Token Replay-Based Object-Centered Business Process Violation Inspection Method and System

Abstract

The present invention discloses a method and system for checking violation of object-centric business processes based on token replay, which belongs to the field of process mining. The method comprises the following steps: constructing an object graph from an object-centric business process event log and extracting process execution; deleting cross-layer redundant dependencies in process execution, dividing process execution by echelons, and obtaining a binding sequence; performing token replay in a multi-object Petri net model according to the binding sequence to obtain a replay result and a deviation result; quantifying compliance according to a fitting metric, and using a local diagnosis method to display the part of the process with serious deviations, and obtaining the violation checking result of the object-centric business process. The system comprises a module for obtaining an object graph and process execution, a module for obtaining a binding sequence, a module for obtaining a replay result and a deviation result, and a module for obtaining a violation checking result of the object-centric business process. For object-centric business processes, the method of the present invention improves the accuracy of its compliance checking.

Description

Technical Field

[0001] The present invention belongs to the technical field of process mining, and particularly relates to an object - centered business process violation checking method and system based on token replay. Background Art

[0002] Traditional process mining is based on the concept of a single case and can only construct a process model from the perspective of a single object. As business processes become more and more complex, they usually involve multiple process objects, which are called object - centered business processes, that is, complex business processes are split into multiple objects with clear business attributes, and the business process is constructed and managed around the interaction and dependency relationships between these objects. The prior art cannot effectively handle the situation where the log contains multiple process objects.

[0003] Therefore, object - centered process mining for handling multi - object interaction behaviors has emerged. To ensure the compliance operation of this technology in complex processes, the compliance issues of object - centered process mining have received more and more attention. However, there are many drawbacks in the prior art when dealing with multi - object compliance issues. For example, the prior art can only perform compliance checks by flattening multi - object complex processes, resulting in the loss of process interaction information existing on some objects, thus leading to problems such as the inability to accurately identify deviations and the difficulty in confirming the source of deviations. Therefore, there is an urgent need for a new compliance checking method that can be applied to object - centered business processes, thereby improving the accuracy of object - centered business process compliance checks. Summary of the Invention

[0004] In order to solve the problems that the traditional method cannot accurately identify deviations and is difficult to confirm the source of deviations, the present invention proposes an object - centered business process violation checking method and system based on token replay, which can accurately and effectively identify object - centered business process violation behaviors, display the severity of deviations, quickly locate the problem links, and provide support for multi - object complex process analysis, and has practicality.

[0005] The technical solution of the present invention is as follows:

[0006] An object - centered business process violation checking method based on token replay includes the following steps:

[0007] Step 1: Obtain an object - centered business process event log, analyze the dependency relationship between objects to construct an object graph, and then use the connected - component extraction method to obtain the process execution;

[0008] Step 2: Use the transitive reduction method to delete cross - layer redundant dependencies, and then use the echelon division mechanism to obtain the binding sequence;

[0009] Step 3: Use the binding execution rule to perform token replay on a multi - object Petri net to obtain the replay result and the deviation result;

[0010] Step 4, calculate the fitness result and the local diagnosis result to obtain the inspection result of the object for the central business process violation.

[0011] Furthermore, the specific process of the said Step 1 is as follows:

[0012] Step 1.1, obtain the event log of the central business process with the object as the center, which consists of a set of events representing operations, and the said events include activities, occurrence times, end times, and basic attributes of associated objects;

[0013] Step 1.2, construct the object graph of the event log of the central business process with the object as the center by analyzing the dependency relationship between objects. The object graph is as follows:

[0014] ;

[0015] Among them, is the set of undirected graph nodes, which consists of all objects in the event log of the central business process with the object as the center; is the set of undirected graph edges, which consists of several pairs of objects with a dependency relationship; The calculation formula of is:

[0016] ;

[0017] Among them, , are two different objects associated with the event ; represents all events in the event log of the central business process with the object as the center; represents the event log of the central business process with the object as the center in which all objects associated with the event , and the calculation formula is:

[0018] ;

[0019] Among them, represents the event sequence associated with the object ;

[0020] Step 1.3, use the connected component extraction method to obtain the process execution; the specific process is as follows:

[0021] First, formalize the connected component extraction method as:

[0022] ;

[0023] Among them, is the connected component extraction method; Represents the process execution obtained from the largest connected subgraph; Is the set of objects in the largest connected subgraph;

[0024] Then, according to The directed graph of, obtain the process execution, Is defined as follows:

[0025] ;

[0026] Among them, Is the set of directed graph nodes, which is composed of all events associated with any object in the object-centered business process event log ; Is the set of directed graph edges, which is composed of several pairs of events with a direct following relationship;

[0027] , The calculation formulas of are respectively:

[0028] ;

[0029] ;

[0030] Among them, Is an empty set; Represents the direct following relationship between events in the object-centered business process event log , and the calculation formula is:

[0031] ;

[0032] Among them, Is an event different from event ; Is the th event; Is the th event; Is the th event.

[0033] Furthermore, the specific process of step 2 is:

[0034] Step 2.1: Use the transitive reduction method to delete the cross-layer redundant dependency relationships in the object graph, and obtain a minimum subgraph that maintains the transitive closure property of the original graph; the cross-layer redundant dependency relationship refers to the node-to-node dependency relationship in the object graph realized through an indirect path, and maintaining the transitive closure property of the original graph means that the reachability relationship between nodes in the graph remains unchanged after deleting the redundant dependencies;

[0035] Step 2.2: Adopt the echelon division mechanism to divide the process execution; the specific process is: In Implement the predecessor-free node recognition algorithm in the middle iteration. Each time, obtain all the current predecessor-free nodes as a echelon, and divide the process execution into several logically independent event echelons. The echelons are unordered within, and follow the established sequential logic between echelons; the predecessor-free node recognition algorithm is to identify the current predecessor-free nodes in the context of the graph and store them in a list;

[0036] Step 2.3 maps each logically independent event echelon to its respective binding to obtain the corresponding binding sequence; define the binding sequence as follows:

[0037] ;

[0038] where, is the th transition, is the specific object of each object type consumed when the th transition is executed; is the binding corresponding to the th transition.

[0039] Furthermore, the specific process of step 3 is as follows:

[0040] Step 3.1, define a multi-object Petri net as follows:

[0041] ;

[0042] where, is a Petri net; is a mapping function from places to object types; is a set of variable arcs, is the set of all arcs of the multi-object Petri net; is the cardinality label function of arcs; and represent the start and end identifiers respectively; The specific mapping relationship of

[0043] ;

[0044] where, is the set of places, is the set of object types;

[0045] Step 3.2, based on the binding sequence, perform token replay on the multi-object Petri net according to the binding execution rules for each process execution to obtain the replay result and deviation result of each process execution;

[0046] The replay result is the number of tokens generated, consumed, missing, and remaining after token replay for each process execution;

[0047] The number of generated tokens is defined as:

[0048] ;

[0049] where, is the set of post-set places of transition ; is the specific object of each object type consumed during the execution of the transition; is the token set of the multi-object Petri net ; is a token, indicating that there is an object in place ;

[0050] The number of consumed tokens is defined as:

[0051] ;

[0052] where, is the set of pre-set places of transition ;

[0053] The number of missing tokens is defined as:

[0054] ;

[0055] The number of remaining tokens is defined as:

[0056] ;

[0057] The deviation result includes the transitions that are not normally enabled during the process execution and the missing token information;

[0058] The judgment rule for the transitions that are not normally enabled is:

[0059] When , binding is enabled under the marking , and transition executes according to the enabled binding under the marking to obtain a new marking , and the calculation formula is:

[0060] ;

[0061] When , binding cannot be in the marking Enabled downward, transition is a transition that is not enabled properly;

[0062] Missing token information is defined as:

[0063] ;

[0064] Missing token information is for the transition in the place where the objects required for proper enabling are missing .

[0065] Furthermore, the specific process of step 4 is as follows:

[0066] Step 4.1, Calculate the fitness, the formula is as follows:

[0067] ;

[0068] where, is the fitness; represents the number of process executions in the object-centered business process event log; represents the number of tokens missing in the th process execution during replay; represents the number of tokens consumed in the th process execution during replay; represents the number of tokens remaining in the th process execution during replay; represents the number of tokens generated in the th process execution during replay;

[0069] Quantify compliance based on the fitness. The closer the fitness is to 1, the better the compliance and the less the deviation;

[0070] Step 4.2, Use the local diagnosis method to calculate the local diagnosis result, the formula is:

[0071] ;

[0072] ;

[0073] where, is the local diagnosis result of the non-properly enabled transition consistency for the process execution ; is the local diagnosis result of the non-properly enabled transition consistency for the object-centered business process event log ; is the non-properly enabled transition in; is the set of non-properly enabled transitions; is the number of tokens missing in the process execution replay; is the number of tokens consumed in the process execution replay; represents the number of process executions of the corresponding log activity included in the object - centered business process event log and is calculated by the formula: ;

[0074] ;

[0075] where is the corresponding log activity;

[0076] The replay result, deviation result, fitness result, and local diagnosis result are the required object - centered business process violation inspection results.

[0077] An object - centered business process violation inspection system based on token replay adopts the object - centered business process violation inspection method described above. The input of this system is the object - centered business process event log and the multi - object Petri net, and the output is the object - centered business process violation inspection result. This system includes the following modules:

[0078] The object graph and process execution acquisition module is used to obtain the object graph according to the dependency relationship between objects in the object - centered business process event log, and then use the connected component extraction method to obtain the process execution from the object graph;

[0079] The binding sequence acquisition module is used to delete the cross - layer redundant dependencies in the process execution according to the transitive reduction method to obtain the simplified process execution, and then use the echelon division mechanism to divide the simplified process execution to obtain the binding sequence;

[0080] The replay result and deviation result acquisition module is used to perform token replay on the multi - object Petri net model according to the binding execution rule to obtain the replay result and deviation result in the object - centered business process violation inspection result;

[0081] The object - centered business process violation inspection result acquisition module is used to substitute the obtained replay result and deviation result into the defined fitness formula to quantify the compliance, and then use the local diagnosis method to display the parts with serious deviations in the process to obtain the fitness result and local diagnosis result parts in the object - centered business process violation inspection result.

[0082] Beneficial technical effects brought by the present invention: For the object-centered business process, the present invention provides a solution, which has been successfully applied to solve the problem of compliance inspection of the object-centered business process and is feasible; for the problem that the existing compliance inspection technology needs to flatten the object-centered business process and the object interaction information is lost, the method of the present invention can fully extract the interaction relationship between objects to discover the actual process execution and is comprehensive; for the problem that the existing compliance inspection technology cannot accurately identify deviations and it is difficult to confirm the source of deviations, the method of the present invention can accurately identify non-compliant behaviors and reduce the possibility of misjudgment or missed judgment, and is accurate; for the problem that the existing compliance inspection technology cannot accurately identify deviations and it is difficult to confirm the source of deviations, the method of the present invention can quickly locate the problem link through quantifying deviations and local diagnosis and correct the problems in the process execution in a timely manner, and is efficient; for the problem that the existing compliance inspection technology is difficult to quantify the deviations of the object-centered business process, the method of the present invention uses the goodness of fit to quantify the deviations and uses the local diagnosis method to show the severity of the deviations in the non-fitting process execution, and can flexibly meet the compliance inspection requirements of different object-centered business processes, and has adaptability and scalability; the present invention can be applied to object-centered business scenarios or fields, such as machine equipment manufacturing processes, fan maintenance processes, medical business processes, service processes, etc., and has high application value and broad prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Figure 1 It is a flowchart of the object-centered business process violation inspection method based on token replay of the present invention.

[0084] Figure 2 It is an object graph obtained by using the method of the present invention in an embodiment of the present invention.

[0085] Figure 3 It is a process execution graph obtained by using the method of the present invention in an embodiment of the present invention.

[0086] Figure 4 It is a process execution graph after deleting cross-layer redundant dependencies obtained by using the method of the present invention in an embodiment of the present invention.

[0087] Figure 5 It is a standard process model graph of the procurement and sales scenario in an embodiment of the present invention.

[0088] Figure 6 It is a multi-object Petri net graph of the standard process model of the procurement and sales scenario obtained by using the method of the present invention in an embodiment of the present invention.

[0089] Figure 7 It is an architecture graph of the object-centered business process violation inspection system based on token replay of the present invention.

[0090] Figure 8This is the multi-object Petri net diagram of the shipping order object type model in the comparative experiment of the present invention. Detailed implementation manners

[0091] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners:

[0092] The flowchart of the present invention is as Figure 1 shown. Starting from the object-centered business process event log, first obtain the object graph and process execution, secondly obtain the binding sequence, then obtain the token replay result and deviation result, then quantify the compliance and display the serious part of the deviation, and finally output the object-centered business process violation inspection result. The present invention also designs an object-centered business process violation inspection system based on token replay applied to the object-centered business process according to the above process. The main functional modules of the system include: an object graph and process execution acquisition module, a binding sequence acquisition module, a replay result and deviation result acquisition module, and an object-centered business process violation inspection result acquisition module.

[0093] An object-centered business process violation inspection method based on token replay includes the following steps:

[0094] Step 1: Obtain the object-centered business process event log, analyze the dependency relationship between objects to construct an object graph, and then use the connected component extraction method to obtain the process execution. The specific process is as follows:

[0095] Step 1.1: The obtained object-centered business process event log consists of a set of events representing operations, and the events include activities, occurrence times, end times, and basic attributes of associated objects;

[0096] Step 1.2: The object graph is an undirected graph describing the dependency relationship between objects in the object-centered business process event log. By analyzing the dependency relationship between objects, the object graph of the object-centered business process event log is as follows:

[0097] ;

[0098] Among them, is the set of undirected graph nodes, consisting of all objects in the object-centered business process event log; is the set of undirected graph edges, consisting of several pairs of objects with a dependency relationship; The calculation formula of

[0099] is:

[0100] Among them, , are events Two different associated objects; The represented object is all events in the central business process event log; The represented object is the central business process event log and all objects associated with the event in it, and the calculation formula is:

[0101] ;

[0102] Among them, represents the event sequence associated with the object , and the specific mapping relationship is:

[0103] ;

[0104] Step 1.3, the connected component extraction method is a method for obtaining the process execution from all the maximum connected subgraphs of the object graph. The process execution is a directed graph that describes the dependencies and interaction information between the associated process events based on the dependencies between the objects in the maximum connected subgraph. The process of using the connected component extraction method to obtain the process execution in the present invention is as follows:

[0105] First, formalize the connected component extraction method as:

[0106] ;

[0107] Among them, is the connected component extraction method; is the maximum connected subgraph of the object graph ; represents the process execution obtained from the maximum connected subgraph; is the set of objects in the maximum connected subgraph;

[0108] Then, obtain the process execution according to the directed graph, and define as:

[0109] ;

[0110] Among them, is the set of directed graph nodes, which is composed of all events associated with any object in the central business process event log ; is the set of directed graph edges, which is composed of several pairs of events with a direct following relationship;

[0111] 、 The calculation formulas of are respectively:

[0112] ;

[0113] ;

[0114] Among them, represents the direct following relationship between events in the object-centered business process event log , and the calculation formula is:

[0115] ;

[0116] Among them, is an event different from event ; is the th event; is the th event; is the th event.

[0117] The method of the present invention is applicable to the object-centered business process, that is, for any given business process, as long as the business process belongs to the object-centered business process, then the business process can use the method of the present invention for business process violation inspection. For example, the standard process of procurement and sales is an object-centered business process. In the standard process of procurement and sales, as shown in Table 1, the object-centered business process event log L1 contains 9 activities (i.e., corresponding to the 9 events from to ), 5 object types, and 7 objects. The activities are as follows: create customer order, create supplier order, receive supplier package, open package and put into storage, create shipping note, pack and ship, deliver, receive payment, empty shipping note. The object types are as follows: customer order (Order), supplier order (Supplier Order), shipping note (Invoice), product details (Item), payment (Payment). The objects are as follows: customer order object O1, supplier order object A, shipping note object I1, product details object X1, product details object X2, product details object X3, payment object P1.

[0118] Table 1 Object-centered business process event log L1

[0119] .

[0120] The object graph obtained by using the method of the present invention As Figure 2 shown, its undirected graph node set is ={A, X1, X2, X3, O1, I1, P1}, and the undirected graph edge set is = {{A, X1, X2, X3}, {O1, X1, X2, X3}, {O1, I1}, {I1, P1}}; Since the object graph itself is a maximum connected subgraph, only one process execution can be obtained using the method of the present invention , as Figure 3 shown, the set of directed graph nodes is , and the set of directed graph edges is .

[0121] Step 2. According to the process execution obtained in Step 1, use the transitive reduction method to delete cross-layer redundant dependencies, and then use the echelon partitioning mechanism to obtain the binding sequence. The specific process is as follows:

[0122] Step 2.1. Use the transitive reduction method to delete cross-layer redundant dependencies; the specific process is as follows: Delete the cross-layer redundant dependency relationships in the object graph to obtain a minimum subgraph that maintains the transitive closure property of the original graph. The cross-layer redundant dependency relationships refer to the node-to-node dependency relationships in the object graph that are realized through indirect paths. The maintaining of the transitive closure property of the original graph means that the reachability relationship between nodes in the graph remains unchanged after deleting the redundant dependencies;

[0123] Step 2.2. Use the echelon partitioning mechanism to partition the process execution; the specific process is as follows: Iteratively implement the predecessor-free node recognition algorithm in , and each time obtain all the current predecessor-free nodes as an echelon, partition the process execution into several logically independent event echelons. The echelons are unordered within, and follow the established sequential logic between echelons; the predecessor-free node recognition algorithm is to identify the nodes that are currently predecessor-free in the context of the graph and store them in a list;

[0124] Step 2.3 Map each logically independent event echelon to its respective binding to obtain the corresponding binding sequence; define the binding sequence as:

[0125] ;

[0126] wherein, is the th transition, is the specific object of each object type (consistent with the object type of the pre-set place of this transition) consumed when the th transition is executed. The number of transitions is the same as the number of events, that is ; is the binding corresponding to the th transition, that is, if the pre-set place of this transition contains For the objects of each object type, binding is enabled and this transition is in a pending execution state; enabling means that when the conditions for the transition to execute are met, the transition is in a state where it can occur.

[0127] The process execution after deleting cross-layer redundant dependencies obtained by using the method of the present invention is as follows Figure 4 shown, that is, due to event and event can be reached through event so the directed graph edges causing cross-layer redundant dependencies are deleted using the transitive reduction method ; The process execution after deleting cross-layer redundant dependencies is processed using the echelon division mechanism, and the divided event echelon is , and the obtained corresponding binding sequence echelon is: [[('CO', {'Order': ['O1']}), ('SO', {'Supplier Order': ['A']}), ('RT', {'Payment': ['P1']})], [('RS', {'Item': ['X1', 'X2', 'X3'], 'Supplier Order': ['A']}), ('CI', {'Invoice': ['I1'], 'Order': ['O1']})], [('UP', {'Item': ['X1'], 'Supplier Order': ['A']}), ('RI', {'Payment': ['P1'], 'Invoice': ['I1']})],[('PS', {'Item': ['X1', 'X2', 'X3'], 'Order': ['O1']})], [('SP', {'Order': ['O1']})]].

[0128] Step 3. According to the binding sequence obtained in Step 2, use the binding execution rule to perform token replay on the multi-object Petri net to obtain the replay result and the deviation result. The specific process is as follows:

[0129] Step 3.1. Define the multi-object Petri net as a six-tuple, specifically:

[0130] ;

[0131] Among them, is the Petri net; is the mapping function from the place to the object type; is the set of variable arcs, is the set of all arcs of the multi-object Petri net; is the cardinality label function of the arc; and represent the start identifier and the end identifier respectively. The identifier is a multiset of tokens, defined as ; is the identifier; is a multiset of tokens; The specific mapping relationship of is:

[0132] ;

[0133] Among them, is the set of places, is the set of object types; The specific mapping relationship of is:

[0134] ;

[0135] Among them, the symbol " " is a variable arc label, indicating that this arc can transport one or more tokens each time. The definition of the token is , indicating that there is an object in the place ; is the token set of the multi-object Petri net , and the calculation formula is:

[0136] ;

[0137] Among them, is the set of objects; is the object type;

[0138] Step 3.2: Based on the binding sequence, perform token replay on the multi-object Petri net according to the binding execution rules for the process execution, and obtain the replay result and deviation result of each process execution;

[0139] The replay result is the number of tokens generated, consumed, missing, and remaining after token replay for each process execution;

[0140] The number of tokens generated is:

[0141] ;

[0142] Among them, is the set of post-set places of the transition ; is the specific object of each object type consumed during the execution of the transition; is the binding that needs to be satisfied for the execution of the transition ;

[0143] The number of tokens consumed is:

[0144] ;

[0145] Among them, is the pre-set place set of the transition ;

[0146] The missing token number is:

[0147] ;

[0148] The remaining token number is:

[0149] ;

[0150] The deviation result is the transitions that are not normally enabled during the process execution and the missing token information added to ensure its smooth execution;

[0151] The judgment rule for the transitions that are not normally enabled is:

[0152] When , the binding is enabled under the marking , and the transition executes according to the enabled binding under the marking to obtain a new marking , that is , and the specific calculation formula is:

[0153] ;

[0154] When , the binding cannot be enabled under the marking , and the transition is a transition that is not normally enabled;

[0155] The missing token information is defined as:

[0156] ;

[0157] The missing token information means that the transition lacks the object required for normal enabling by binding in the place .

[0158] The standard process model of the procurement and sales scenario is as Figure 5As shown, there are five types of objects in this process, namely product details (i.e., Item), supplier (Supplier Order), customer order (Order), invoice (Invoice), and payment (Payment). The places and transitions of the business process for each object type are identified with a unique background color. The label of the place indicates its object type, and the label of the transition indicates its corresponding log activity. At the same time, the activities shared by two object types use two background colors; Figure 5 In it, Item1, Item2, Item3, and Item4 are four product details; Supplier Orde1, SupplierOrde2, Supplier Orde3, and Supplier Orde4 are four suppliers; Orde1, Orde2, Orde3, and Orde4 are four customer orders; Invoice1, Invoice2, and Invoice3 are three invoices; Payment1, Payment2, and Payment3 are three payments; SO is to create a supplier order; CO is to create a customer order; RT is to receive payment; RS is to receive a supplier package; CI is to create a shipping note; UP is to open the package; RI is to empty the shipping note; PS is to pack and ship; SP is to deliver.

[0159] Using the method of the present invention in Figure 5 The token replay is performed on the model, and the multi-object Petri net of the standard process model for the purchase and sales scenario is as Figure 6 shown, Figure 6 The binding sequence of the standard process model for the purchase and sales scenario is as shown in Table 2 below:

[0160] Table 2 Binding Sequence Table of the Standard Process Model for the Purchase and Sales Scenario

[0161] .

[0162] According to the binding sequence in Table 2, the token replay is performed on the Figure 6 multi-object Petri net, and the number of generated tokens is 25, the number of consumed tokens is 25, the number of missing tokens is 2, and the number of remaining tokens is 2 (i.e., there are two remaining objects X2 and X3 at the Item2 place).

[0163] Due to the existence of missing tokens and remaining tokens, when the token replay is performed on the standard process model for the purchase and sales scenario during this process execution, it cannot be fully fitted. The deviation results are shown in Table 3. The deviation results shown in Table 3 include three parts: the process execution serial number, the transitions that are not normally enabled, and the information on adding missing tokens, indicating that there is a transition PS (i.e., pack and ship) that is not normally enabled in process execution 1. The reason for the abnormal enabling is that there are two X products (i.e., X3 and X2) that should be packed and shipped missing in the PS link. The specific content of Table 3 is as follows:

[0164] Table 3 Deviation Result Table

[0165] 。

[0166] Step 4. According to the replay result and deviation result obtained in Step 3, use the fitness formula to quantify compliance, and then use the local diagnosis method to display the parts with serious deviations in the process, and obtain the inspection result of the violation of the object-centered business process. The inspection result of the violation of the object-centered business process includes not only the replay result and deviation result, but also the fitness result and local diagnosis result; the specific calculation processes of the fitness result and local diagnosis result are as follows:

[0167] Step 4.1. Define the fitness formula as follows:

[0168] ;

[0169] Among them, is the fitness; represents the number of process executions in the object-centered business process event log; represents the number of tokens missing in the replay of the th process execution; represents the number of tokens consumed in the replay of the th process execution; represents the number of tokens remaining in the replay of the th process execution; represents the number of tokens generated in the replay of the th process execution;

[0170] Quantify compliance based on the fitness value. The closer the fitness is to 1, the better the compliance and the fewer the deviations.

[0171] Step 4.2. Use the local diagnosis method to calculate the local diagnosis result; the local diagnosis method refers to focusing on key transitions, diagnosing the number of abnormal objects in the "added after missing" part among all objects consumed by the process execution in the places required for key transitions, and identifying and analyzing possible deviations in the process, mainly including two quantification methods: the consistency of unnormally enabled transitions for a single process execution and the consistency of unnormally enabled transitions in the event log. The former is for a single process execution, and the latter is for the entire event log. The calculation methods are as follows:

[0172] ;

[0173] ;

[0174] Among them, is the local diagnosis result of the consistency of unnormally enabled transitions for the process execution ; Event log for object-centric business processes The local diagnosis results of the abnormally enabled transition consistency; yes Changes that are not enabled normally; It is a set of transitions that are not enabled normally; yes In process execution The number of tokens missing in the replay; yes In process execution The number of tokens consumed in the replay; Represents an object-centric business process event log Included The number of process executions corresponding to the log activity is calculated as follows:

[0175] ;

[0176] in, for Corresponding log activities;

[0177] Will Figure 6 After the replay, the number of generated tokens 25, the number of consumed tokens 25, the number of missing tokens 2, and the number of remaining tokens 2 are substituted into the fitting formula and the result is Substituting into the fitting formula, the fitting result is 0.92; the local diagnosis result obtained by the method of the present invention includes two parts: the consistency of abnormally enabled transitions executed by a single process and the consistency of abnormally enabled transitions in the event log. Among them, the local diagnosis result of the consistency of abnormally enabled transitions executed by a single process is 0.5. However, the embodiment only includes one process execution, so the local diagnosis result of the consistency of abnormally enabled transitions in the event log is also 0.5, that is, when executing the transition PS, half of the objects are abnormal objects that do not exist in the previous set library and are additionally added, indicating that the degree of deviation of this activity in the actual execution of the process is relatively serious and should be checked and adjusted in time.

[0178] Based on the above method, the present invention designs an object-centric business process violation checking system based on token replay. The object-centric business process event log and the standard multi-object Petri net are input into the system, and the object-centric business process violation checking result is obtained as output, such as Figure 7 As shown, the system includes the following modules:

[0179] The object graph acquisition and process execution module is used to acquire the object graph according to the dependencies between objects in the object-centric business process event log, and then obtain the process execution from the object graph using the connected component extraction method;

[0180] The binding sequence acquisition module is used to delete cross-layer redundant dependencies in the process execution according to the transfer reduction method, obtain the simplified process execution, and then use the echelon division mechanism to divide the simplified process execution to obtain the binding sequence;

[0181] The replay result and deviation result acquisition module is used to perform token replay on the multi-object Petri net model according to the binding execution rule to obtain the replay result and deviation result in the object-centered business process violation inspection result;

[0182] The object-centered business process violation inspection result acquisition module is used to substitute the obtained replay result and deviation result into the defined fitness formula to quantify compliance, and then use the local diagnosis method to display the parts with serious deviations in the process, and obtain the fitness result and local diagnosis result parts in the object-centered business process violation inspection result.

[0183] The present invention also designs a storage medium storing a program, and when the program is executed by a processor, it implements the above-mentioned object-centered business process violation inspection method based on token replay and the object-centered business process violation inspection system based on token replay.

[0184] The present invention also designs a computing device, including a processor and a memory for storing a program executable by the processor. When the processor executes the program stored in the memory, it implements the above-mentioned object-centered business process violation inspection method based on token replay and the object-centered business process violation inspection system based on token replay. The computing device can be a desktop computer, a laptop computer, a smart phone, a PDA handheld terminal, a tablet computer, a programmable logic controller (PLC), or other terminal devices with processor functions.

[0185] The present invention provides a new solution to the problems that existing compliance inspection technologies cannot accurately identify deviations and are difficult to confirm the sources of deviations when dealing with object-centered business processes. It can accurately identify non-compliant behaviors, reduce the possibility of misjudgment or missed judgment, make the object-centered compliance inspection technology more accurate, and use the fitness to quantify deviations and use the local diagnosis method to display the severity of deviations in the non-fitting process execution, making it more adaptable and scalable. In addition, it can also be applied to fields such as machine equipment manufacturing processes, medical business processes, and service processes, and has practical promotion value and is worthy of promotion.

[0186] To prove the feasibility and superiority of the present invention, the following comparative experiments are given.

[0187] The comparative experiment uses the purchase and sales event log shown in Table 1, and this process is an object-centered business process event log.

[0188] Compare and analyze the method of the present invention with the existing flat token replay compliance checking technology. Figure 6 For the multi-object Petri net of the standard process model of the procurement and sales scenario using the present invention, the replay result is used as a comparison index. If the replay result can fully reflect the existing deviations, it indicates that the compliance checking method is better. Through token replay, the method of the present invention fully reflects the existing deviations. Table 4 shows the event log flattened by the shipping order object type:[[]]END]]

[0189] Table 4 Event log flattened by the shipping order object type

[0190] .

[0191] Figure 8 For the multi-object Petri net of the shipping order object type model, using the existing flat token replay compliance checking technology to replay the log in Table 4 on the shipping order object type model, the binding sequence table of the shipping order object type model is shown in Table 5:[[]]END]]

[0192] Table 5 Binding sequence table of the shipping order object type model

[0193] .

[0194] Replay the binding sequence in Table 5 on the Figure 8 multi-object Petri net, and the generated token number is 3, the consumed token number is 3, the missing token number is 0, and the remaining token number is 0; due to the lack of dependency information between objects, this method cannot detect that two X commodities are missing when the commodity details object type executes the PS activity. Therefore, the compliance checking of the method of the present invention is more comprehensive and accurate. This can prove the feasibility and superiority of the present invention.

[0195] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by those skilled in the art within the scope of the essence of the present invention should also fall within the protection scope of the present invention.

Claims

1. An object-centric business process violation detection method based on token replay, characterized in that: The steps include: Step 1: Obtain object-centric business process event logs, analyze dependencies between objects to construct object graphs, and then use connected component extraction methods to obtain process execution; Step 2: Use the transitive reduction method to remove cross-layer redundant dependencies, and then use the echelon partitioning mechanism to obtain the binding sequence; Step 3: Use the binding execution rules to perform token replay on the multi-object Petri net to obtain the replay result and deviation result; Step 4: Calculate the fitting result and the local diagnosis result to obtain the violation inspection result of the object-centered business process; The specific process of step 1 is as follows: Step 1.1, the object-centric business process event log L obtained consists of a set of events representing operations, wherein the event includes activities, occurrence time, end time and basic attributes of associated objects; Step 1.2: By analyzing the dependencies between objects, the object graph of the object-centric business process event log L is constructed as follows: G L =(O,C O ); Where O is an undirected graph node set consisting of all objects in the object-centric business process event log; C O It is an undirected graph edge set, consisting of several pairs of objects with dependency relationships; C O The calculation formula is: Among them, o1 and o2 are two different objects associated with event e; E indicates that the object is all events in the central business process event log; obj L (e) represents all objects associated with event e in the event log L of the central business process. The calculation formula is: obj L (e)={o∈O|e∈trace(o)}; Among them, trace(o) represents the event sequence associated with object o; Step 1.3: Use the connected component extraction method to obtain process execution; the specific process is as follows: First, the formal connected component extraction method is: Among them, ext comp (·) is the connected component extraction method; pO′ represents the process execution obtained from the maximum connected subgraph; O′ is the object set in the maximum connected subgraph; Then, according to the directed graph of O′, the process execution is obtained, and pO′ is defined as follows: p O′ =(E′,D); Where E′ is a directed graph node set consisting of all events in the object-centric business process event log associated with any object in O′; D is a directed graph edge set consisting of several pairs of events with direct follow-up relationships; The calculation formulas of E′ and D are: D=with L ∩(E′×E′); in, is an empty set; con L The object represents the direct follow-up relationship between events in the central business process event log L. The calculation formula is: Where e′ is an event different from event e; e n is the nth event; e i is the i-th event; e i+1 is the i+1th event; The specific process of step 2 is as follows: Step 2.1, using the transitive reduction method to delete the cross-layer redundant dependencies in the object graph, and obtain a minimum subgraph that maintains the transitive closure characteristics of the original graph; the cross-layer redundant dependencies refer to the dependencies between nodes in the object graph realized by indirect paths, and maintaining the transitive closure characteristics of the original graph means that the reachability relationship between nodes in the graph remains unchanged after deleting the redundant dependencies; Step 2.2: Use the echelon division mechanism to divide the process execution; the specific process is: O′ The no-predecessor node identification algorithm is iteratively implemented, and each time all the current no-predecessor nodes are obtained as an echelon, and the process execution is divided into several logically independent event echelons, which are disordered within the echelon and follow the established sequential logic between echelons; the no-predecessor node identification algorithm is to identify the nodes that currently have no predecessor nodes in the context of the graph and store them in a list; Step 2.3 maps each logically independent event echelon to its own binding to obtain the corresponding binding sequence; the binding sequence σ is defined as follows: in, For the A change, For the The specific objects of each object type consumed when a transition is executed; For the The binding corresponding to each transition; The specific process of step 3 is as follows: Step 3.1, define the multi-object Petri net MOPN as follows: Where N is a Petri net; It is the mapping function from library to object type; is a variable arc set, F is the set of all arcs in the multi-object Petri net; W is the cardinality labeling function of the arc; M init and M final Represent the start mark and end mark respectively; The specific mapping relationship is: Among them, P is the set of places, is a collection of object types; Step 3.2: Based on the binding sequence, the process execution is performed on the multi-object Petri net according to the binding execution rules to perform token replay, and the replay result and deviation result of each process execution are obtained; The replay result is the number of tokens generated, consumed, missing, and remaining after token replay for each process execution; The number of tokens produced prod is defined as: Among them, t· is the set of post-set libraries of transition t; b is the specific object of each object type consumed when the transition is executed; Q MOPN It is the token set of the multi-object Petri net MOPN; Token, which means the library There is an object o in The number of consumed tokens cons is defined as: Among them, ·t is the set of the previous set of transition t; The number of missing tokens mis is defined as: The remaining token number rem is defined as: rem(t,b)=mis(t,b); Deviation results include transitions that are not enabled normally during process execution and missing token information; The judgment rules for transitions that are not enabled normally are: When cons(t,b)≤M, the binding (t,b) is enabled under the flag M, and the transition t is executed under the flag M according to the enabled binding to obtain the new flag M′. The calculation formula is: M′=M-cons(t,b)+prod(t,b); When cons(t,b)>M, the binding (t,b) cannot be enabled under the flag M, and the transition t is an unenabled transition. Missing token informationinfo miss Defined as: Missing token information is the transition t in the library The object o required for normal binding is missing; The specific process of step 4 is as follows: Step 4.1, calculate the degree of fit, the formula is as follows: Where fit(·) is the fit; K represents the number of process executions in the event log of the object-centric business process; mis k represents the number of tokens missing in the replay of the k-th process execution; k Represents the number of tokens consumed by the kth process execution in the replay; rem k represents the number of tokens remaining in the replay of the kth process execution; prod k represents the number of tokens generated by the k-th process execution in the replay; Compliance is quantified based on the fit measure, and the closer the fit is to 1, the better the compliance and the less the deviation; Step 4.2: Use the local diagnosis method to calculate the local diagnosis result. The formula is: Among them, dia p (·) is the local diagnosis result of the abnormally enabled transition consistency of process execution p; L (·) is the local diagnosis result of the abnormally enabled transition consistency of the object-centric business process event log L; t un ∈T un It is a transition that is not normally enabled in MOPN; T un The transition set is not enabled normally; mis p (t un ) is t un The number of tokens missing in the replay of process execution p; cons p (t un ) is t un The number of tokens consumed in the process execution p replay; Indicates that the object is a central business process event log L containing t un The number of process executions corresponding to the log activity is calculated as follows: Among them, acti(t un ) is t un Corresponding log activities; The replay results, deviation results, fit results and local diagnosis results are the required object-centric business process violation inspection results.

2. An object-centric business process violation checking system based on token replay, characterized in that: The object-centric business process violation checking method based on token replay as claimed in claim 1 is adopted, the input of the system is the object-centric business process event log and the multi-object Petri net, and the output is the object-centric business process violation checking result; the system includes the following modules: The object graph acquisition and process execution module is used to acquire the object graph according to the dependencies between objects in the object-centric business process event log, and then obtain the process execution from the object graph using the connected component extraction method; The binding sequence acquisition module is used to delete cross-layer redundant dependencies in the process execution according to the transitive reduction method to obtain a simplified process execution, and then use the echelon partitioning mechanism to divide the simplified process execution to obtain a binding sequence; A module for obtaining replay results and deviation results is used to perform token replay on a multi-object Petri net model according to binding execution rules, and obtain replay results and deviation results in the violation inspection results of the object-centric business process; The object to be acquired is the central business process violation inspection result module, which is used to substitute the obtained replay results and deviation results into the defined fit formula to quantify compliance, and then use the local diagnosis method to display the part of the process with serious deviations. The object to be acquired is the fit results and local diagnosis results in the central business process violation inspection results.

Images