve these goals, the present invention adopts the following technical scheme that：
A kind of process model mining method of free single cycle structure, comprises the following steps：
The definition of s1 basic conceptions
Define multiple set：
If S is a set, the multiple set on a S is defined as a function Z:B (S) represents the institute on S Have a set that multiset is combined into, a multiple set by a vector representation,Represent a natural number set；
Defined nucleotide sequence：
If S is a set, S* is the set of all finite sequence compositions on S；
Define mark：
If A is active set, a mark σ ∈ A* is a movable queue；
Define event log：
Event log L is mark σ multiple set, is designated as L ∈ B (A*)；
Define Petri network：
Four-tuple PN=(P, T；F, M) it is referred to as a Petri network, and if only if：
(1) N=(P, T；F) it is a net；Wherein：
P is a limited set of library, and T is the limited transition collection of active set in a corresponding event log, It is a finite arc collection；
(2)M:P → N is referred to as a mark for netting PN, wherein, miFor initial marking, mfIdentified to terminate；
(3) there are PN following transition rule occurs：
(a) for changing t ∈ T, ifThen claim transition t to be enabled in the case where identifying M, be designated as M [t>；
If (b) M [t>, then in the case where identifying M, transition t can occur, and trigger transition t to obtain a new mark from mark M M ', it is designated as M [t>M ', and it is right
Wherein, M (p) represents mark place p at, and M ' (p) represents to trigger the mark changed after t at place p,●T tables Show t input set, t●Represent t output collection；
Define Workflow net：
Hexa-atomic group of WFN=(P, T；F,M,in,on) being referred to as a Workflow net, and if only if：
(1) Σ=(P, T；F, M) it is a Petri network；
(2) an initial place i be presentn∈ P, have
(3) a terminate storehouse institute o be presentn∈ P, have
(4) forY is at one from inTo onPath on；
Wherein,●inRepresent inInput set,●onRepresent onOutput collection；
Definition mining algorithm α：
W is made to represent the workflow logs for covering limited transition collection T；α (W) is defined as follows：
(1)
(2)
(3)
(4)
(5)
(6)PW={ p (AW,BW)|(AW,BW)∈YW}∪{iW,oW}；
(7)FW={ (aW,p(AW,BW))|(AW,BW)∈YW∧aW∈AW}∪{(p(AW,BW),bW)|(AW,BW)∈YW∧bW ∈BW}∪{(iW,t)|t∈TI}∪{(t,oW)|t∈TO}；
(8) α (W)=(PW,TW,FW)；
Wherein, TWIt is the transition in Workflow net corresponding to the activity occurred in daily record；
TIIt is the set to come into play, i.e., all movable set of first position is appeared in mark；
TOIt is the set of ending activity, i.e., all movable set of last position is appeared in mark；
p(AW,BW) it is a place, wherein, AWIt is p (AW,BW) input transition set,●p(AW,BW)=AW；BWIt is p (AW,BW) output transition set, p (AW,BW)●=BW, all A for meeting the requirement of (4) stepW, BWX is combined into the collection of compositionW；
For XWEvery a pair of (AW,BW), nonempty setAnd nonempty setThere is (AW′,BW′) ∈XW, in algorithm (4) step, all is not that maximum set is all removed, and obtained collection is combined into YW；
PWInclude a unique source place iWWith a unique terminate storehouse institute oW；
The set F of the arc of (7) step generation Workflow netW, TIIn all changes move the capital to another place with iWAs input place, TOIn All changes move the capital to another place with OWAs output place；
All place p (AW,BW) all with AWAs input node, with BWAs output node；
The result finally given is Petri network α (W)=(PW,TW,FW), it describes the behavior that daily record W is reflected；
First (σ) represents first activity in mark σ, and last (σ) represents last activity in mark σ；
S2 defines free single cycle structure
Define free single cycle structure：
A WFN free single cycle structure is shown as FSL=by triple table<Tstart,Tend,Tset>；
Wherein：Tstart、Tend∈ T and Tset∈ T*, T* represent the sequence of all activity compositions in limited transition collection T；
So that Tstart∈p0 ●And Tend∈●p0；
Sequence SE=be present<p0,Tstart,p1,…,pn,Tend,p0>, there is Tset=<Tstart,…,Tend>；
For a free single cycle structure：
TstartFirst transition in structure are represented, i.e., are changed corresponding to incipient activity of the structure in daily record；TendTable Show last transition in structure, i.e., changed corresponding to termination activity of the structure in daily record；TsetRepresent institute in the structure By transition corresponding to the sequence that forms in order of activity in daily record, i.e., free single cycle sequence；
The order of the sequence is each movable order when the loop structure first pass occurs in the mark of daily record；
It is defined for the completeness of the daily record containing free single cycle structure：
Define free single cycle Complete Log：
Make WfslIt is the complete workflow logs of a free single cycle, A is active set, and and if only ifAnd σ1、σ2∈WfslSo that a ∈ σ1, σ1(a) >=2 and
For each mark in the daily record, the activity continuously repeated is all not present, but the work that occurrence number is more than 1 be present It is dynamic；
For the relation between the activity in daily record, free single cycle relation is introducedTo represent in same freely single The relation between front and rear adjacent activity in loop structure；
Definition includes the active sequence relation of free single cycle relation：
Make WfslIt is the complete workflow logs of a free single cycle, A is active set, and T is limited transition corresponding to A Collection, Wfsl∈P(T*)；Make a, b ∈ T：
- a ＞Wb：And if only if exist a mark σ=<a1,a2,a3,…,an>, i ∈ { 1 ..., n-1 } are so that σ ∈ W, ai =a and ai+1=b；
—a→Wb：And if only if a ＞WB and b ≯Wa；
—a#Wb：And if only if a ≯WB and b ≯Wa；
—a||Wb：And if only if a ＞WB and b ＞Wa；
—aa：And if only if, and a is present in the free single cycle structure that length is 1；
—ab：And if only if has the free single cycle sequence T that a length is nset=<t1,t2,t3,…,tn>, Meet Tset∈T*,ti=a and ti+1=b；t1,t2,t3,…,tnRepresent TsetIn activity；
Daily record WfslIt is the multiple set of the mark of transition composition, meets that free single cycle is complete；A, b are limited transition Activity corresponding to two transition in set；＞WExpression follows relation, →WRepresent causality, #WExclusive relation is represented, | |W Represent parallel relation,Represent free single cycle relation；
Before free single cycle structure is excavated, initial order relation movable in daily record is obtained according to algorithm 1, wherein not Include free single cycle relation；R represents the movable initial order set of relationship obtained based on daily record；
Algorithm 1 lists initial order relation movable in daily record, and it is comprised the following steps that：
Step 1:Active sequence set of relationship R is initialized as empty set；
Step 2:Belong to the complete workflow logs W of free single cycle for allfslMark σ, perform following operate：
Step 3:For the movable a in the mark σ of a n activityi, from 1 to n, execution is following to be operated i values：
Step 4:By ai＞Wai+1Add active sequence set of relationship R；
Repeat step 4, until i is equal to n；
Repeat step 3, until having traveled through all mark σ；
Step 5:For each element in active sequence set of relationship R, following operate is performed：
Each element in wherein R is one group of relation being made up of two parameters, and movable a is the previous ginseng of relation Number, movable b is the latter parameter of relation；
Step 6:If 1) relation a ＞ be presentWB and b ≯WA, then by relation a →WB adds active sequence set of relationship R；
If 2) relation a ＞ be presentWB and b ＞WA, then by relation a | |WB adds active sequence set of relationship R；
If 3) exist relation a ≯WB and b ≯WA, then by relation a#WB adds active sequence set of relationship R；
Repeat step 6, all elements perform end of operation in R；
Step 7:Export active sequence set of relationship R；
The free single cycle structure mining algorithms of s3
S3.1 finds the activity of repetition
According to the definition of free single cycle structure, all occurrence numbers are more than 1 activity in a mark of daily record, all belong to The activity included in free single cycle structure；Definition repeats active set to deposit these activities：
Definition repeats active set RepetitiveActivitySet：
If A is active set, WfslFor the complete workflow logs of free single cycle,When and only When So that ar∈ σ, and σ (ar)≥2；
Algorithm 2 finds the activity of repetition and adds all repetition activities and repeats in active set, and it is comprised the following steps that：
Step 1:Repeat active set RepetitiveActivitySet and be initialized as empty set；
Step 2:Belong to the complete workflow logs W of free single cycle for allfslMark σ, perform following operate：
Step 3:For the movable a in a mark σ comprising n activityi,aj, i values from 1 to j, j values from i to n-1, Perform following operate：
Step 4:If aiWith ajEqual and aiIt is not belonging to repeat active set RepetitiveActivitySet, then by ai Add and repeat active set RepetitiveActivitySet；
Repeat step 4, until i is equal to n-1；
Repeat step 3, until having traveled through all mark σ；
Step 5:Output repeats active set RepetitiveActivitySet；
S3.2 has found to repeat forerunner's activity of activity and succeeding activity
The activity repeated in active set is not divided into clearly among each free single cycle structure；
Obtained repetition activity is present in below by the movable forerunner's activity of mark repetition and succeeding activity is defined Which free single cycle structure makes a distinction offer support；
Define the movable forerunner's activity of mark repetition and succeeding activity：
If A is active set, WfslIt is the complete workflow logs of free single cycle, RepetitiveActivitySet is WfslRepetition active set；
ForSo that ar∈ σ, there is ai=ar, 1 ＜ I ＜ n, then claim ai-1For arForerunner activity, ai+1For arSucceeding activity；a1,a2,…,anRepresent the activity in mark σ.
For any activity in repetition active set, all movable forerunners' activity compositions in each mark of daily record The movable mark collects before repeating activity, and all appear in after the movable succeeding activity forms movable mark repetition activity collects；
The activity that algorithm 3 repeats for each, collect after obtaining the movable preceding collection of mark repetition activity and mark repetition activity, Comprise the following steps that：
Step 1:For repeating each movable a in active set RepetitiveActivitySetr, perform following grasp Make：
Step 2:For the complete workflow logs W of the free single cycle of daily recordfslIn each mark σ, perform following grasp Make：
Step 3:For the movable a in the mark σ comprising n activityi, from 2 to n-1, execution is following to be operated i values：
Step 4:If 1) aiWith arIt is equal, and ai-1It is not belonging to arMark repeat activity before collect, then by ai-1Add arMark Collect a before repeating activityr.TracePreSet；
If 2) aiWith arIt is equal, and ai+1It is not belonging to arMark repeat activity after collect, then by ai+1Add arMark repeat live Collect a after dynamicr.TracePostSet；
Repeat step 4, until i is equal to n-1；
Repeat step 3, until having traveled through all mark σ；
Repeat step 2, all movable a in RepetitiveActivitySetrPerform end of operation；
Step 5:The each movable a of outputrMark repeat activity before collection ar.TracePreSet collect after repeating activity with mark ar.TracePostSet；
Movable mark by obtaining each repetition repeats to collect after the preceding collection of activity repeats activity with mark, to distinguish its presence Which prepared in free single cycle structure；
S3.3 excavates free single cycle structure
Algorithm 4 is excavated to all free single cycle structures, and it is comprised the following steps that：
Step 1:Free single cycle structure set FreeSingleLoopSet is initialized as empty set, and length is 2 initial live Dynamic set Length2TstartSet is initialized as empty set, and the termination active set Length2TendSet that length is 2 is initialized For empty set, length is not initialized as empty set for 2 repetition active set NoLength2ActivitySet, other length cycles collection Close OtherLoopSet and be initialized as empty set；
Step 2:For repeating each movable a in active set RepetitiveActivitySetr, perform following grasp Make：
Step 3:If 1) movable arMark repeat activity after collect ar.TracePostSet collected before its mark repeats activity ar.TracePreSet proper subclass, then by arAdd the initial active set Length2TstartSet that length is 2；
If 2) movable arMark repeat activity before collection ar.TracePreSet collected after its mark repeats activity ar.TracePostSet proper subclass, then by arAdd the termination active set Length2TendSet that length is 2；
If 3) movable arMark repeat activity before collection ar.TracePreSet collect after repeating activity with its mark ar.TracePostSet not proper subclass each other, then by arAdd the repetition active set that length is not 2 NoLength2ActivitySet；
Repeat step 3, all movable a in RepetitiveActivitySetrPerform end of operation；
Step 4:Each the movable a in initial active set Length2TstartSet for being 2 for lengths, perform Operate below：
Step 5:One free single cycle sequence TsetIt is initialized as sky；
Step 6:If in the presence of a movable bsBelong to the termination active set Length2TendSet that length is 2, for certainly By the complete workflow logs W of single cyclefslIn each mark σ, movable asWith movable bsOccurrence number phase in current mark σ Together, then in the presence of a free single cycle sequence TsetFor<as,bs>；
Step 7:By free single cycle sequence Tset=<as,bs>Add free single cycle structure set FreeSingleLoopSet；
Step 8:It is not each activity in 2 repetition active set NoLength2ActivitySet for length at, perform following operate：
Step 9:Unordered set Set is initialized as { at, a free single cycle sequence TsetIt is initialized as sky；
Step 10:For each movable btBelong to the repetition active set that length is not 2 NoLength2ActivitySet and btNot equal to at, perform following operate：
Step 11:If movable atMark repeat activity before collection at.TracePreSet with movable btMark repeat activity before collect bt.TracePreSet common factor is equal to movable atMark repeat activity after collect at.TracePostSet with movable btMark repeat Collect b after activityt.TracePostSet common factor, then by btAdd unordered set Set；
Repeat step 11, it is all in NoLength2ActivitySet to be different from movable atActivity perform operation Finish；
Step 12:By each activity in unordered set Set according to this in presence these movable mark σ in daily record The order sequence that a little activities occur first, obtains a free single cycle sequence Tset；
Step 13:By free single cycle sequence TsetAdd free single cycle structure set FreeSingleLoopSet；
Repeat step 9 arrives step 13, all activity execution end of operations in NoLength2ActivitySet；
Step 14:Export free single cycle structure set FreeSingleLoopSet；
S4 establishes the orbution of log activity
For each free single cycle structure in free single cycle structure set, the orbution between activity therein Need to redefine according to the definition of free single cycle relation；
Algorithm 5 redefines the orbution between free single cycle structure internal activity, and it is comprised the following steps that：
Step 1:Active sequence set of relationship R is initialized as the active sequence set of relationship R exported in algorithm 1；
Step 2:For all free single cycle sequences belonged in free single cycle structure set FreeSingleLoopSet Arrange Tset=<t1,t2,t3,…,tn>, for TsetIn each movable ti, from 1 to n-1, execution is following to be operated i values：
Step 3:1) if n is equal to 1, by relation t1 t1Add active sequence set of relationship R；
Step 4:If 2) n is not equal to 1, following operate is performed：
Step 5:By relation ti ti+1Add active sequence set of relationship R；
Repeat step 5, until i is equal to n-1, and by relation tk t1Add active sequence set of relationship R；
Step 6:Export active sequence set of relationship R.
The invention has the advantages that：
The present invention is to solve the problems, such as the excavation of short circulation in parallel construction, on the basis of α algorithms, for freely singly following Ring structure proposes αfslAlgorithm, the algorithm have redefined the completeness of the daily record comprising loop structure, and in basic activity New cyclic order relation is with the addition of on the basis of order relation；By pre-processing daily record, the activity repeated in daily record is extracted, The neighbouring relations of repetition activity are listed, therefrom find loop structure present in daily record, in a manner of the later stage adds loop structure Carry out excacation stream process model.Finally, α is verified by the instance analysis to certain computer maintenance companyfslAlgorithm excavates effective Property and correctness.
Brief description of the drawings
Fig. 1 is N in the present invention0Length is the schematic diagram that 1 loop structure and single-unit activity are in parastate；
Fig. 2 is N in the present invention1The loop structure that the loop structure that length is 1 is 2 with length is in the signal of parastate Figure；
Fig. 3 is three kinds of different N in the present invention2The loop structure of length is in the schematic diagram of parastate；
Fig. 4 is N in the present invention3Length is the schematic diagram of 1 free single cycle structure；
Fig. 5 is N in the present invention4Length is the schematic diagram of 2 free single cycle structure；
Fig. 6 is N in the present invention5Length is the schematic diagram of 3 free single cycle structure；
Fig. 7 is α in the present inventionfslThe model schematic that algorithm excavates；
Fig. 8 is α in the present invention+The model schematic that algorithm excavates；
Fig. 9 is the model schematic that ILP algorithms excavate in the present invention；
Figure 10 is the model schematic that IM algorithms excavate in the present invention；
Figure 11 is the degree of fitting comparison schematic diagram of different each algorithm mining models of daily record in the present invention；
Figure 12 is for the accuracy comparison schematic diagram of each algorithm mining model of different daily records in the present invention.
Embodiment
Below in conjunction with the accompanying drawings and embodiment is described in further detail to the present invention：
A kind of process model mining method of free single cycle structure, it comprises the following steps：
The definition of s1 basic conceptions
Multiple set, sequence, mark, event log, Petri network, Workflow net, mining algorithm α etc. are described in this step Basic conception.In ensuing content,A natural number set is represented, for example,
Define multiple set：
If S is a set, the multiple set on a S is defined as a function Z:B (S) is represented on S The set that all multisets are combined into.One multiple set can be by a vector representation.Such as：
If S={ a, b, c }, Z=[a, b2,c3] it is multiple set on a S, there is Z (a)=1, Z (b)=2, Z (c)= 3。
Defined nucleotide sequence：
If S is a set, S* is the set of all finite sequence compositions on S.σ=<σ[1],…,σ[n]>It is a sequence Row, σ [i] represent i-th of element in sequence σ, | σ | represent sequence σ length.
Forσ ' (a) represents numbers of a in the middle appearance of σ '.Such as σ '=<a,a,a,b,b,c>, σ ' (a)=3.
Define mark：
If A is active set, a mark σ ∈ A* is a movable queue.
Define event log：
Event log L is mark σ multiple set, is designated as L ∈ B (A*).
The oriented bipartite graph that Petri is made up of place (place) and transition (transition).
Define Petri network：
Four-tuple PN=(P, T；F, M) it is referred to as a Petri network, and if only if：
(1) N=(P, T；F) it is a net；Wherein：P is a limited set of library, and T is living in a corresponding event log The limited transition collection of dynamic set,It is a finite arc collection；
(2)M:P → N is referred to as a mark for netting PN, wherein, miFor initial marking, mfIdentified to terminate；
(3) there are PN following transition rule occurs：
(a) for changing t ∈ T, ifThen claim transition t to be enabled in the case where identifying M, be designated as M [t>；
If (b) M [t>, then in the case where identifying M, transition t can occur, and trigger transition t to obtain a new mark from mark M M ', it is designated as M [t>M ', and it is right
Wherein, M (p) represents mark place p at, and M ' (p) represents to trigger the mark changed after t at place p,●T tables Show t input set, t●Represent t output collection.
Workflow net is a Petri network for having unique initial place and terminate storehouse institute.
Define Workflow net：
Hexa-atomic group of WFN=(P, T；F,M,in,on) being referred to as a Workflow net, and if only if：
(1) Σ=(P, T；F, M) it is a Petri network；
(2) an initial place i be presentn∈ P, have
(3) a terminate storehouse institute o be presentn∈ P, have
(4) forY is at one from inTo onPath on；
Wherein,●inRepresent inInput set,●onRepresent onOutput collection；
Definition mining algorithm α：
W is made to represent the workflow logs for covering limited transition collection T；α (W) is defined as follows：
(1)
(2)
(3)
(4)
(5)
(6)PW={ p (AW,BW)|(AW,BW)∈YW}∪{iW,oW}；
(7)FW={ (aW,p(AW,BW))|(AW,BW)∈YW∧aW∈AW}∪{(p(AW,BW),bW)|(AW,BW)∈YW∧bW ∈BW}∪{(iW,t)|t∈TI}∪{(t,oW)|t∈TO}；
(8) α (W)=(PW,TW,FW)；
Wherein, TWIt is the transition in Workflow net corresponding to the activity occurred in daily record；
TIIt is the set to come into play, i.e., all movable set of first position is appeared in mark；
TOIt is the set of ending activity, i.e., all movable set of last position is appeared in mark；
p(AW,BW) it is a place, wherein, AWIt is p (AW,BW) input transition set,●p(AW,BW)=AW；BWIt is p (AW,BW) output transition set, p (AW,BW)●=BW, all A for meeting the requirement of (4) stepW, BWX is combined into the collection of compositionW；
For XWEvery a pair of (AW,BW), nonempty setAnd nonempty setThere is (AW′,BW′) ∈XW, in algorithm (4) step, all is not that maximum set is all removed, and obtained collection is combined into YW；
PWInclude a unique source place iWWith a unique terminate storehouse institute oW；
The set F of the arc of (7) step generation Workflow netW, TIIn all changes move the capital to another place with iWAs input place, TOIn All changes move the capital to another place with OWAs output place；
All place p (AW,BW) all with AWAs input node, with BWAs output node；
The result finally given is Petri network α (W)=(PW,TW,FW), it describes the behavior that daily record W is reflected；
First (σ) represents first activity in mark σ, and last (σ) represents last activity in mark σ；
S2 defines free single cycle structure
α algorithms and its derivative algorithm, limitation when excavating loop structure nested in parallel construction be present.
First, α algorithms and its derivative algorithm, when being excavated for the loop structure of different length, situations below be present：
(1) it is 3 and longer loop structure for length, basic α algorithms just can solve the problem that, because being included in circulation During the activity of 3 and the above, using the relation that follows directly after i.e. (>L) can clearly express and distinguish parallel and loop structure.
(2) for the loop structure (i.e. from ring structure) that length is 1, basic α algorithms can be by the activity in loop structure It is identified as an activity with model other parts independence.
(3) for loop structure that length is 2, it is found that two activities in loop structure exist direct from each other Follow relation and completely the same with the definition of parallel relation, that is, the caused footprint in event log is identical.
In the derivative algorithm of α algorithms, including α+The methods of algorithm etc. passes through post processing with the addition of the circulation knot that length is 1 Structure；The methods of redefining parallel relation by pretreatment, the differentiation for solving loop structure and parallel construction that length is 2 are asked Topic.
When loop structure is present in parallel construction, if there is no the feature for including clear and definite loop structure in daily record (the activity continuous several times included in loop structure occur), existing α algorithms and its part derivative algorithm just can not be good It was found that the loop structure being present in parallel construction.Concrete condition is as shown in Figure 1, Figure 2 and shown in Fig. 3.
The present invention is only for as shown in Figure 1, Figure 2 and the loop structures of Fig. 3 shown types is excavated.
When the activity or active set that do not include continuous several times appearance in any mark in daily record, such as：
σ=<…a,a,a…>Or σ=<…a,b,a,b…>Situations such as, using α algorithms and its part derivative algorithm, just not The such loop structure included in parallel construction can effectively be excavated.
For this kind of loop structure in Fig. 1, Fig. 2 and Fig. 3, its initial storehouse institute and terminate storehouse institute are same place, Be referred to as free single cycle structure in the present invention, be the characteristics of this loop structure maximum exist in daily record a mark do not include it is this Activity in loop structure, and the loop structure is a single cycle, i.e., and it is sequential organization that circulation is internal.
Define free single cycle structure：
WFN=(P, T；F, M, i, o) it is a Workflow net, a WFN free single cycle structure is represented by triple For FSL=<Tstart,Tend,Tset>；
Wherein：Tstart,Tend∈ T and Tset∈ T*, T* represent the sequence of all activity compositions in limited transition collection T.So that Tstart∈p0 ●And Tend∈●p0。
Sequence S=be present<p0,Tstart,p1,…,pn,Tend,p0>, there is Tset=<Tstart,…,Tend>。
For a free single cycle structure, TstartFirst transition in structure are represented, i.e., the structure is in daily record Changed corresponding to incipient activity；TendLast transition in structure are represented, i.e. termination activity of the structure in daily record corresponds to Transition；TsetThe sequence that the activity in daily record corresponding to transition all in the structure forms in order is represented, i.e. freedom is singly followed Ring sequence.The order of the sequence is each movable order when the loop structure first pass occurs in the mark of daily record.
The example of the free single cycle structure for illustrating different length in Fig. 4, Fig. 5 and Fig. 6.
For the N in Fig. 65Represented length is its incipient activity T for 3 free single cycle structurestartFor B, Termination activity TendFor D, free single cycle sequence TsetFor<B,C,D>.
In the case where not considering to represent preference, there are same movable feelings if there is multiple in a mark in daily record Loop structure be present in condition, i.e. explanation.For α algorithms, movable a, b be present, according to following relation ＞LDefinition, if a ＞LB with And b ＞LA, it just can not judge that movable a and movable b is present in length for two activities of 2 loop structure or is present in flat Two activities in row structure.In the derivative algorithm α of α algorithms+In algorithm, by redefine parallel relation and causality come The ring that length is 2 is handled, its parallel relation a defined | |WIn b be not present same mark in exist σ=<…a,b, a…>Situation, in this case, be considered as causality a →WB processing.
In the case of nested loop structure in parallel construction, the repetition activity in any mark of daily record is not continuous Occur, but interted the activity of other parallel constructions, such as σ=<…a,x,a…>Or σ=<…a,b,x,a,b…>It is this Situation, α algorithms and its derivative algorithm can not be excavated correctly.These algorithms can identify the loop structure being nested in parallel construction For common sequential organization or the structure of mistake, it is impossible to which performance repeats the situation of activity.But for the excavation of loop structure When, when the activity duplicated in any mark of daily record, the influence of noise is excluded, the activity repeated is all that loop structure is deposited Embodiment.For the daily record comprising loop structure, existing has activity occurrence number not less than twice in a certain bar mark, that is, circulates Structure occurs at least twice in a mark, but necessarily need not continuously occur.
It is defined for the completeness of the daily record containing free single cycle structure：
Define free single cycle Complete Log：
Make WfslIt is the complete workflow logs of a free single cycle, A is active set, and and if only ifAnd σ1, σ2∈WfslSo that a ∈ σ1, σ1(a) >=2 and
Two marks in the complete daily record of free single cycle is met be present, there is activity occurrence number to be not less than 2 in a mark, And the activity does not occur in another mark.Such as the N in Fig. 10, its daily record for meeting that free single cycle is complete can To be W1={ ACD, ABCD, ACBD, ABCBD }.
For each mark in the daily record, the activity continuously repeated is all not present, but exist and repeat movable B.
For the relation between the activity in daily record, present invention introduces free single cycle relationTo represent in same The relation between front and rear adjacent activity in free single cycle structure.
Definition includes the active sequence relation of free single cycle relation：
Make WfslIt is the complete workflow logs of a free single cycle, A is active set, and T is limited transition corresponding to A Collection, Wfsl∈P(T*)；Make a, b ∈ T：
- a ＞Wb：And if only if exist a mark σ=<a1,a2,a3,…,an>, i ∈ { 1 ..., n-1 } are so that σ ∈ Wfsl, ai=a and ai+1=b；
—a→Wb：And if only if a ＞WB and b ≯Wa；
—a#Wb：And if only if a ≯WB and b ≯Wa；
—a||Wb：And if only if a ＞WB and b ＞Wa；
—aa：And if only if, and a is present in the free single cycle structure that length is 1；
—ab：And if only if has the free single cycle sequence T that a length is nset=<t1,t2,t3,…,tn>, Meet Tset∈T*,ti=a and ti+1=b；t1,t2,t3,…,tnRepresent TsetIn activity；
Daily record WfslIt is the multiple set of the mark of transition composition, meets that free single cycle is complete；A, b are limited transition Activity corresponding to two transition in set；＞WExpression follows relation, →WRepresent causality, #WExclusive relation is represented, | |W Represent parallel relation,Represent free single cycle relation.
Before free single cycle structure is excavated, initial order relation movable in daily record can be obtained according to algorithm 1, its In do not include free single cycle relation.R represents the movable initial order set of relationship obtained based on daily record.
Algorithm 1 lists initial order relation movable in daily record, and it is comprised the following steps that：
Step 1:Active sequence set of relationship R is initialized as empty set；
Step 2:Belong to the complete workflow logs W of free single cycle for allfslMark σ, perform following operate：
Step 3:For the movable a in the mark σ of a n activityi, from 1 to n, execution is following to be operated i values：
Step 4:By ai＞Wai+1Add active sequence set of relationship R；
Repeat step 4, until i is equal to n；
Repeat step 3, until having traveled through all mark σ；
Step 5:For each element in active sequence set of relationship R, following operate is performed：
Each element in wherein R is the relation of one group of oil, two parameter compositions, and movable a is the previous ginseng of relation Number, movable b is the latter parameter of relation；
Step 6:If 1) relation a ＞ be presentWB and b ≯WA, then by relation a →WB adds active sequence set of relationship R；
If 2) relation a ＞ be presentWB and b ＞WA, then by relation a | |WB adds active sequence set of relationship R；
If 3) exist relation a ≯WB and b ≯WA, then by relation a#WB adds active sequence set of relationship R；
Repeat step 6, all elements perform end of operation in R；
Step 7:Export active sequence set of relationship R.
In algorithm 1, travel through daily record in each mark, obtain it is two neighboring activity between follow relation, according to footprint The definition of matrix, obtain comprising the cause and effect in addition to free single cycle relation, parallel and exclusive relation movable initial Orbution.In comprising free single cycle relation footprint matrix, the activity that will be present in loop structure is singly followed using free Ring relational symbol is identified.For the free single cycle structure that length is 1, a is usedA replaces conventional a#WA this generations The incoherent relation identity of table；For the free single cycle structure that length is 2 and the above, according to free single cycle sequence TsetIn Order, make aB relation form represents TsetIn front and rear two adjacent activities.
The free single cycle structure mining algorithms of s3
For how to find unrestricted choice structure, the main algorithm for proposing three phases of the present invention.
(1) find and activity is repeated present in daily record.
(2) by repeating the work in activity free single cycle structure different from the relation differentiation between fore-aft travel in daily record It is dynamic.
(3) each free single cycle structure is excavated, obtains its free single cycle sequence Tset。
S3.1 finds the activity of repetition
The present invention does not consider to represent preference, according to the definition of free single cycle structure, in a mark of daily record it is all go out Occurrence number is more than 1 activity, belongs to the activity included in free single cycle structure.
Invention defines repeat active set to deposit these activities.
Definition repeats active set RepetitiveActivitySet：
If A is active set, WfslFor the complete workflow logs of free single cycle,When and only When So that ar∈ σ, and σ (ar)≥2。
Algorithm 2 describes how to find the activity of repetition and add all repetition activities to repeat active set In RepetitiveActivitySet, comprise the following steps that：
Step 1:Repeat active set RepetitiveActivitySet and be initialized as empty set；
Step 2:Belong to the complete workflow logs W of free single cycle for allfslMark σ, perform following operate：
Step 3:For the movable a in a mark σ comprising n activityi,aj, i values from 1 to j, j values from i to n-1, Perform following operate：
Step 4:If aiWith ajEqual and aiIt is not belonging to repeat active set RepetitiveActivitySet, then by ai Add and repeat active set RepetitiveActivitySet；
Repeat step 4, until i is equal to n-1；
Repeat step 3, until having traveled through all mark σ；
Step 5:Output repeats active set RepetitiveActivitySet.
The step 2 of algorithm 2 arrives step 4, judges with the presence or absence of the activity repeated in a mark in daily record, if it does, It is added into repetition active set RepetitiveActivitySet.
For the activity in RepetitiveActivitySet, being merely illustrative these activities is repeated in the mark of daily record Activity in the activity of appearance, that is, free single cycle structure, which free single cycle structure be not divided into specifically In.
Such as the N in Fig. 32, meet the complete daily record W of free single cycle for its one2=AH, ABH, ACDH, AEFGH,ABCDEFGH,ABECFDGH,ACBDH,ABCBDBEBFBGBH,ABCDEFGCDH,ABCEBDFCBDGH,AECFDGH, AEFCGDCDEFGH, ABCEDFGCDH, ABCBDECDFCDGH, ABCBDECDFCDGEBFGH, ABCEDFGBEFGCEBFBGDH }, The repetition active set RepetitiveActivitySet obtained according to algorithm 2 is { B, C, D, E, F, G }, contain it is all The activity repeated in a certain bar mark of daily record.
S3.2 has found to repeat forerunner's activity of activity and succeeding activity
Movable each freedom that is not divided into clearly repeated in active set RepetitiveActivitySet is singly followed Among ring structure.The present invention is deposited by defining the movable forerunner's activity of mark repetition and succeeding activity to obtained repetition activity It is which free single cycle structure makes a distinction offer support.
Define the movable forerunner's activity of mark repetition and succeeding activity：
If A is active set, WfslIt is the complete workflow logs of free single cycle.For So that ar∈ σ, there is ai=ar, 1 ＜ i ＜ n, then claim ai-1For arForerunner activity, ai+1For ar Succeeding activity, a1,a2,…,anRepresent the activity in mark σ.For repeating in active set RepetitiveActivitySet Any activity, all movable forerunners' activity composition movable marks collect before repeating activity in each mark of daily record TracePreSet, all appear in after the movable succeeding activity forms movable mark repetition activity collect TracePostSet. The activity that algorithm 3 repeats for each, collect after obtaining the movable preceding collection of mark repetition activity and mark repetition activity, to distinguish it It is present in which free single cycle structure is prepared.
The activity that algorithm 3 repeats for each, collect after obtaining the movable preceding collection of mark repetition activity and mark repetition activity, Comprise the following steps that：
Step 1:For repeating each movable a in active set RepetitiveActivitySetr, perform following grasp Make：
Step 2:For the complete workflow logs W of the free single cycle of daily recordfslIn each mark σ, perform following grasp Make：
Step 3:For the movable a in the mark σ comprising n activityi, from 2 to n-1, execution is following to be operated i values：
Step 4:If 1) aiWith arIt is equal, and ai-1It is not belonging to arMark repeat activity before collect, then by ai-1Add arMark Collect a before repeating activityr.TracePreSet；
If 2) aiWith arIt is equal, and ai+1It is not belonging to arMark repeat activity after collect, then by ai+1Add arMark repeat live Collect a after dynamicr.TracePostSet；
Repeat step 4, until i is equal to n-1；
Repeat step 3, until having traveled through all mark σ；
Repeat step 2, all movable a in RepetitiveActivitySetrPerform end of operation；
Step 5:The each movable a of outputrMark repeat activity before collection ar.TracePreSet collect after repeating activity with mark ar.TracePostSet。
The step 2 of algorithm 3 arrives step 4, each mark for including repetition activity, and the forerunner's activity that will repeat activity adds Enter the movable mark and repeat collection TracePreSet before activity, succeeding activity is added after the movable mark repeats activity and collected TracePreSet。
For previously mentioned example, the N in Fig. 32And its daily record W2, collect before movable B mark repetition activity can be obtained TracePreSet is { A, C, D, E, F, G }, and it is { C, D, E, F, G, H } to collect TracePostSet after movable B mark repetition activity. Collection TracePreSet is { A, B, D, E, F, G } before the mark of mobile C repeats activity, is collected after the mark repetition activity of mobile C TracePostSet is { B, D, E, F, G }.It is { B, C, E, F, G } that movable D mark, which repeats collection TracePreSet before activity, movable D Mark repeat activity after collect TracePostSet be { B, C, E, F, G, H }.Movable E mark repeats collection TracePreSet before activity For { A, B, C, D, G }, it is { B, C, D, F } to collect TracePostSet after movable E mark repetition activity.Movable F mark repeats activity Preceding collection TracePreSet is { B, C, D, E }, and it is { B, C, D, G } to collect TracePostSet after movable F mark repetition activity.Movable G Mark repeat activity before collection TracePreSet be { B, C, D, F }, movable G mark repeat to collect after activity TracePostSet be B, C,D,E,H}。
S3.3 excavates free single cycle structure
For distinguishing the activity in each free single cycle structure, here in the presence of two kinds of situations.
(1) firstly, for be present in same length be 2 free single cycle structure two activities, incipient activity Collection collects more activities, the i.e. direct cause and effect activity of a forerunner in the presence of the incipient activity than thereafter before mark repeats activity, there is Apr →WTstart.Likewise, the mark of termination activity collects than the more activities of collection before it after repeating activity, i.e., the termination activity is follow-up Direct cause and effect activity Apo, there is Tstart→WApo.Therefore, collection belongs to Qian Ji after a movable mark repeats activity, illustrates the work Dynamic is the T in the free single cycle structure that a length is 2startIncipient activity in corresponding daily record, sort of activity is put into Set Length2TstartSet comprising sort of activity.If collection belongs to Hou Ji before the movable mark repeats activity, illustrate this Activity is the T in the free single cycle structure that a length is 2endTermination activity in corresponding daily record, sort of activity is put into The set Length2TendSet of sort of activity.For each incipient activity in Length2TstartSet, travel through in daily record Each mark, there will necessarily be in Length2TendSet a termination activity all the time with the incipient activity occurrence number phase Together, illustrating this incipient activity and this termination activity is present among the free single cycle structure that same length is 2.
(2) gather not proper subclass each other before and after a movable mark revolving-door, illustrate that the activity is present in length not For in 2 free single cycle structure, these activities are saved in into other repetition active set NoLength2ActivitySet. For the activity in NoLength2ActivitySet, it is in if there is two activities in same free single cycle structure, this Two movable marks repeat the preceding collection common factor of activity and rear collection common factor should be identical, i.e., comprising other all and free single cycles The activity of parallelism structural is all empty set.Therefore, judge that the mark between each two activity repeats collection common factor and mark weight before activity Bring back to life to collect after moving and occur simultaneously, if the common factor of preceding collection is identical with the common factor integrated afterwards or all as empty set, illustrate that the two activities belong to Same free single cycle structure.All activities in same free single cycle structure are stored in set in the form of unordered In Set.It is more than 2 free single cycle structure for each length, finds movable order therein, it is necessary to travel through in daily record Mark.When there is the activity in the structure in the presence of a mark, the order occurred according to movable first pass is to each in Set Activity sequence, obtain the structurally ordered free single cycle sequence Tset。
Algorithm 4 is excavated according to above-mentioned two situations to all free single cycle structures, and it is comprised the following steps that：
Step 1:Free single cycle structure set FreeSingleLoopSet is initialized as empty set, and length is 2 initial live Dynamic set Length2TstartSet is initialized as empty set, and the termination active set Length2TendSet that length is 2 is initialized For empty set, length is not initialized as empty set for 2 repetition active set NoLength2ActivitySet；
Step 2:For repeating each movable a in active set RepetitiveActivitySetr, perform following grasp Make：
Step 3:If 1) movable arMark repeat activity after collect ar.TracePostSet collected before its mark repeats activity ar.TracePreSet proper subclass, then by arAdd the initial active set Length2TstartSet that length is 2；
If 2) movable arMark repeat activity before collection ar.TracePreSet collected after its mark repeats activity ar.TracePostSet proper subclass, then by arAdd the termination active set Length2TendSet that length is 2；
If 3) movable arMark repeat activity before collection ar.TracePreSet collect after repeating activity with its mark ar.TracePostSet not proper subclass each other, then by arAdd the repetition active set that length is not 2 NoLength2ActivitySet；
Repeat step 3, all movable a in RepetitiveActivitySetrPerform end of operation；
Step 4:Each the movable a in initial active set Length2TstartSet for being 2 for lengthsPerform with Lower operation：
Step 5:One free single cycle sequence TsetIt is initialized as sky；
Step 6:If in the presence of a movable bsBelong to the termination active set Length2TendSet that length is 2, for certainly By the complete workflow logs W of single cyclefslIn each mark σ, movable asWith movable bsOccurrence number phase in current mark σ Together, then in the presence of a free single cycle sequence TsetFor<as,bs>；
Step 7:By free single cycle sequence Tset=<as,bs>Add free single cycle structure set FreeSingleLoopSet；
Step 8:It is not each activity in 2 repetition active set NoLength2ActivitySet for length at, perform following operate：
Step 9:Unordered set Set is initialized as { at, a free single cycle sequence TsetIt is initialized as sky；
Step 10:For each movable btBelong to the repetition active set that length is not 2 NoLength2ActivitySet and btNot equal to at, perform following operate：
Step 11:If movable atMark repeat activity before collection at.TracePreSet with movable btMark repeat activity before collect bt.TracePreSet common factor is equal to movable atMark repeat activity after collect at.TracePostSet with movable btMark repeat Collect b after activityt.TracePostSet common factor, then by btAdd unordered set Set；
Repeat step 11, it is all in NoLength2ActivitySet to be different from movable atActivity perform operation Finish；
Step 12:By each activity in unordered set Set according to this in presence these movable mark σ in daily record The order sequence that a little activities occur first, obtains a free single cycle sequence Tset；
Step 13:By free single cycle sequence TsetAdd free single cycle structure set FreeSingleLoopSet；
Repeat step 9 arrives step 13, all activity execution end of operations in NoLength2ActivitySet；
Step 14:Export free single cycle structure set FreeSingleLoopSet.
Algorithm 4 will repeat repetition activity all in active set RepetitiveActivitySet according to different feelings Condition is divided into the unrestricted choice structure being individually present, and arrangement obtains respective free single cycle sequence T in orderset, and most It is saved in eventually in all free single cycle structure set FreeSingleLoopSet.Part I, which is distinguished, is present in length for 2 Free single cycle structure in activity, the step 2 of algorithm 4 to step 3 judges the free single cycle structure that all length is 2 TstartCorresponding incipient activity and TendCorresponding termination activity, and these activities are respectively put into set Length2TstartSet and set Length2TendSet.Each mark in daily record is traveled through, to each incipient activity, is looked for To corresponding termination activity.This group of incipient activity and termination activity are sequentially added to a free single cycle sequence Tset In, finally it is added in FreeSingleLoopSet.Part II is distinguished in the free single cycle structure in the presence of other length Activity.Step 3 in algorithm, other activities are added in set NoLength2ActivitySet.By to each two activity Mark repeat collection before activity occur simultaneously and rear collection occur simultaneously whether the identical or judged result that is all not present, will be present in same freedom Activity in single cycle structure makes a distinction, and is added in set Set.All it is a unordered collection for each set Set Close.Step 12 in algorithm 4 arrives step 13, to the activity in each Set, contains these activities in the presence of a mark in daily record. The activity in Set is sequentially added to orderly free single cycle sequence according to the order that these activities occur first in this mark TsetIn, finally it is added in FreeSingleLoopSet.
For the N in previously mentioned example Fig. 32And its daily record W2, the loop structure that length therein is 2 is found first In activity, each mark for repeating activity combined according to the circulation obtained before in RepetitiveActivitySet repeats Collection TracePreSet and mark collect TracePostSet after repeating activity before movable, it can be found that mobile CTherefore C belongs to T in the loop structure that some length is 2startCorresponding starting is lived It is dynamic, add set Length2TstartSet.It can be found that movable DTherefore D belongs to certain The T for the loop structure that one length is 2endCorresponding termination activity, add set Length2TendSet.Travel through in daily record Each, it is found that mobile C is identical with the number that movable D occurs in every mark.Therefore, mobile C belongs to same length with movable D The free single cycle structure for 2 is spent, by Tset=<C,D>Add all free single cycle structure set FreeSingleLoopSet.For remaining movable B, E, F, G, add other and repeat active set NoLength2ActivitySet.Mark by judging each two activity collects common factor and rear collection common factor, Ke Yifa before repeating activity Existing movable B and the preceding collection of other activities occur simultaneously and rear collection common factor is different from, therefore movable B oneself belongs to a free single cycle Structure, by Tset=<B>Add all free single cycle structure set FreeSingleLoopSet.For movable E, F, G two-by-two Between mark repeat collection before activity occur simultaneously and rear collection to occur simultaneously all be { B, C, D }, therefore movable E, F, G belong to same freedom and singly followed Ring structure, it is added in a unordered set Set.For each length be more than 2 Set, any one in daily record Comprising mark movable in the free single cycle structure, such as { ABECFDGH }, can obtain these movable ordered arrangements from By single cycle sequence Tset=<E,F,G>, and it is added into all free single cycle structure set FreeSingleLoopSet.Extremely This, all free single cycle structures are all found in daily record, all free single cycle structure set FreeSingleLoopSet=<C,D>,<B>,<E,F,G>}.
S4 establishes the orbution of log activity
It is therein for each free single cycle structure in free single cycle structure set FreeSingleLoopSet Orbution between activity needs to redefine according to the definition of free single cycle relation.
Algorithm 5 realizes redefining for orbution between free single cycle structure internal activity, and its specific steps is such as Under：
Step 1:Active sequence set of relationship R is initialized as the active sequence set of relationship R exported in algorithm 1；
Step 2:For all free single cycle sequences belonged in free single cycle structure set FreeSingleLoopSet Arrange Tset=<t1,t2,t3,…,tn>, for TsetIn each movable ti, from 1 to n-1, execution is following to be operated i values：
Step 3:1) if n is equal to 1, by relation t1 t1Add active sequence set of relationship R；
Step 4:If 2) n is not equal to 1, following operate is performed：
Step 5:By relation ti ti+1Add active sequence set of relationship R；
Repeat step 5, until i is equal to n-1, and by relation tk t1Add active sequence set of relationship R；
Step 6:Export active sequence set of relationship R.
Initial order relation between each free single cycle structure internal activity is newly defined as freely singly following by algorithm 5 Ring relation.The step 3 of algorithm 5 uses t for the free single cycle structure that length is 11 t1Instead of initial order relation t1#W t1This mark for representing exclusive relation.The step 4 of algorithm 5 is more than 1 free single cycle knot to step 5 for length Structure, according to free single cycle sequence TsetIn order, use ti ti+1Relation form represent TsetIn front and rear adjacent two Individual activity, for TendCorresponding termination activity and TstartCorresponding initial active, free single cycle relation T be presentend Tstart.According to the N in example Fig. 3 before2And its daily record W2And obtained free single cycle structure set FreeSingleLoopSet, for the activity in each free single cycle, the relation between the activity of successive is changed intoThat is BB,CD,DC,EF,FG,GE。
For N2And its daily record W2Footprint matrix is as shown in table 1 corresponding to initial active sequence：
The W of table 12Initial footprint matrix
A B C D E F G H
A # → → # → # # →
B ← # || || || || || →
C ← || # || || || || #
D # || || # || || || →
E ← || || || # → ← #
F # || || || ← # → #
G # || || || → ← # →
H ← ← # ← # # ← #
Footprint matrix comprising free single cycle relation is as shown in table 2：
The W of table 22The footprint matrix for including free single cycle relation
The α excavated to free single cycle structure of the present inventionfslAlgorithm is realized based on α algorithms.
Daily record is pre-processed first, finds repeatedly activity present in it, and then be sequentially stored in activity is repeated In respective free single cycle structure.By being redefined to order between free single cycle structure internal activity, in footprint Preferably represent each free single cycle structure be present in daily record in matrix.
Petri network is excavated according to α algorithms according to the active sequence relation beyond the relation inside circulation.Analyze loop structure In activity circulation external activity between orbution find, each free single cycle structure exist a forerunner it is straight Meet cause and effect activity Apr', there is Apr′→WTstart, and a follow-up directly cause and effect activity Apo', there is Tend→WApo', find movable Apr′ And Apo', free single cycle structure is inserted into corresponding transition Apr' and transition Apo' between place.According to the method in α On the basis of algorithm excavates to model, the free single cycle structure of discovery is added, finally gives correct model.
In addition, the present invention also passes through the feasibility and analysis result of following experimental verification said process method for digging.
By taking the operation flow that certain computer maintenance company repairs on the net as an example, the computer of client breaks down, and applies for maintenance industry It is engaged in successfully including a series of activity to repairing.Client needs to apply for maintenance (Application) first, is then carried out on the net Maintenance registration (Register).After being registrated successfully, the equipment of oneself is sent to maintenance computer by client by the form of mailing Company (Delivery).Computer maintenance company will carry out fault detect (Malfunction after receiving the equipment of client Detection), equipment fault generally comprises hardware failure (Hardware Malfunction) and software breaks down (Software Malfunction), can two kinds of failures all exist, can also only exist first, trouble-free feelings even be present The problem of the problem of condition, i.e. client's use environment rather than equipment itself.For the hardware fault of the equipment detected, maintenance personal Hardware maintenance (Hardware Fix) is carried out, same will also carry out software maintenance (Software for software fault Fix), the process that detection is out of order and repaired may repeatedly occur, it is also possible to not occur.All breakdown maintenances finish, relevant people Member is counted (Expense Statistics) to maintenance cost, and the equipment packing of client is sent back into client (Return).Client is after the equipment of return is received, and carry out an acceptance inspection (Check), needs to pay maintenance cost after the completion of examination (Pay).Last computer maintenance company will carry out regular return visit to client, track the equipment service condition of client in a short time (Follow Up)。
For the orchestration instance, definition of the present invention according to free single cycle Complete Log above, hand simulation Three groups meet the complete controlling stream daily record of free single cycle, as indicated at 3.In these daily records, activity is repeated not comprising any Situation about continuously occurring, i.e. σ<…a,a,a…>Or σ<…a,b,a,b…>This kind of situation.
The free single cycle Complete Log of 3 groups of table
The algorithm realized in sourced processes excavate framework ProM as plug-in unit, it using the daily record of XES forms as Input, using the process model that WF-net is represented as output.
The daily record obtained above is excavated to obtain model with AlphaFSLMiner plug-in units and sees Fig. 7.
Because α algorithms do not support the excavation to short loop structure, therefore as a comparison, have chosen can excavate short follow respectively The α of ring+Algorithm, ILP algorithms and IM algorithms excavate to same daily record, and the result of mining model is shown in Fig. 8, Fig. 9 and Figure 10.
According to the operation flow of computer maintenance company, it can be found that it is to exist that hardware detection, which is out of order with hardware maintenance, simultaneously , it is simultaneous that same software detection, which is out of order with hardware maintenance,.Because an equipment is made up of multiple hardware, hardware goes out The situation of existing failure has a lot, to detect and repair one by one.Therefore, hardware detection be out of order and hardware maintenance the two work Dynamic to be likely to occur repeatedly, same software detection is out of order and software maintenance also occurs repeatedly.For an equipment, or even meeting There is the no failure i.e. situation of client's error in judgement, such case is not in the work of hardware and software failure and software and hardware maintenance It is dynamic.
For several situations mentioned above, αfslAlgorithm can excavate the Work flow model for meeting these situations.Pass through figure 7 as can be seen that hardware detection be out of order (Hardware Malfunction) be present in hardware maintenance (Hardware Fix) In same free single cycle structure, illustrate to find a kind of failure of hardware every time, it is repaired.Occur for software Failure (Software Malfunction) and software maintenance (Software Fix) are equally applicable.Hardware fault and hardware The free single cycle structure and the free single cycle structure of software fault and software maintenance composition for repairing composition are in parallel junction In structure, the detection and maintenance that illustrate software and hardware can occur simultaneously, one of which can also only occur, it might even be possible to do not send out It is raw.
For α+Algorithm, ILP algorithms and IM algorithms, it is corresponding that these three algorithms can not correctly excavate the operation flow Model.α+Algorithm and ILP algorithms excavate mistake in itself be present in model, hardware detection is out of order (Hardware Malfunction) and software break down (Software Malfunction) the two change no follow-up places, hardware dimension (Hardware Fix) and the two transition of software maintenance (Software Fix) are repaiied without preceding after place, also do not show this Should existing free single cycle structure.In the model that IM algorithms excavate, do not show hardware fault and hardware maintenance is inevitable Situation about occurring simultaneously, software fault and software maintenance are also such.Figure 11 and Figure 12 is respectively from each algorithm under different daily records The degree of fitting of the model excavated and the performance of accuracy give experimental result.
From Figure 11 it can be found that for each daily record, αfslThe model-fitting degree that algorithm and IM algorithms excavate it is identical and And it is very high, and α+The degree of fitting for the model that algorithm and ILP algorithms obtain is relatively low.Although it can be found that α from Figure 12fslAlgorithm is dug Pick obtains model as the accuracy that increases of mark bar number in daily record slightly reduces, but remains in higher level.By contrast, IM The model accuracy that algorithm excavates is relatively low, and α+The accuracy for the model that algorithm and ILP algorithms obtain is ineffective.
Certainly, described above is only presently preferred embodiments of the present invention, and the present invention is not limited to enumerate above-described embodiment, should When explanation, any those skilled in the art are all equivalent substitutes for being made, bright under the teaching of this specification Aobvious variant, all falls within the essential scope of this specification, ought to be protected by the present invention.

Claims (1)

  1. A kind of 1. process model mining method of free single cycle structure, it is characterised in that comprise the following steps：
    The definition of s1 basic conceptions
    Define multiple set：
    If S is a set, the multiple set on a S is defined as a function Z:B (S) represents all more on S Collect the set being combined into again, a multiple set by a vector representation,Represent a natural number set；
    Defined nucleotide sequence：
    If S is a set, S* is the set of all finite sequence compositions on S；
    Define mark：
    If A is active set, a mark σ ∈ A* is a movable queue；
    Define event log：
    Event log L is mark σ multiple set, is designated as L ∈ B (A*)；
    Define Petri network：
    Four-tuple PN=(P, T；F, M) it is referred to as a Petri network, and if only if：
    (1) N=(P, T；F) it is a net；Wherein：P is a limited set of library, and T is active set in a corresponding event log The limited transition collection closed,It is a finite arc collection；
    (2)M:P → N is referred to as a mark for netting PN, wherein, miFor initial marking, mfIdentified to terminate；
    (3) there are PN following transition rule occurs：
    (a) for changing t ∈ T, ifThen claim transition t to be enabled in the case where identifying M, be designated as M [t>；
    If (b) M [t>, then in the case where identifying M, transition t can occur, and trigger transition t to obtain a new mark M ' from mark M, remember For M [t>M ', and it is right
    Wherein, M (p) represents mark place p at, and M ' (p) represents to trigger the mark changed after t at place p,·T represents t Input set, t·Represent t output collection；
    Define Workflow net：
    Hexa-atomic group of WFN=(P, T；F,M,in,on) being referred to as a Workflow net, and if only if：
    (1) Σ=(P, T；F, M) it is a Petri network；
    (2) an initial place i be presentn∈ P, have
    (3) a terminate storehouse institute o be presentn∈ P, have
    (4) forY is at one from inTo onPath on；
    Wherein,·inRepresent inInput set,·onRepresent onOutput collection；
    Definition mining algorithm α：
    W is made to represent the workflow logs for covering limited transition collection T；α (W) is defined as follows：
    (1)
    (2)
    (3)
    (4)
    (5)
    (6)PW={ p (AW,BW)|(AW,BW)∈YW}∪{iW,oW}；
    (7)FW={ (aW,p(AW,BW))|(AW,BW)∈YW∧aW∈AW}∪{(p(AW,BW),bW)|(AW,BW)∈YW∧bW∈BW} ∪{(iW,t)|t∈TI}∪{(t,oW)|t∈TO}；
    (8) α (W)=(PW,TW,FW)；
    Wherein, TWIt is the transition in Workflow net corresponding to the activity occurred in daily record；
    TIIt is the set to come into play, i.e., all movable set of first position is appeared in mark；
    TOIt is the set of ending activity, i.e., all movable set of last position is appeared in mark；
    p(AW,BW) it is a place, wherein, AWIt is p (AW,BW) input transition set,·p(AW,BW)=AW；BWIt is p (AW,BW) Output transition set, p (AW,BW)·=BW, all A for meeting the requirement of (4) stepW, BWX is combined into the collection of compositionW；
    For XWEvery a pair of (AW,BW), nonempty setAnd nonempty setThere is (AW′,BW′)∈XW, In algorithm (4) step, all is not that maximum set is all removed, and obtained collection is combined into YW；
    PWInclude a unique source place iWWith a unique terminate storehouse institute oW；
    The set F of the arc of (7) step generation Workflow netW, TIIn all changes move the capital to another place with iWAs input place, TOIn it is all Change is moved the capital to another place with OWAs output place；
    All place p (AW,BW) all with AWAs input node, with BWAs output node；
    The result finally given is Petri network α (W)=(PW,TW,FW), it describes the behavior that daily record W is reflected；
    First (σ) represents first activity in mark σ, and last (σ) represents last activity in mark σ；
    S2 defines free single cycle structure
    Define free single cycle structure：
    A WFN free single cycle structure is shown as FSL=by triple table<Tstart,Tend,Tset>；
    Wherein：Tstart、Tend∈ T and Tset∈ T*, T* represent the sequence of all activity compositions in limited transition collection T；
    So that Tstart∈p0 ·And Tend∈·p0；
    Sequence SE=be present<p0,Tstart,p1,…,pn,Tend,p0>, there is Tset=<Tstart,…,Tend>；
    For a free single cycle structure：
    TstartFirst transition in structure are represented, i.e., are changed corresponding to incipient activity of the structure in daily record；TendRepresent knot Last transition in structure, i.e., change corresponding to termination activity of the structure in daily record；TsetRepresent all in the structure The sequence that activity in daily record corresponding to transition forms in order, i.e., free single cycle sequence；
    The order of the sequence is each movable order when the loop structure first pass occurs in the mark of daily record；
    It is defined for the completeness of the daily record containing free single cycle structure：
    Define free single cycle Complete Log：
    Make WfslIt is the complete workflow logs of a free single cycle, A is active set, and and if only ifAnd σ1、σ2∈ WfslSo that a ∈ σ1, σ1(a) >=2 and
    The activity for continuously repeating appearance is all not present for each mark in the daily record, but the work that occurrence number is more than 1 be present It is dynamic；
    For the relation between the activity in daily record, free single cycle relation is introducedTo represent to be in same free single cycle The relation between front and rear adjacent activity in structure；
    Definition includes the active sequence relation of free single cycle relation：
    Make WfslIt is the complete workflow logs of a free single cycle, A is active set, and T is limited transition collection corresponding to A, Wfsl∈P(T*)；Make a, b ∈ T：
    - a ＞Wb：And if only if exist a mark σ=<a1,a 2,a 3,…,a n>, i ∈ { 1 ..., n-1 } are so that σ ∈ Wfsl, ai=a and ai+1=b；
    —a→Wb：And if only if a ＞WB and b ≯Wa；
    —a#Wb：And if only if a ≯WB and b ≯Wa；
    —a||Wb：And if only if a ＞WB and b ＞Wa；
    —aa：And if only if, and a is present in the free single cycle structure that length is 1；
    —ab：And if only if has the free single cycle sequence T that a length is nset=<t1,t2,t3,…,tn>, meet Tset∈T*,ti=a and ti+1=b；t1,t2,t3,…,tnRepresent TsetIn activity；
    Daily record WfslIt is the multiple set of the mark of transition composition, meets that free single cycle is complete；A, b are limited transition set In two transition corresponding to activity；＞WExpression follows relation, →WRepresent causality, #WExclusive relation is represented, | |WRepresent Parallel relation,Represent free single cycle relation；
    Before free single cycle structure is excavated, initial order relation movable in daily record is obtained according to algorithm 1, wherein not including Free single cycle relation；R represents the movable initial order set of relationship obtained based on daily record；
    Algorithm 1 lists initial order relation movable in daily record, and it is comprised the following steps that：
    Step 1:Active sequence set of relationship R is initialized as empty set；
    Step 2:Belong to the complete workflow logs W of free single cycle for allfslMark σ, perform following operate：
    Step 3:For the movable a in the mark σ of a n activityi, from 1 to n, execution is following to be operated i values：
    Step 4:By ai＞W ai+1Add active sequence set of relationship R；
    Repeat step 4, until i is equal to n；
    Repeat step 3, until having traveled through all mark σ；
    Step 5:For each element in active sequence set of relationship R, following operate is performed：
    Each element in wherein R is one group of relation being made up of two parameters, and movable a is the previous parameter of relation, living Dynamic b is the latter parameter of relation；
    Step 6:If 1) relation a ＞ be presentWB and b ≯WA, then by relation a →WB adds active sequence set of relationship R；
    If 2) relation a ＞ be presentWB and b ＞WA, then by relation a | |WB adds active sequence set of relationship R；
    If 3) exist relation a ≯WB and b ≯WA, then by relation a#WB adds active sequence set of relationship R；
    Repeat step 6, all elements perform end of operation in R；
    Step 7:Export active sequence set of relationship R；
    The free single cycle structure mining algorithms of s3
    S3.1 finds the activity of repetition
    According to the definition of free single cycle structure, all occurrence numbers are more than 1 activity in a mark of daily record, belong to certainly By the activity included in single cycle structure；Definition repeats active set to deposit these activities：
    Definition repeats active set RepetitiveActivitySet：
    If A is active set, WfslFor the complete workflow logs of free single cycle,Repeat active setAnd if only if by A ∈ repeats active set,So that ar∈ σ, and σ (ar)≥2；
    Algorithm 2 finds the activity of repetition and adds all repetition activities and repeats in active set, and it is comprised the following steps that：
    Step 1:Repeat active set RepetitiveActivitySet and be initialized as empty set；
    Step 2:Belong to the complete workflow logs W of free single cycle for allfslMark σ, perform following operate：
    Step 3:For the movable a in a mark σ comprising n activityi,aj, from 1 to j, j values perform i values from i to n-1 Operate below：
    Step 4:If aiWith ajEqual and aiIt is not belonging to repeat active set RepetitiveActivitySet, then by aiAdd Repeat active set RepetitiveActivitySet；
    Repeat step 4, until i is equal to n-1；
    Repeat step 3, until having traveled through all mark σ；
    Step 5:Output repeats active set RepetitiveActivitySet；
    S3.2 has found to repeat forerunner's activity of activity and succeeding activity
    The activity repeated in active set is not divided into clearly among each free single cycle structure；
    Which is present in obtained repetition activity below by the movable forerunner's activity of mark repetition and succeeding activity is defined Individual free single cycle structure makes a distinction offer support；
    Define the movable forerunner's activity of mark repetition and succeeding activity：
    If A is active set, WfslIt is the complete workflow logs of free single cycle, RepetitiveActivitySet is Wfsl Repetition active set；
    ForSo that ar∈ σ, there is ai=ar, 1 ＜ i ＜ N, then claim ai-1For arForerunner activity, ai+1For arSucceeding activity；a1,a2,…,anRepresent the activity in mark σ；
    For any activity in repetition active set, in each mark of daily record, all movable forerunners are movable forms the work Dynamic mark collects before repeating activity, and all appear in after the movable succeeding activity forms movable mark repetition activity collects；
    The activity that algorithm 3 repeats for each, collect after obtaining the movable preceding collection of mark repetition activity and mark repetition activity, specifically Step is as follows：
    Step 1:For repeating each movable a in active set RepetitiveActivitySetr, perform following operate：
    Step 2:For the complete workflow logs W of the free single cycle of daily recordfslIn each mark σ, perform following operate：
    Step 3:For the movable a in the mark σ comprising n activityi, from 2 to n-1, execution is following to be operated i values：
    Step 4:If 1) aiWith arIt is equal, and ai-1It is not belonging to arMark repeat activity before collect, then by ai-1Add arMark repeat Collection a before movabler.TracePreSet；
    If 2) aiWith arIt is equal, and ai+1It is not belonging to arMark repeat activity after collect, then by ai+1Add arMark repeat activity after Collect ar.TracePostSet；
    Repeat step 4, until i is equal to n-1；
    Repeat step 3, until having traveled through all mark σ；
    Repeat step 2, all movable a in RepetitiveActivitySetrPerform end of operation；
    Step 5:The each movable a of outputrMark repeat activity before collection ar.TracePreSet collect after repeating activity with mark ar.TracePostSet；
    Movable mark by obtaining each repetition repeats to collect after the preceding collection of activity repeats activity with mark, to distinguish which it is present in One free single cycle structure is prepared；
    S3.3 excavates free single cycle structure
    Algorithm 4 is excavated to all free single cycle structures, and it is comprised the following steps that：
    Step 1:Free single cycle structure set FreeSingleLoopSet is initialized as empty set, and length is 2 initial active collection Close Length2TstartSet and be initialized as empty set, the termination active set Length2TendSet that length is 2 is initialized as sky Collection, length are not initialized as empty set for 2 repetition active set NoLength2ActivitySet, other length cycles set OtherLoopSet is initialized as empty set；
    Step 2:For repeating each movable a in active set RepetitiveActivitySetr, perform following operate：
    Step 3:If 1) movable arMark repeat activity after collect ar.TracePostSet collected before its mark repeats activity ar.TracePreSet proper subclass, then by arAdd the initial active set Length2TstartSet that length is 2；
    If 2) movable arMark repeat activity before collection ar.TracePreSet it is to collect a after its mark repeats activityr.TracePostSet Proper subclass, then by arAdd the termination active set Length2TendSet that length is 2；
    If 3) movable arMark repeat activity before collection ar.TracePreSet a is collected after repeating activity with its markr.TracePostSet Not proper subclass each other, then by arAdd the repetition active set NoLength2ActivitySet that length is not 2；
    Repeat step 3, all movable a in RepetitiveActivitySetrPerform end of operation；
    Step 4:Each the movable a in initial active set Length2TstartSet for being 2 for lengthsPerform following grasp Make：
    Step 5:One free single cycle sequence TsetIt is initialized as sky；
    Step 6:If in the presence of a movable bsBelong to the termination active set Length2TendSet that length is 2, for freely singly following The complete workflow logs W of ringfslIn each mark σ, movable asWith movable bsOccurrence number in current mark σ is identical, then In the presence of a free single cycle sequence TsetFor<as,bs>；
    Step 7:By free single cycle sequence Tset=<as,bs>Add free single cycle structure set FreeSingleLoopSet；
    Step 8:It is not each movable a in 2 repetition active set NoLength2ActivitySet for lengtht, perform Operate below：
    Step 9:Unordered set Set is initialized as { at, a free single cycle sequence TsetIt is initialized as sky；
    Step 10:For each movable btBelong to length for 2 repetition active set NoLength2ActivitySet and btNot equal to at, perform following operate：
    Step 11:If movable atMark repeat activity before collection at.TracePreSet with movable btMark repeat activity before collect bt.TracePreSet common factor is equal to movable atMark repeat activity after collect at.TracePostSet with movable btMark repeat Collect b after activityt.TracePostSet common factor, then by btAdd unordered set Set；
    Repeat step 11, it is all in NoLength2ActivitySet to be different from movable atActivity perform end of operation；
    Step 12:Each activity in unordered set Set is lived according to these in presence these movable mark σ in daily record The dynamic order sequence occurred first, obtains a free single cycle sequence Tset；
    Step 13:By free single cycle sequence TsetAdd free single cycle structure set FreeSingleLoopSet；
    Repeat step 9 arrives step 13, all activity execution end of operations in NoLength2ActivitySet；
    Step 14:Export free single cycle structure set FreeSingleLoopSet；
    S4 establishes the orbution of log activity
    For each free single cycle structure in free single cycle structure set, it is therein activity between orbution according to The definition of free single cycle relation needs to redefine；
    Algorithm 5 redefines the orbution between free single cycle structure internal activity, and it is comprised the following steps that：
    Step 1:Active sequence set of relationship R is initialized as the active sequence set of relationship R exported in algorithm 1；
    Step 2:For all free single cycle sequences belonged in free single cycle structure set FreeSingleLoopSet Tset=<t1,t2,t3,…,tn>, for TsetIn each movable ti, from 1 to n-1, execution is following to be operated i values：
    Step 3:1) if n is equal to 1, by relation t1 t1Add active sequence set of relationship R；
    Step 4:If 2) n is not equal to 1, following operate is performed：
    Step 5:By relation ti ti+1Add active sequence set of relationship R；
    Repeat step 5, until i is equal to n-1, and by relation tk t1Add active sequence set of relationship R；
    Step 6:Export active sequence set of relationship R.
CN201710635987.8A 2017-07-31 2017-07-31 A kind of process model mining method of free single cycle structure Pending CN107391715A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710635987.8A CN107391715A (en) 2017-07-31 2017-07-31 A kind of process model mining method of free single cycle structure

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710635987.8A CN107391715A (en) 2017-07-31 2017-07-31 A kind of process model mining method of free single cycle structure

Publications (1)

Publication Number Publication Date
CN107391715A true CN107391715A (en) 2017-11-24

Family

ID=60342811

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710635987.8A Pending CN107391715A (en) 2017-07-31 2017-07-31 A kind of process model mining method of free single cycle structure

Country Status (1)

Country Link
CN (1) CN107391715A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110084517A (en) * 2019-04-28 2019-08-02 西安理工大学 A kind of workflow method of oriented towards education data logging

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100017168A1 (en) * 2006-07-03 2010-01-21 Inchron Gmbh Method of testing the real-time capability of a system
CN105095491A (en) * 2015-08-18 2015-11-25 山东科技大学 Process model repair method based on Petri net basic structures

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100017168A1 (en) * 2006-07-03 2010-01-21 Inchron Gmbh Method of testing the real-time capability of a system
CN105095491A (en) * 2015-08-18 2015-11-25 山东科技大学 Process model repair method based on Petri net basic structures

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
杜玉越 等: "一种基于逻辑Petri网的过程挖掘方法", 《电子学报》 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110084517A (en) * 2019-04-28 2019-08-02 西安理工大学 A kind of workflow method of oriented towards education data logging
CN110084517B (en) * 2019-04-28 2024-01-30 北京第一因科技有限公司 Workflow method for educational data log

Similar Documents

Publication Publication Date Title
Moreira et al. Polynomial time verification of decentralized diagnosability of discrete event systems
Shakeri et al. Sequential testing algorithms for multiple fault diagnosis
CN113361136B (en) Diagnostic verification method for automatic manufacturing system based on Petri network
Viana et al. Codiagnosability analysis of discrete-event systems modeled by weighted automata
Ye et al. Competing failure modeling for performance analysis of automated manufacturing systems with serial structures and imperfect quality inspection
Deng et al. State-based decentralized diagnosis of bi-fuzzy discrete event systems
Yang et al. Improve the effectiveness of test case generation on efsm via automatic path feasibility analysis
CN109936479A (en) Control plane failure diagnostic system and its implementation based on Differential Detection
Li et al. On–the–fly diagnosability analysis of bounded and unbounded labeled petri nets using verifier nets
US20100235143A1 (en) Method for the generation of analytical redundancy relations for system diagnostics
CN110333694B (en) Fuzzy Petri network-based numerical control equipment fault diagnosis method
CN107391715A (en) A kind of process model mining method of free single cycle structure
Aleem et al. Business process mining approaches: a relative comparison
Cabasino et al. A comparison among tools for the diagnosability of discrete event systems
Kang et al. CPS-based fault-tolerance method for smart factories: Simulation-based framework and fault management application
Luo et al. A new algorithm for testing diagnosability of fuzzy discrete event systems
de MC Machado et al. Event-based automaton model for identification of discrete-event systems for fault detection
Desrochers et al. Complex-valued token Petri nets
Esgin et al. Process Profiling based Synthetic Event Log Generation.
CN109685453B (en) Method for intelligently identifying effective paths of workflow
CN112784447B (en) Nuclear power plant accident modeling method for DET and RELAP5 program dynamic coupling framework
Dutta et al. Failure diagnosis in real time stochastic discrete event systems
Yi et al. Novel Approach to Discover Precise Process Model by Filtering out Log Chaotic Activities
Bouché et al. Improving the performance of production lines with an expert system using a stochastic approach
Gascard et al. A polynomial-time algorithm for diagnosability verification of discrete event systems

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication

Application publication date: 20171124

RJ01 Rejection of invention patent application after publication