CN109509547B - Process model correction method for selecting nested concurrency - Google Patents

Abstract

The invention discloses a process model correction method for selecting nested concurrency, which aims to determine the position of deviation in the treatment process, find the places where the actual treatment process and the hospital published business process can be improved, provide the simplest calibration concept, and add an reachable library set to help determine the position of deviation in the treatment process. The diagnosis model obtained by correction can correctly reflect the actual behaviors, the problem of branch jumping in a selection structure is solved, and the relation between activities can be expressed more simply compared with the common Petri network due to the addition of the logical relation. The diagnosis model based on the logic Petri net correction greatly reduces the complexity of the model, so that the simplicity is higher. And the accuracy of the corrected model is higher because the self-loop heel invisible transition is not generated.

Description

Process model correction method for selecting nested concurrency

Technical Field

The invention relates to a method for modifying a process model by selecting nested concurrency.

Background

The goal of process mining is to establish an efficient correlation between the process model and the event logs generated from modern information systems. The process mining mainly comprises three applications: process discovery, consistency detection, process improvement. The discovery technique uses an event log generation model that does not include any a priori information. The consistency check is to compare a known process model with the event log generated by the process model and check whether the actual conditions recorded in the log are consistent with the model. Process improvement is the use of event logs generated by real-world activities to extend or improve an existing process.

Determining the quality of a process mining result needs to be described from the following 4 dimensions: degree of fit, simplicity, cleanliness, accuracy and generalization. Fitness refers to the ability of a model to allow behavior reflected by an event log to occur. The fit of this model is very good if the model can replay all the traces in the log from start to finish. The simple degree means that the model can explain the behavior in the log and is the simplest model. Accuracy refers to how relevant the resulting model is to the event log. Generalization refers to the behavior of the model not being limited to the examples seen in the log. A well-generalized model should not be a special model that only allows sample logs to occur, and models that are not generalized should be over-fitted.

Fahland et al propose a model restoration method, which collects non-fitting sub-logs by using the deviation between the calibration discovery model and the logs, excavates the sub-logs by means of an indectiveminer excavation algorithm, and adds the excavated sub-processes to the appropriate position of the original model as a whole. The model repaired by the repairing method has extremely high fitting degree, namely most traces in the event log can be replayed in the repaired model. However, since the sub-process is added to the model as a whole, the transition is repeated, the model becomes redundant and complex, and the repaired model is simple and has low accuracy.

The Goldratt method and knapack method are modified in such a way that individual active self-loops or invisible transitions are added according to different constraints. Intuitively, the model repaired by the three methods cannot correctly express the business process, and the accuracy and the simplicity of the repaired model are not high due to the addition of the self-loop and the invisible transition, so that the invention provides the simplest logical calibration to find the deviation position and tries to correct the model on the basis of keeping the basic structure of the original model.

The existing hospital business process has many disadvantages, for example, because the national medical insurance catalogs are different in different years, some medicines are newly brought into the national medical insurance in the pharmacy dispensing process, the patient cannot reimburse before, and the patient can reimburse at present. These flows occur to illustrate the case where there is a jump in the model from one branch of the selection branch to another. In addition, because the flow is not optimized, the existing flow model is not time-optimal, so that the congestion of the patient on some activities is caused, and meanwhile, some activities which should be carried out are in an idle state, so that the great waste of medical resources is caused, and the medical experience of the patient is not good. It can be seen that the existing hospital business processes need further improvement.

Disclosure of Invention

The invention aims to provide a process model correction method for selecting nested concurrency, so as to shorten the waiting time of a patient during treatment, enable the whole treatment process to be more efficient and improve the treatment experience of the patient.

In order to achieve the purpose, the invention adopts the following technical scheme:

the method for selecting the nested concurrent process model modification comprises the following steps:

defining a Petri Net

The quadruple PN (P, T; F, M) is called a Petri net, if and only if:

(P, T, F) is a net;

wherein, P is a finite library set used for representing various states in the clinic model, T is a finite transition set used for representing names of activities in the clinic flow,

is a finite arc set;

②

referred to as an identification of the net PN,

representing a set formed by all identification Petri nets;

③ PN has the following transition occurrence rules:

a. for T e.T, if

I.e. p contains tokken, t is said to be enabled under the M designation, denoted M [ t>Indicating that the activity is conditioned for occurrence;

b. if M [ t >, then under M, t can occur, deriving a new identifier M 'from M trigger t, denoted M [ t > M', indicating a transition from one activity to another in the visit procedure;

and is aligned with

All have:

wherein, t represents the front set of transition t, and t represents the rear set of transition t;

defining calibration

Is provided with

Is a set of activities, a represents the activity of a visit procedure,

a set representing a visit flow activity;

sigma epsilon is a trace on A, PN (P, T; F, M) is a Petri net on A;

the binary group (a, T) belongs to A > >. times T > > > >/{ (> > > > > > > > > > > > > > > > > >) } is a movement and represents a track performed in the treatment flow;

the calibration γ ∈ (a > > × T >) is a movement sequence (a, T) · between the trace σ and the model PN, and satisfies:

(1)π₁(γ)_↓Aσ, i.e. the movement sequence of the visit procedure generates the trajectory of the visit activity;

(2)

namely, the moving sequence of the model generates a complete visiting procedure sequence;

wherein > represents no movement, a > > > > > > < a { > > > > >;

π₁(γ)_↓Athe first row element in the calibration is represented, which is the activity the patient has undergone during the actual visit;

π₂(γ)_↓Tthe second row element in the calibration is represented, which is the activity passed through in the hospital visit procedure;

represents that no activity occurred in the flow;

m_ian input representing a sequence of visit model movements; m is_fAn output representing a sequence of visit model movements;

i represents a pre-input state, and f represents a post-output state;

for the doublet (a, t) ∈ γ in the calibration, the definition is as follows:

a) if a belongs to A and t ═ > >, the action is a log action, which means that some treatment activities occur in the real treatment process, but do not occur according to the hospital flow chart, which indicates that the existing hospital flow chart does not change along with the change of the requirements of the patients, and the existing hospital flow chart does not meet the real requirements of the patients;

b) if a is ═ and T belongs to T, the model action is taken, which means that some treatment activities do not occur in the real treatment process and occur according to the hospital flow chart;

c) if a belongs to A and T belongs to T, the operation is synchronous, and the flow chart of the representative hospital conforms to the real clinic process;

d) otherwise, the action is illegal;

defining a process tree

Is provided with

Given an operator set, τ is an implicit transition, then:

(ii) a ∈ A { [ τ } is a process tree;

② setting PT₁，…，PT_nIs a process tree, then

Is also a process tree, n>0；

Defining optimal calibration

Calibration gamma epsilon gamma between trace sigma and process model PN_σ,PNIs an optimal calibration, if and only if

Wherein lc (a, t) is a probability cost function and represents the deviation degree between the hospital flowchart and the real clinic process;

Γ_σ，PNrepresents the set of all calibrations between trace σ and model PN;

γ' represents other than the optimal calibration;

∑_(a，t)∈γlc ((a, t)) represents the set of least likely cost functions for optimal calibration;

∑_{(a，t)∈γ′}lc ((a, t)) represents the set of least likely cost functions for non-optimal calibration;

defining select recognition pairs and recognition pair sets

Let PN be (P, T; F, M), PT is the process tree of PN;

the selection recognition pair crp is a doublet (p)_left,p_right)，p_left＝·(LLN(K))，p_right(rln (k)), and satisfies:

and K ═ x;

wherein K represents a leaf node;

LLN (K) represents the leftmost leaf node of the selected structure in the visit model;

RLN (K) represents the rightmost leaf node of the selected structure in the visit model;

p_leftrepresenting the front set of leftmost leaf nodes, p_rightRepresenting the postset of the rightmost leaf node, x representing the product sign; selecting the set of identification pairs CRPS as a set comprising all identification pairs, there is:

defining a logical Petri Net

A logical Petri net is defined as a six-tuple LPN ═ P, T; F, I, O, M) if and only if:

(1) p is a finite set of libraries;

(2)T＝T_D∪T_I∪T_Ois a finite set of transitions that are,

if T is equal to T_I∩T_OThen, then

a.T_DRepresenting a set of transitions in a Petri net;

b.T_Irepresenting a set of logical input transitions, pair

t input library receives a logical expression f_I(t) a limit;

c.T_Orepresenting sets of logical output transitions, pair

t output library receives logical expression f_O(t) a limit;

(3) f ═ P × T (T × P) is a finite arc set;

(4) i is the mapping of the transitions from the logical inputs to the logical input function, for

(5) O is the mapping of the transition from the logical output to the logical output function, for

O(t)＝f_O(t)；

(6)M:P→Z₀Is a representation function of a Petri network, Z₀＝{0,1,2，…}；

Define simplest calibration

Let (a, T, H) e (a { > >) }) × (T { > > }) × (H) be an action;

simplest calibration β ═ a₁,t₁,h₁)…(a_|β|,t_|β|,h_|β|) Is an action queue, and satisfies:

(1)Γ_σ，PNis the set of all calibrations between trace σ and model PN;

(2)

namely, the action queue generated in the model is calibrated in the simplest way, and finally the termination identifier is reached;

(3) h is a Token reachable library set generated by transition triggering in the M state, and is called as a reachable library set;

wherein h represents a Tokenreachable library generated by transition triggering;

(a₁,t₁,h₁) Representing an initial action of the action queue, (a)_|β|,t_|β|,h_|β|) An end action representing an action queue;

π2(π₁(β)_↓A) ↓ T represents the second row element after the simplest calibration;

defining a logical simplest calibration

Logic simplest calibration LPN_β＝(a₁,t₁,h₁)…(a_|β|,t_|β|,h_|β|) Is an action queue, and satisfies:

(1)Γ_σ，LPNis the set of all calibrations between trace σ and model LPN;

(2)

namely:

the logic simplest calibrates an action queue generated in the model, and finally the termination identifier is reached;

(3) simplest calibration LPN between trace sigma and logical Petri Net model LPN_β∈Γ_σ,LPNIs a logical simplest calibration, if and only if

Wherein, | h_iI represents the number of reachable libraries;

| P | represents the number of states in the encounter model;

input representing a sequence of visit model movements under a logical Petri net;

representing the output of the visit model movement sequence under the logical Petri net;

the following method 1 is proposed for finding the selective identification pair of the selected structure in the hospital visit process, and the specific process is as follows:

(1) initializing the selection recognition pair crp (p)_left,p_right) And selecting a set of identification pairs CRPS;

(2) traversing all child nodes in the process tree;

(3) if the child node is "x";

(4) finding a leaf node under the child node;

(5) the following two elements of the selection recognition pair are obtained, namely:

the front set of the leftmost leaf node (lln (k)) and the back set of the rightmost leaf node (rln (k));

(6) the obtained selection recognition pair crp (p)_left,p_right) Merging into a selection recognition pair set CRPS;

the following method 2 is proposed for determining the deviation position of the treatment process, and the specific process is as follows:

(1) deviation position Dev (t) of initialization visit process_m,t_n)；

(2) Traversing the simplest calibration beta, and if log actions occur, performing the next step;

(3) when a missing Token library appears, recording the position of the activity at the moment;

(4) if a library carried "-" or at h appears in the reachable library set_iWhere the library appears multiple times, is denoted as p_kIndicates an abnormal libraryTherefore, the abnormal state appears at the moment;

(5) store the abnormal place p_kAssigning the previous set of values to the starting position of the deviation position, and assigning the abnormal position p to the abnormal position_kIs assigned to the end position of the deviation position and the end position of the deviation position cannot be the leftmost leaf node lln (k);

(6) deviation position Dev (t) of return visit procedure_m,t_n)；

Wherein, t_mIndicating the starting position of the deviation position, t_nAn end position indicating a deviation position;

the following method 3 is proposed for correcting the hospital visiting flow, determining the deviation position between the actual situation and the given hospital business flow, restoring the original hospital business flow at the deviation position, and adding the flow relationship, and the specific process is as follows:

(1) initializing a logical Petri network LPN, initializing an activity t of a selected branch_i；

(2) Calling the method 1 to obtain a selective recognition pair set CRPS;

(3) calling method 2 to obtain deviation position Dev (t) of the clinic flow_m,t_n)；

(4) Traversing the simplest calibration beta, and if log actions occur, performing the next step;

(5) if the occurrence condition is met:

starting position t of offset position_mIs p_leftAnd branch t is selected_iIs not the leftmost leaf node lln (k);

(6) add the postset from the leftmost leaf node (lln (k)) to the option branch t_iThe flow relationship of (a);

(7) if the occurrence condition is met: end position t of deviation position_nIs p_rightAnd t is_nNot the rightmost leaf node;

(8) then add the node from the rightmost leaf node rln (k) to the biased position end position t_nThe flow relationships of the preceding set of (2);

(9) mining a transition triggering condition by using an LPN mining algorithm to obtain a logic transition and a logic input and output function;

(10) returning the revised visit logic Petri network LPN;

the corrected hospital clinic procedure is a time-optimized business procedure model, represents the complete fitting of the actual clinic procedure and the clinic model, optimizes and selects the activity of the nested concurrent structure, and shortens the waiting time of the patient during clinic.

The invention has the following advantages:

in order to locate the deviation in the treatment process, the invention provides concepts of simplest calibration and logic simplest calibration, and determines the deviation position in the treatment process according to the relation between the reachable library set and the selection recognition pair set. And mining the triggering condition of the transition to obtain a logic transition function and a logic input and output function, and correcting the diagnosis model. The revised diagnosis model can correctly reflect the actual behaviors, the problem of branch jumping in the selection structure in the diagnosis model is solved, and the relation between activities can be expressed more simply compared with a common Petri network due to the addition of the logical relation. The diagnosis model based on the logic Petri net correction greatly reduces the complexity of the original model, so that the simplicity is higher. And the accuracy of the corrected model is higher because the self-loop heel invisible transition is not generated. The model is a time-optimized business process model, the complete fitting between the actual treatment process and the model can be realized, the whole model structure is clearer, the described model is more accurate, the activity of selecting the nested concurrent structure is optimized, the waiting time of the patient during treatment is shortened, the whole treatment process is more efficient, and the treatment experience of the patient is improved.

Drawings

FIG. 1 is a schematic diagram of a logical Petri net model.

FIG. 2 is a Petri net model N for selecting nested concurrent structures₁Schematic representation.

FIG. 3 shows trace σ₁And model N₁Schematic diagram of the simplest calibration.

FIG. 4 shows trace σ₂And model N₁Schematic diagram of the simplest calibration.

FIG. 5 is a simple Petri Net N₂Schematic representation.

FIG. 6 shows trace σ₃And model N₂Schematic diagram of the simplest calibration.

FIG. 7 is a logical Petri Net model LPN₁Schematic representation.

FIG. 8 shows trace σ₃And model LPN₁Schematic of the logic simplest calibration.

FIG. 9 shows model N₁Corresponding process tree PT₁Schematic representation.

FIG. 10 is a logical Petri Net model LPN₂Schematic representation.

FIG. 11 shows trace σ₁And model LPN₂Schematic of the logic simplest calibration.

FIG. 12 shows trace σ₂And model LPN₂Schematic of the logic simplest calibration.

FIG. 13 is a graph of model N corrected based on the Fahland correction method₃Schematic representation.

Fig. 14 is a schematic view of a business process model of hospital orthopedics.

Fig. 15 is a schematic diagram of an orthopedic encounter model modified based on the Fahland method.

Fig. 16 is a schematic diagram of an orthopedic visit model modified based on Goldratt and knapack methods.

Fig. 17 is a schematic view of an orthopedic examination model modified by the method of the present invention.

FIG. 18 is a graph showing the variation of the accuracy of different model correction methods at different trace scales.

FIG. 19 is a graph showing the variation of the degree of compactness of different model modification methods.

Detailed Description

The invention is described in further detail below with reference to the following figures and detailed description:

the method for selecting the nested concurrent process model modification comprises the following steps:

defining a Petri Net

The quadruple PN (P, T; F, M) is called a Petri net, if and only if:

(P, T, F) is a net;

wherein P is a finite library set representing various states in the encounter model, and T is a finite variableThe migration set is used for representing the names of the activities in the treatment process,

is a finite arc set.

②

Referred to as an identification of the net PN,

representing a set of all identified Petri nets.

③ PN has the following transition occurrence rules:

a. for T e.T, if

I.e. p contains tokken, t is said to be enabled under the M designation, denoted M [ t>Indicating that the activity is conditioned for occurrence;

b. if M [ t >, then under M, t can occur, deriving a new identifier M 'from M trigger t, denoted M [ t > M', indicating a transition from one activity to another in the visit procedure;

and is aligned with

All have:

where, t represents the front set of transition t, and t represents the back set of transition t.

Wherein, various states in the process of seeing a doctor include: idle state, ongoing state.

Wherein, the names of the activities in the treatment process include:

the system comprises the following activities of self-service registration machine registration, emergency treatment, selection of department, expert outpatient service, common outpatient service, CT making, nuclear magnetic resonance MRI making, doctor diagnosis, basic treatment, diagnosis and treatment bill, payment, pharmacy dispensing, hospitalization and the like.

Defining calibration

Is provided with

Is a set of activities, a represents the activity of a visit procedure,

a set representing a visit flow activity;

sigma epsilon is a trace on A, PN (P, T; F, M) is a Petri net on A;

the doublet (a, T) ∈ A > > × T > > > { (> > >, >) } is a movement, representing the trajectory that is performed in the visit procedure.

The calibration γ ∈ (a > > × T >) is a movement sequence (a, T) · between the trace σ and the model PN, and satisfies:

(1)π₁(γ)_↓Aσ, i.e. the movement sequence of the hospital visit procedure generates the trajectory of the visit activity;

(2)

i.e. the sequence of movements of the model results in a complete sequence of the visit procedure.

Wherein > represents no movement, a > > > > > > < a { > > > > >;

π₁(γ)_↓Athe first row element in the calibration is represented, which is the activity the patient has undergone during the actual visit;

π₂(γ)_↓Tthe second row element in the calibration is represented, which is the activity passed through in the hospital visit procedure;

represents that no activity occurred in the flow;

m_ian input representing a sequence of visit model movements; m is_fAn output representing a sequence of visit model movements;

i represents a pre-input state and f represents a post-output state.

For the doublet (a, t) ∈ γ in the calibration, the definition is as follows:

a) if a belongs to A and t ═ > >, the action is a log action, which means that some treatment activities occur in the real treatment process, but do not occur according to the hospital flow chart, which indicates that the existing hospital flow chart does not change along with the change of the requirements of the patients, and the existing hospital flow chart does not meet the real requirements of the patients;

b) if a > and T belongs to T, the model action is represented, which means that some treatment activities do not occur in the real treatment process and occur according to the hospital flow chart;

c) if a belongs to A and T belongs to T, the operation is synchronous, and the flow chart of the representative hospital conforms to the real clinic process;

d) otherwise, it is illegal action.

Defining a process tree

Is provided with

Given an operator set, τ is an implicit transition, then:

(ii) a ∈ A { [ τ } is a process tree;

② setting PT₁，…，PT_nIs a process tree, then

Is also a process tree, n>0。

Operator set

The operators in (1) have 4 kinds:

x represents a selection relation, meaning that only one sub-tree corresponding to the operator occurs;

→ represents the order relationship, meaning that the sub-tree corresponding to the operator will occur sequentially;

indicating a cyclic relationship in which PT₁Denotes the circulation volume, PT₂，…，PT_nRepresenting a cyclic path, n > 0;

Λ represents a parallel relationship.

Defining optimal calibration

Calibration gamma epsilon gamma between trace sigma and process model PN_σ,PNIs an optimal calibration, if and only if

Wherein lc (a, t) is a probability cost function and represents the deviation degree between the hospital flowchart and the real clinic process;

Γ_σ,PNrepresents the set of all calibrations between trace σ and model PN;

γ' represents other than the optimal calibration;

∑_(a,t)∈γlc ((a, t)) represents the set of least likely cost functions for optimal calibration;

∑_{(a,t)∈γ′}lc ((a, t)) represents the set of least likely cost functions for non-optimal calibration.

In order to be able to measure the cost function, the performance cost function lc (a, t) is given a fixed value in the present invention.

In the present invention, the cost function of the log action lc (a, t) and the model action lc (<, t) is defined as 1, and the synchronization action lc (a, t) is 0. Since there may be more than one calibration with the lowest cost, there may be more than one optimal calibration.

Defining select recognition pairs and recognition pair sets

Is provided with

And PN is (P, T; F, M), PT is the process tree of PN;

the selection recognition pair crp is a doublet (p)_left,p_right)，p_left＝·(LLN(K))，p_right(rln (k)), and satisfies:

and K ═ x;

wherein K represents a leaf node;

LLN (K) represents the leftmost leaf node of the selected structure in the visit model;

RLN (K) represents the rightmost leaf node of the selected structure in the visit model;

p_leftrepresenting the front set of leftmost leaf nodes, p_rightRepresenting the postset of the rightmost leaf node, x represents the product sign.

Selecting the set of identification pairs CRPS as a set comprising all identification pairs, there is:

defining a logical Petri Net

A logical Petri net is defined as a six-tuple LPN ═ P, T; F, I, O, M) if and only if:

(1) p is a finite set of libraries;

(2)T＝T_D∪T_I∪T_Ois a finite set of transitions that are,

if T is equal to T_I∩T_OThen, then

a.T_DRepresenting a set of transitions in a Petri net;

b.T_Irepresenting a set of logical input transitions, pair

t input library receives a logical expression f_I(t) a limit;

c.T_Orepresenting sets of logical output transitions, pair

t ofLogical expression f received by output library_O(t) a limit;

(3) f ═ P × T (T × P) is a finite arc set;

(4) i is the mapping of the transitions from the logical inputs to the logical input function, for

(5) O is the mapping of the transition from the logical output to the logical output function, for

(6)M:P→Z₀Is a representation function of a Petri network, Z₀＝{0,1,2，…}。

An example of a logical Petri net is given in FIG. 1.

t₁In order for the logic input to transition,

is t₁Is used to generate a logical input function of (c),

represents t₁On triggering, p₁And p₂Cannot both exist.

Represents t₁The trigger condition is satisfied in two cases:

1)p₁and p₃Presence of a token; 2) p is a radical of₂And p₃There is a tobken.

t₂The triggering rule is the same as the classic Petri net in the transition of the classic Petri net.

t₃For logic output transition, its logic output function O (t)₃)＝p₇∨p₈Represents t₃There are three cases after initiation:

1)p₇and p₈Wherein there is a token; 2) p is a radical of₇Wherein there is a token; 3) p is a radical of₈Zhongzhiken。

The event log is calibrated with the process model, log actions and model actions may occur, which indicates that an error occurs between the original process model and the event log. At this time, the model cannot accurately express the event log, that is, the event log and the model cannot be completely fitted. Model modification belongs to the third category of applications of process mining: and (5) improvement.

The purpose of the revision is to modify the model to better reflect real-world traffic. The method takes an event log and an original model as input, and analyzes the log to find out the position of deviation and repair the deviation so as to obtain a repaired model. The repaired model should be similar to the original model, i.e. the basic structure of the original model cannot be changed.

Log actions and model actions can be found by using calibration, but the positions of the actions in the Petri network cannot be determined, in order to further improve the correction result and accurately position the position of the deviation, the calibration is expanded, the simplest calibration and the logic simplest calibration are provided, and the formal definition is given as follows.

Define simplest calibration

Let (a, T, H) e (a { > >) }) × (T { > > }) × (H) be an action;

simplest calibration β ═ a₁,t₁,h₁)…(a_|β|,t_|β|,h_|β|) Is an action queue, and satisfies:

(1)Γ_σ,PNis the set of all calibrations between trace σ and model PN;

(2)

namely, the action queue generated in the model is calibrated in the simplest way, and finally the termination identifier is reached;

(3) h is a Token reachable library set generated by transition triggering in the M state, and is called as a reachable library set;

wherein h represents a Tokenreachable library generated by transition triggering;

(a₁,t₁,h₁) Representing action queuesInitial motion, (a)_|β|,t_|β|,h_|β|) An end action representing an action queue;

π2(π₁(β)_↓A) ↓ T denotes the second row element after the simplest alignment.

In the definition of the simplest calibration, if the Token absence required for transition triggering is present, the flag is negative, indicating that a deviation has occurred here. Taking the Petri net in FIG. 2 as an example, if the transition t₃Triggering the required library p₂Missing Token, then it is marked as-p₂This means that the transition, although it has occurred, is absent.

Example 1, a Petri Net model N for selecting nested concurrency structures is shown in FIG. 2₁Wherein the initial depot P₁There is an initial identification. If there is a log L₁＝{<t₁,t₂,t₃,t₅,t₆,t₇,t₉>,<t₁,t₂,t₃,t₅,t₆,t₇,t₈,t₉>Their simplest alignment is shown in fig. 3 and 4. Wherein FIGS. 3 and 4 are for traces β with deviations₁、β₂The simplest calibration of (1).

Trace beta₁、β₂The cost function corresponding to the simplest calibration of (1) is lc (beta) respectively₁)＝1、lc(β₂)＝2。

For beta₁Expanding action (t) in (1)₁,t₁,h₁) Simplest calibration beta₁Reachable library set h₁＝{p₂For (a)_|β1|,t_|β1|,h_|β1|) In particular, it can be expressed as h₂＝{p₃}，h₃＝{-p₂,p₃,p₄,p₅}，h₄＝{-p₂,p₃,p₅,p₇}，h₅＝{-p₂,p₃,p₇,p₈}，h₆＝{-p₂,p₃,p₉}，h₇＝{-p₂,p₃,p₁₀}. For beta₂Expanding action (t) in (1)₁,t₁,h₁) Simplest calibration beta₂Reachable library set h₁＝{p₂For (a)_|β2|,t_|β2|,h_|β2|) In particular, it can be expressed as h₂＝{p₃}，h₃＝{-p₂,p₃,p₄,p₅}，h₄＝{-p₂,p₃,p₅,p₇}，h₅＝{-p₂,p₃,p₇,p₈}，h₆＝{-p₂,p₃,p₉}，h₇＝{-p₂,-p₆,p₃,p₉}，h₈＝{-p₂,-p₆,p₃,p₁₀}。

The location of the deviation can be found through a simplest calibration, but the cost function is still high. Based on different network models, the logic simplest calibration is provided in the logic Petri network, so that the log and the model can be fitted in the modified logic Petri network, and the log action and the model action can be avoided, so that the cost function value is 0.

A formal definition of the logical simplest calibration is given below.

Defining a logical simplest calibration

Logic simplest calibration LPN_β＝(a₁,t₁,h₁)…(a_|β|,t_|β|,h_|β|) Is an action queue, and satisfies:

(1)Γ_σ,LPNis the set of all calibrations between trace σ and model LPN;

(2)

namely:

the logic simplest calibrates an action queue generated in the model, and finally the termination identifier is reached;

(3) simplest calibration LPN between trace sigma and logical Petri Net model LPN_β∈Γ_σ,LPNIs a logical simplest calibration, if and only if

Wherein, | h_iI represents the number of reachable libraries;

| P | represents the number of possible states in the encounter model;

input representing a sequence of visit model movements under a logical Petri net;

representing the output of the sequence of movements of the visit model under the logical Petri net.

Example h₁＝{p₁,p₂,p₃H, then | h₁And | represents that the reachable library has a value of 3.

As shown in FIG. 5, with N₂For example, P₁There is an initial identification. If there is a log L₂＝{<t₁,t₃,t₂,t₅>H, trace β after the simplest calibration₃As shown in fig. 6. For beta₃Expanding action (t) in (1)₁,t₁,h₁) Simplest calibration beta₃Reachable library set h₁＝{p₂For (a)_|β3|,t_|β3|,h_|β3|) In particular, it can be expressed as h₂＝{p₃}，h₃＝{-p₂,p₃,p₄}，h₄＝{-p₂,p₃,p₅}。

By the net N₂For example, the following figures are N₂Logic Petri net LPN (low power density network) repaired by logic simplest calibration for original model₁。P₁In which there is an initial identity, σ₃＝<t₁,t₃,t₂,t₅>. And N₂In contrast, a stream relation is added and at transition t₂Add logic input condition

As can be seen from fig. 8, after the logic-simple calibration, the probability cost function lc (a, t) is 0. Trace sigma₃Is a perfect fit. Logic simplest calibration

Reachable library set h₁＝{p₂For

In particular, it can be expressed as h₂＝{p₃}，h₃＝{-p₂,p₃,p₄}，h₄＝{-p₂,p₃,p₅}. If in the common Petri network, only one branch can be selected to continue running after encountering the selected branch, the method is impossible when jumping to another branch. In model LPN₁When t is in₂After adding a logic input condition, only p₂Or p₃One of them is present with Token, t₂A transition may occur. When a more complex logic relationship is encountered, the logic Petri net can be more effectively expressed.

Observe the given Petri Net model N₁The method is a structure for selecting nested concurrency. In the selection structure of the Petri net, all branches have the same initial library place and termination library place, and p is used for convenience of description_startInitial library representing selection Structure, p_endA termination library of selection structures is indicated. In order to locate the position of the deviation, the concept of a process tree is introduced, and the information of a Petri net structure and transition is recorded in the process tree. Using Petri model N₁An exemplary process tree is shown in fig. 9.

By traversing the process tree, the nodes of the selection structure can be found, and then the leftmost leaf node and the rightmost leaf node thereof can be found. For convenience of representation, the leftmost leaf node is denoted as lln (k) and the rightmost leaf node is denoted as rln (k). Because all the information recorded in the process tree is transition information, in the Petri network, the front set of the leftmost leaf child node and the rear set of the rightmost leaf child node are initial library places p in the selection structure_startAnd terminating the library p_end. Will select the knotAnd a binary group formed by the constructed initial library and the constructed termination library is called a selection identification pair, and all the selection identifications appearing in one Petri net are symmetrical to be a selection identification pair set.

The following method 1 is proposed for finding the selective identification pair of the selected structure in the hospital visit process, and the specific process is as follows:

(1) initializing the selection recognition pair crp (p)_left,p_right) And selecting a set of identification pairs CRPS;

(2) traversing all child nodes in the process tree;

(3) if the child node is "x";

(4) finding a leaf node under the child node;

(5) the following two elements of the selection recognition pair are obtained, namely:

the front set of the leftmost leaf node (lln (k)) and the back set of the rightmost leaf node (rln (k));

(6) the obtained selection recognition pair crp (p)_left,p_right) Merging into a selection recognition pair set CRPS;

with process tree PT₁For example, there is (t)₂,t₃) A structure is selected. At PT₁In (2), the leftmost leaf node of the selection nodes is t₂The rightmost leaf node is t₇. The front set of leftmost leaf nodes is p₂The last set of rightmost leaf nodes is p₉. Obtaining a selection identification pair (p)₂,p₉)。

The following method 2 is proposed for determining the deviation position of the treatment process, and the specific process is as follows:

(1) deviation position Dev (t) of initialization visit process_m,t_n)；

(2) Traversing the simplest calibration beta, and if log actions occur, performing the next step;

(3) when a missing Token library appears, recording the position of the activity at the moment;

(4) if a library carried "-" or at h appears in the reachable library set_iWhere the library appears multiple times, is denoted as p_kThe abnormal library represents that an abnormal state occurs at the moment;

(5)store the abnormal place p_kAssigning the previous set of values to the starting position of the deviation position, and assigning the abnormal position p to the abnormal position_kIs assigned to the end position of the deviation position and the end position of the deviation position cannot be the leftmost leaf node lln (k);

(6) deviation position Dev (t) of return visit procedure_m,t_n)；

Wherein, t_mIndicating the starting position of the deviation position, t_nIndicating the end position of the deviation position.

After the simplest calibration beta is carried out in the method, the next step is carried out when the log action occurs, when the activity state of the same treatment process occurs for multiple times in the reachable library set H, the deviation is determined to occur, and the treatment activity name with the deviation is returned.

The following method 3 is proposed for correcting the hospital visiting flow, determining the deviation position between the actual situation and the given hospital business flow, restoring the original hospital business flow at the deviation position, and adding the flow relationship, and the specific process is as follows:

(1) initializing a logical Petri network LPN, initializing an activity t of a selected branch_i；

(2) Calling the method 1 to obtain a selective recognition pair set CRPS;

(3) calling method 2 to obtain deviation position Dev (t) of the clinic flow_m,t_n)；

(4) Traversing the simplest calibration beta, and if log actions occur, performing the next step;

(5) if the occurrence condition is met:

starting position t of offset position_mIs p_leftAnd branch t is selected_iIs not the leftmost leaf node lln (k);

(6) add the postset from the leftmost leaf node (lln (k)) to the option branch t_iThe flow relationship of (a);

(7) if the occurrence condition is met: end position t of deviation position_nIs p_rightAnd t is_nNot the rightmost leaf node;

(8) then add the node from the rightmost leaf node rln (k) to the biased position end position t_nFlow off of the preceding set ofIs a step of;

(9) mining a transition triggering condition by using an LPN mining algorithm to obtain a logic transition and a logic input and output function;

(10) returning the revised visit logic Petri network LPN;

the model is a time-optimized business process model, the complete fitting of the actual treatment process and the model can be realized, the whole model structure is clearer, the described model is more accurate, the activity of selecting the nested concurrent structure is optimized, the waiting time of the patient during treatment is shortened, the whole treatment process is more efficient, and the treatment experience of the patient is improved.

The correction method provided by the invention can ensure that only the input-output logical relationship between the flow relationship and the transition is increased on the basis of the original model, the whole model structure is clearer, the described model is more accurate, and the activity of selecting the nested concurrent structure is optimized, so that a correct process model is obtained by correction, and the whole diagnosis process is more efficient.

Example 2: FIG. 10 illustrates a model LPN of a logical Petri net₂，P₁There is an initial identification.

With log L₁＝{<t₁,t₂,t₃,t₅,t₆,t₇,t₉>,<t₁,t₂,t₃,t₅,t₆,t₇,t₈,t₉>For example, by a logic simplest calibration, a model LPN is found₂And log L₁There is a deviation therebetween. The logical simplest calibration is given in fig. 11 and 12. The main idea is to find the position of the deviation, add a logical relationship to the transition at the position of the deviation, and formulate the input and output conditions. And adding logical judgment to the transition by adding arcs on the premise of not greatly changing the original model, so that the log is fitted with the model.

As shown in FIGS. 11 and 12, two traces σ₁And σ₂And model LPN₂Is a perfect fit. At transition t₃To increase the input condition

And output condition O (t)₃)＝p₄∧p₅. At transition t₇Is increased by an input condition I (t)₇)＝p₇∧p₈And output conditions

In the case of a logical Petri net,

only one of the representations a and b can occur, and a ^ b represents that a and b can occur simultaneously. At transition t, unlike the conventional Petri nets₃Where only p is allowed₂Or p₃Can be input at transition t₃Subsequent output post p₄And p₅Can occur simultaneously. Whereas in the ordinary Petri network, transition t₃Requires p to occur₂And p₃With a tobken. In the same way, t₇Input of (2) requires p₇And p₈At the same time, t₇Output of only allowing p₆And p₉Whereas in a normal Petri net, once a transition t occurs₇Occurrence will yield two tokens, each assigned to p₆And p₉. For beta_σ1Expanding action (t) in (1)₁,t₁,h₁) Simplest calibration beta_σ1Reachable library set h₁＝{p₂For

In particular, it can be expressed as h₂＝{p₃}，h₃＝{p₄,p₅}，h₄＝{p₅,p₇},h₅＝{p₇,p₈},h₆＝{p₉},h₇＝{p₁₀}。

For beta_σ2Expanding action (t) in (1)₁,t₁,h₁) Simplest calibration beta_σ2Reachable library set h₁＝{p₂For

In particular, it can be expressed as h₂＝{p₃}，h₃＝{p₄,p₅}，h₄＝{p₅,p₇}，h₅＝{p₇,p₈}，h₆＝{p₆}，h₇＝{p₉}，h₈＝{p₁₀}。

The visit model modified by the Fahland method is shown in fig. 13, and a subprocess obtained by digging two traces is added to the appropriate position of the original model. Adding invisible transitions, skipping the add sub-process. After the restoration, a lot of repeated activities occur, and the complexity of the model is also improved, so that the replay of the trace is only realized, but the correct model is not mined.

Due to the addition of the self-loop, the repaired process model allows this self-loop to occur indefinitely, which is not possible in practice during the visit, resulting in poor fitting and accuracy of the visit model.

The method based on the logic simplest calibration can completely fit the trace and the model, the whole model structure is clearer, the described model is more accurate, and the fitting degree, the simplicity and the accuracy of the model are improved.

The standard of cleanliness is the principle of the Oncam razor. In process mining, this principle means that the best model is the simplest model that can interpret the behavior seen in the log. Since invisible transitions are internal transitions and invisible to the outside, they are ignored in the conciseness of the new definition. The compactness of the new definition is as follows:

in the formula: l (σ) represents the frequency of occurrence of a trace in the event log, Σ_σ∈LL (σ) represents the total number of traces in the event LogP is a finite set of libraries, T is a finite set of transitions, L (σ, P, T) represents the sum of transitions and libraries in a trace in the event log, Σ_σ∈LL (σ, P, T) represents the total number of live transitions and bins in the event log. The model and event log used in the experiment originated from a certain hospital in Qingdao. Experiments for the Fahland method, the Goldratt method and the knapack method were obtained by excavation with a process excavation tool ProM 6.6. Because the logic Petri net has no corresponding experimental tool, the experimental result of the invention adopts manual simulation.

Fig. 14 is a model of a business process of orthopedics department in a certain hospital. In order to express the method of the present invention more briefly, some unimportant parts are deleted, and the main business process after the granularity division is described as follows:

first, if the disease condition is serious, the Patient (Patient) can be registered by Self-help Registration Machine (Self-Service Registration Machine) and the Department (Choose Department) can be selected on the Registration Machine for treatment. After the selection of the department, the expert Clinic (professional Clinic) or the General Clinic (General Clinic) can be selected according to the condition. After the patients are out-patient according to the order of calling and registration, doctors schedule to make CT (computed tomography) or Magnetic Resonance Imaging (Magnetic Resonance Imaging) according to the condition of different patients for better analysis of the disease conditions. After the doctor has finished the Diagnosis of the tablet (Diagnosis), the patient is divided into two cases: selecting severe disease, and Further examining (Further administration) and then hospitalizing (Hospitalization), and after the disease is improved, reimbursing (Reimburse) with medical insurance card and leaving the hospital; if the condition is mild, the patient is given Basic Treatment (Basic Treatment). Doctors make Medical Treatment lists on computers, patients pay (Payment) and go to Pharmacy dispensing (Pharmacy dispensing), and patients can leave the Pharmacy dispensing (Leaving Hospital) after taking the medicines.

In the actual medical examination, doctors often cannot make correct diagnosis only by means of CT or MRI, and further detailed examination is needed, and basic treatment is performed after the detailed examination is correct. This also ensures that the patient heals as quickly as possible. Because the national medical insurance catalogs are different in different years, some medicines are newly brought into the national medical insurance in the pharmacy process, and the patient cannot be reimbursed before but can be reimbursed at present. These flows occur to illustrate the case where there is a jump in the model from one branch of the selection branch to another. The method can ensure that only the input-output logic relationship between the flow relationship and the transition is increased on the basis of the original model, thereby correcting to obtain the correct process model.

Table 1 gives the main information of the three sets of event logs, including the number of traces, the length of the traces, and the number of deviations.

Log L₁1107 tracks are arranged in the middle, the length range of the tracks is between 9 and 13, and 1029 deviations exist; log L₂1417 tracks are arranged in the mark, the length range of the tracks is 9-13, and 1276 deviations exist; log L₃There are 2109 traces, the length of the trace is between 9-13, and there are 1956 deviations. The download link of the event log used in this experiment is:

https://pan.baidu.com/s/1bIXT4wIKMXVg104UELIErg。

TABLE 1 three sets of Log information

The Fahland method uses the deviation between the calibration discovery model and the log to collect sub-logs that do not fit. And mining the sub-logs by an inductive mining algorithm, and adding the mined sub-process to a proper position of the original model as a self-loop. Repeated sub-processes are added, so that the model has the same name activity for many times, and the accuracy and the simplicity of the model are improved. The Goldratt method and knapack method are modified in such a way that a single active self-loop is added according to different constraints. The addition of self-loops and invisible transitions, while improving the fit, greatly reduces the accuracy of the model. FIG. 15 is a Fahland method modified model. Fig. 16 is a model modified by Goldratt method and knapack method. Because the model structure is simple, the model modified by the Goldrtat method and the Knapack method is the same model. FIG. 17 is based on logic PetThe ri-net corrected orthopedic diagnosis model has the advantages that log actions appear in event logs for many times, the deviation is determined through simplest calibration, the logical relationship is mined, and the logical relationship is added in common transition. Obtaining a logical input function

I (pharmacy) P₁₂∧P₁₃. Logic output function O (basic therapy) ═ P₉∧P₁₀、

The common Petri network cannot well describe the logical relation among the activities, and the simplicity, the cleanliness and the accuracy of the model are greatly improved due to the fact that self-loop and invisible transition do not exist.

TABLE 2 comparison of correction results for various methods

Table 2 shows a comparison of the elements of the model obtained with three sets of event logs, modified by different methods. The comparison of the number of important elements such as libraries, transitions, invisible transitions, and flow relationships required by a process model can be seen from the table. It can be seen that the change between each element based on the correction method provided by the invention and the original model is minimum, and the simplicity is reflected from the side surface to be higher. Tables 3, 4 and 5 are the correction results of three correction methods, respectively, including increased library number | P |, increased transition number | T + τ |, increased flow relationship | F |, repeated transition number | T |, and Simplicity.

TABLE 3 repair results of the Fahland method

TABLE 4 repair results of the Goldrtat, Knapack method

TABLE 5 correction results of the method of the invention

From the analysis of tables 3, 4 and 5, the method of the invention has the advantages that various elements of the added model are less than those of other methods, the modified model is simpler and more practical, and the logical relationship among activities can be better expressed than those of other methods. Meanwhile, the simple cleanliness is superior to other methods. The invention adopts event logs with different orders of magnitude and carries out model correction by the four methods. A comparison of the accuracy and simplicity of the different methods obtained with event logs of different orders of magnitude is given below. Since the degrees of fit corrected by the different methods are all 1, no comparison is necessary. The accuracy of a model represents the ability of the model to replay event logs, with a higher accuracy of one model indicating that the model generates fewer traces beyond a given event log record. The variation in accuracy of the four methods at different trace scales is shown in fig. 18. As shown in fig. 19, since the original model has a simple structure, and the graphs obtained by correcting the knapack method and the Goldratt method are consistent, the simplicity of the two methods is also consistent. The simplicity of the correction method based on the present invention is always maintained at a high level, significantly higher than the Fahland method, knapack and Goldratt methods.

It should be understood, however, that the description herein of specific embodiments is not intended to limit the invention to the particular forms disclosed, but on the contrary, the intention is to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the invention as defined by the appended claims.

Claims (1)

1. The method for selecting the nested concurrent process model modification is characterized by comprising the following steps of:

defining a Petri Net

The quadruple PN (P, T; F, M) is called a Petri net, if and only if:

(P, T, F) is a net;

wherein, P is a finite library set used for representing various states in the clinic model, T is a finite transition set used for representing names of activities in the clinic flow,

is a finite arc set;

②M：

referred to as an identification of the net PN,

representing a set formed by all identification Petri nets;

③ PN has the following transition occurrence rules:

a. for T e.T, if

M (p) is not less than 1, i.e. if p contains tokken, t is called to be enabled under the mark M and is marked as M [ t ≧ 1>Indicating that the activity is conditioned for occurrence;

b. if M [ t >, then under M, t can occur, deriving a new identifier M 'from M trigger t, denoted M [ t > M', indicating a transition from one activity to another in the visit procedure;

and is aligned with

All have:

wherein, t represents the front set of transition t, and t represents the rear set of transition t;

defining calibration

Is provided with

Is a set of activities, a represents the activity of a visit procedure,

a set representing a visit flow activity; sigma epsilon is a trace on A, PN (P, T; F, M) is a Petri net on A;

the binary group (a, T) belongs to A > >. times T > > > >/{ (> > > > > > > > > > > > > > > > > >) } is a movement and represents a track performed in the treatment flow;

the calibration γ ∈ (a > > × T >) is a movement sequence (a, T) · between the trace σ and the model PN, and satisfies:

(1)π₁(γ)_↓Aσ, i.e. the movement sequence of the visit procedure generates the trajectory of the visit activity;

(2)

namely, the moving sequence of the model generates a complete visiting procedure sequence;

wherein > represents no movement, a > > > > > > < a { > > > > >;

π₁(γ)_↓Athe first row element in the calibration is represented, which is the activity the patient has undergone during the actual visit;

π₂(γ)_↓Tthe second row element in the calibration is represented, which is the activity passed through in the hospital visit procedure;

represents that no activity occurred in the flow;

m_ian input representing a sequence of visit model movements; m is_fAn output representing a sequence of visit model movements;

i represents a pre-input state, and f represents a post-output state;

for the doublet (a, t) ∈ γ in the calibration, the definition is as follows:

a) if a belongs to A and t ═ > >, the action is a log action, which means that some treatment activities occur in the real treatment process, but do not occur according to the hospital flow chart, which indicates that the existing hospital flow chart does not change along with the change of the requirements of the patients, and the existing hospital flow chart does not meet the real requirements of the patients;

b) if a is ═ and T belongs to T, the model action is taken, which means that some treatment activities do not occur in the real treatment process and occur according to the hospital flow chart;

c) if a belongs to A and T belongs to T, the operation is synchronous, and the flow chart of the representative hospital conforms to the real clinic process;

d) otherwise, the action is illegal;

defining a process tree

Is provided with

Given an operator set, τ is an implicit transition, then:

(ii) a ∈ A { [ τ } is a process tree;

② setting PT₁，…，PT_nIs a process tree, then

Is also a process tree, n>0；

Defining optimal calibration

Calibration gamma epsilon gamma between trace sigma and process model PN_σ,PNIs an optimal calibration, if and only if

∑_(a,t)∈γlc((a,t))≤∑_{(a,t)∈γ′}lc((a,t))；

Wherein lc (a, t) is a probability cost function and represents the deviation degree between the hospital flowchart and the real clinic process;

Γ_σ,PNrepresents the set of all calibrations between trace σ and model PN;

γ' represents other than the optimal calibration;

∑_(a,t)∈γlc ((a, t)) represents the set of least likely cost functions for optimal calibration;

∑_{(a,t)∈γ′}lc ((a, t)) represents the set of least likely cost functions for non-optimal calibration;

defining select recognition pairs and recognition pair sets

Let PN be (P, T; F, M), PT is the process tree of PN;

the selection recognition pair crp is a doublet (p)_left,p_right)，p_left＝·(LLN(K))，p_right(rln (k)), and satisfies:

and K ═ x;

wherein K represents a leaf node;

LLN (K) represents the leftmost leaf node of the selected structure in the visit model;

RLN (K) represents the rightmost leaf node of the selected structure in the visit model;

p_leftrepresenting the front set of leftmost leaf nodes, p_rightRepresenting the postset of the rightmost leaf node, x representing the product sign;

selecting the set of identification pairs CRPS as a set comprising all identification pairs, there is:

defining a logical Petri Net

A logical Petri net is defined as a six-tuple LPN ═ P, T; F, I, O, M) if and only if:

(1) p is a finite set of libraries;

(2)T＝T_D∪T_I∪T_Ois a finite set of transitions that are,

if T is equal to T_I∩T_OThen, then

a.T_DRepresenting a set of transitions in a Petri net;

b.T_Irepresenting a set of logical input transitions, pair

t input library receives a logical expression f_I(t) a limit;

c.T_Orepresenting sets of logical output transitions, pair

t output library receives logical expression f_O(t) a limit;

(3) f ═ P × T (T × P) is a finite arc set;

(4) i is the mapping of the transitions from the logical inputs to the logical input function, for

I(t)＝f_I(t)；

(5) O is the mapping of the transition from the logical output to the logical output function, for

O(t)＝f_O(t)；

(6)M:P→Z₀Is a representation function of a Petri network, Z₀＝{0,1,2，…}；

Define simplest calibration

Let (a, T, H) e (a { > >) }) × (T { > > }) × (H) be an action;

simplest calibration β ═ a₁,t₁,h₁)…(a_|β|,t_|β|,h_|β|) Is an action queue, and satisfies:

(1)Γ_σ,PNall alignment between trace sigma and model PNGathering;

(2)

namely, the action queue generated in the model is calibrated in the simplest way, and finally the termination identifier is reached;

(3) h is a Token reachable library set generated by transition triggering in the M state, and is called as a reachable library set;

wherein h represents a Tokenreachable library generated by transition triggering;

(a₁,t₁,h₁) Representing an initial action of the action queue, (a)_|β|,t_|β|,h_|β|) An end action representing an action queue;

π2(π₁(β)_↓A) ↓ T represents the second row element after the simplest calibration;

defining a logical simplest calibration

Logic simplest calibration LPN_β＝(a₁,t₁,h₁)…(a_|β|,t_|β|,h_|β|) Is an action queue, and satisfies:

(1)Γ_σ,LPNis the set of all calibrations between trace σ and model LPN;

(2)

namely:

the logic simplest calibrates an action queue generated in the model, and finally the termination identifier is reached;

(3) simplest calibration LPN between trace sigma and logical Petri Net model LPN_β∈Γ_σ,LPNIs a logical simplest calibration, if and only if

|h_i|≤|P|；

Wherein, | h_iI represents the number of reachable libraries;

| P | represents the number of states in the encounter model;

input representing a sequence of visit model movements under a logical Petri net;

representing the output of the visit model movement sequence under the logical Petri net;

the following method 1 is proposed for finding the selective identification pair of the selected structure in the hospital visit process, and the specific process is as follows:

(1) initializing the selection recognition pair crp (p)_left,p_right) And selecting a set of identification pairs CRPS;

(2) traversing all child nodes in the process tree;

(3) if the child node is "x";

(4) finding a leaf node under the child node;

(5) the following two elements of the selection recognition pair are obtained, namely:

the front set of the leftmost leaf node (lln (k)) and the back set of the rightmost leaf node (rln (k));

(6) the obtained selection recognition pair crp (p)_left,p_right) Merging into a selection recognition pair set CRPS;

the following method 2 is proposed for determining the deviation position of the treatment process, and the specific process is as follows:

(1) deviation position Dev (t) of initialization visit process_m,t_n)；

(2) Traversing the simplest calibration beta, and if log actions occur, performing the next step;

(3) when a missing Token library appears, recording the position of the activity at the moment;

(4) if a library carried "-" or at h appears in the reachable library set_iWhere the library appears multiple times, is denoted as p_kThe abnormal library represents that an abnormal state occurs at the moment;

(5) store the abnormal place p_kAssigning the previous set of values to the starting position of the deviation position, and assigning the abnormal position p to the abnormal position_kIs assigned to the end position of the deviation position and the end position of the deviation position cannot be the leftmost leaf node lln (k);

(6) deviation position Dev (t) of return visit procedure_m,t_n)；

Wherein, t_mIndicating the starting position of the deviation position, t_nAn end position indicating a deviation position;

the following method 3 is proposed for correcting the hospital visiting flow, determining the deviation position between the actual situation and the given hospital business flow, restoring the original hospital business flow at the deviation position, and adding the flow relationship, and the specific process is as follows:

(1) initializing a logical Petri network LPN, initializing an activity t of a selected branch_i；

(2) Calling the method 1 to obtain a selective recognition pair set CRPS;

(3) calling method 2 to obtain deviation position Dev (t) of the clinic flow_m,t_n)；

(4) Traversing the simplest calibration beta, and if log actions occur, performing the next step;

(5) if the occurrence condition is met:

starting position t of offset position_mIs p_leftAnd branch t is selected_iIs not the leftmost leaf node lln (k);

(6) add the postset from the leftmost leaf node (lln (k)) to the option branch t_iThe flow relationship of (a);

(7) if the occurrence condition is met: end position t of deviation position_nIs p_rightAnd t is_nNot the rightmost leaf node;

(8) then add the node from the rightmost leaf node rln (k) to the biased position end position t_nThe flow relationships of the preceding set of (2);

(9) mining a transition triggering condition by using an LPN mining algorithm to obtain a logic transition and a logic input and output function;

(10) returning the revised visit logic Petri network LPN;

the corrected hospital clinic procedure is a time-optimized business procedure model, represents the complete fitting of the actual clinic procedure and the clinic model, optimizes and selects the activity of the nested concurrent structure, and shortens the waiting time of the patient during clinic.

Images