Business process instantiation dynamic generation method and device

Abstract

The application relates to the field of information technology, and discloses a business process instantiation dynamic generation method and device, which comprises the following steps: in response to a business application of a self-service terminal, obtaining a business type and a structured data packet; according to the business type, calling a standard process template predefined in a directed graph; analyzing the structured data packet to extract an instantiation condition factor; matching the instantiation condition factor with a condition response rule set of the standard process template to obtain a matching rule; and according to the matching rule, dynamically reconstructing a directed graph structure of the standard process template, so as to obtain an instantiation process model matched with the business application. Through real-time analysis of user materials and intelligent reconstruction of a process model, the application realizes accurate adaptation of one case to one process, eliminates redundancy from the source, optimizes a path, and significantly improves business handling efficiency and a personalized service level.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of information technology, in particular to a business process instantiation dynamic generation method and device. BACKGROUND

[0002] Government service hall widely deploys self-service terminals, and gradually extends to community, industrial park and other places, aiming to provide citizens with nearby and self-service convenience. This measure effectively diverts the flow of the hall and relieves the pressure of queuing. However, when the service tentacles extend to complex businesses, such as business start-up, project approval and other series processes involving multiple departments and links, the existing system exposes its deep structural bottleneck.

[0003] At present, whether it is the sequential handling of offline windows or the electronic flow of online systems or self-service terminals, its essence is a kind of solid assembly line mode. The system regards each complex business as a preset and unchangeable task sequence, and strictly pushes each link according to the established order; although the self-service terminal allows users to remotely initiate the process, the subsequent approval flow is still trapped in the rigid order chain. The fundamental defect of this mode is that it cannot perceive the specific content and state of the user's submitted business, and applies the same complete process template to all applications. For example, even if the user has completed and uploaded the "name pre-approval" certificate, the system will mechanically retain this link, resulting in invalid link flow and resource consumption; at the same time, it cannot dynamically adjust the priority or dependency of the link according to the urgency or specific attributes of the business (such as the green channel for small and micro enterprises).

[0004] Therefore, although the existing technology optimizes the entry of the application, it fails to touch the intelligent reconstruction of the process core, and the overall handling time of the complex business still depends on the accumulation of all links, and any delay of any link will block the entire chain. SUMMARY

[0005] Therefore, the technical problem to be solved by the present application is to overcome the defects of the existing complex business process rigidity and inability to dynamically adjust according to the user's personalized application content, thereby providing a business process instantiation dynamic generation method and device, intelligently reconstructing the process model according to the real-time analysis of the user's submitted materials, realizing the precise adaptation of one case to one process, eliminating redundant links and optimizing the processing path from the source, and significantly improving the business handling efficiency and personalized service level.

[0006] In the first aspect, in order to solve the above technical problems, the present application provides a business process instantiation dynamic generation method, comprising:

[0007] In response to the business application of the self-service terminal, obtaining the business type and the structured data package of the business application;

[0008] According to the business type, a corresponding standard process template is called, and the standard process template is predefined in a directed graph structure; wherein the nodes of the directed graph structure represent atomic processing nodes of the business, and the edges represent the dependency relationship between the nodes;

[0009] The structured data packet is parsed to extract at least one instantiated conditional factor;

[0010] The instantiated conditional factor is matched with a conditional response rule set of the standard process template to obtain a matching rule;

[0011] According to the matching rule, the directed graph structure of the standard process template is dynamically reconstructed to obtain an instantiated process model matched with the business application.

[0012] Preferably, parsing the structured data packet to extract at least one instantiated conditional factor comprises: parsing the structured data packet to obtain a set of atomic data items containing a plurality of atomic data items; inputting each atomic data item into a factor value prediction model to obtain a factor value prediction value corresponding to each atomic data item; wherein the factor value prediction model is a machine learning model obtained by training based on historical business data, each sample in the training data of the factor value prediction model includes a historical atomic data item and a corresponding process reconstruction association label; the process reconstruction association label represents the influence degree of the historical atomic data item on the reconstruction of the process template in the historical business instance; according to the factor value prediction value, the set of atomic data items is screened to obtain one or more effective atomic data items; and the effective atomic data items are formatted into the instantiated conditional factor.

[0013] Preferably, after obtaining the factor value prediction value corresponding to each atomic data item, it further comprises: determining a scene category to which the business application belongs based on the business type; loading a predefined scene weight configuration table according to the scene category, wherein a scene adjustment coefficient is defined for each type of atomic data item in the scene weight configuration table; using the scene adjustment coefficient to correct the factor value prediction value to obtain a corrected prediction value; and using the corrected prediction value to replace the factor value prediction value.

[0014] Preferably, matching the instantiated conditional factor with the conditional response rule set of the standard process template to obtain a matching rule comprises: according to the business type, reading a corresponding conditional response rule subset from a rule storage; for each rule in the conditional response rule subset, parsing the trigger condition of the rule to obtain a set of sub-condition constraints; comparing each sub-condition constraint with the instantiated conditional factor to determine whether there is an instantiated conditional factor that satisfies the sub-condition constraint; if the set of sub-condition constraints are all satisfied, it is determined that the rule is matched successfully, and the matching rule is obtained.

[0015] Preferably, the condition-response rule subset is stored in a structured manner based on the trigger condition of the rule; wherein each rule is associated with at least one index key, which is determined according to the type of at least one instantiated condition factor that must appear in the trigger condition of the rule.

[0016] Preferably, the dynamic reconstruction operation includes at least one of the following: if the instantiated condition factor indicates that any of the atomic processing nodes has been completed, removing the corresponding node and all associated edges from the directed graph structure; if the instantiated condition factor includes user-specified priority information, adding an attribute label to the corresponding node.

[0017] Preferably, after removing the corresponding node and all associated edges from the directed graph structure, further comprising: identifying all upstream and downstream dependent nodes of the removed node; for each pair of nodes consisting of an upstream dependent node and a downstream dependent node, performing the following judgment: judging whether there is a direct connection between the node pair, if not, further judging whether the removed node constitutes the only dependent edge of the node pair in the standard process template;

[0018] if not, creating a new dependent edge between the upstream dependent node and the downstream dependent node constituting the node pair; if yes, marking the downstream dependent node of the node pair as a node to be manually checked.

[0019] Preferably, when scheduling tasks based on the instantiated process model, reading the attribute label attached to the node and mapping the attribute label to a scheduling instruction; the scheduling instruction includes raising the priority of the task queue, allocating specific execution resources, setting the processing deadline or adding task pre-check items.

[0020] Preferably, before dynamically reconstructing the directed graph structure of the standard process template according to the matching rule, further comprising: analyzing the reconstruction operation indicated by the matching rule, if there is a pair of mutually exclusive operations for the same graph element, triggering conflict resolution.

[0021] The second aspect, to solve the above technical problems, the present application provides a kind of business process instantiation dynamic generation device, for executing the business process instantiation dynamic generation method, comprising:

[0022] Application receiving module receives the business application of self-service terminal, obtains business type and the structured data package of the business application;

[0023] a template management module, which invokes a corresponding standard flow template according to the business type, the standard flow template being predefined in a directed graph structure; wherein a node of the directed graph structure represents an atomic processing node of a business, and an edge represents a dependency relationship between nodes;

[0024] a factor extraction module, which parses the structured data packet and extracts at least one instantiated condition factor;

[0025] a rule matching module, which matches the instantiated condition factor with a condition response rule set of the standard flow template to obtain a matching rule;

[0026] a flow reconstruction engine, which dynamically reconstructs the directed graph structure of the standard flow template according to the matching rule to obtain an instantiated flow model matched with the business application.

[0027] The above technical solution of the present application has the following beneficial effects compared with the prior art:

[0028] The business flow instantiation dynamic generation method and device according to the present application intelligently reconstruct a flow model according to real-time analysis of user submission materials, realizes precise adaptation of one case to one flow, eliminates redundant links and optimizes processing paths from the source, and significantly improves business handling efficiency and individualized service level.

[0029] Firstly, the traditional government affair system mechanically applies a complete flow to all applicants, regardless of the user material preparation state and individualized needs, resulting in a large amount of invalid circulation and waiting. After receiving a business application, the present application actively parses a structured data packet, extracts instantiated condition factors reflecting the current business-specific state, such as “material pre-audit completion” and “user selects urgent handling”, and matches these factors with preset business rules; after successful matching, the standard flow template is dynamically reconstructed according to the rules to generate a flow model completely matching the specific circumstances of the current applicant. This technical path directly realizes precise adaptation of the business flow from the source, so that each user obtains the most optimal handling guide, fundamentally overcomes the disadvantages of rigid flow, and greatly improves the individualized level and user satisfaction of government affair service.

[0030] Secondly, the dynamic reconstruction operation of the present application can eliminate redundant links from the logical level, directly optimize the processing path, thereby significantly compressing the overall handling time. The prior art, even if it allows to skip some links, often makes the judgment during execution, and cannot simplify from the process definition. The present application is different, its dynamic reconstruction is to directly edit the directed graph structure of the standard process template, for example, when it is identified that a certain pre-approval has been completed, the system will completely remove the corresponding node and its dependent relationship from the flowchart; this is not just skipping the link, but deleting this task from the process logic model, so that all subsequent scheduling and execution are based on this more simplified model. From the source, eliminate invalid task assignment, state monitoring and data flow conversion overhead. The technical effect of the present application is direct and significant: it generates a personalized process with the optimal structure, reduces the number of tasks and waiting time that must be experienced by the business handling from the beginning, and realizes the step-by-step improvement of processing efficiency.

[0031] Finally, by decoupling the variable business rules from the stable process engine, the present application gives the system great flexibility and maintainability, and ensures that personalized services can be continuously and efficiently operated. Once the process is solidified in the traditional system, any adjustment of business rules needs to modify the program code, which is high in cost, long in cycle, and high in risk. In the present application, the condition response rule set is an independent and configurable knowledge base that encapsulates all business adjustment logic; when the policy changes or new service situations need to be added, the system can be provided with new process personalization capability by business personnel adding, deleting or modifying the rule set configuration, without the need to redevelop and deploy the core process template and reconstruction engine. BRIEF DESCRIPTION OF DRAWINGS

[0032] In order to make the content of the present application more easily understood, the present application will be further described in detail below according to specific embodiments of the present application and in conjunction with the drawings, in which:

[0033] Figure 1 The flowchart of the business process instantiation dynamic generation method in the preferred embodiment of the present application;

[0034] Figure 2 The flowchart of the parsing structured data packet to extract instantiated condition factors in the preferred embodiment of the present application;

[0035] Figure 3 The flowchart of obtaining a matching rule in the preferred embodiment of the present application;

[0036] Figure 4 The flowchart of the node removal and dependent relationship self-repairing scheme in the preferred embodiment of the present application;

[0037] Figure 5 The structural block diagram of the business process instantiation dynamic generation in the preferred embodiment of the present application. DETAILED DESCRIPTION

[0038] The present application will be further described below in conjunction with the drawings and specific embodiments so that those skilled in the art can better understand and implement the present application, but the embodiments are not intended to limit the present application.

[0039] The purpose of the embodiments of the present application is to overcome the defects of complex business process rigidity and low processing efficiency caused by the inability to dynamically adjust according to user personalized application content in the prior art.

[0040] Embodiment one: refer to Figure 1 As shown in the figure, the embodiments of the present application disclose a business process instantiation dynamic generation method, comprising:

[0041] S100, in response to the business application of the self-service terminal, obtaining the business type and the structured data packet of the business application;

[0042] S200, according to the business type, calling the corresponding standard process template, the standard process template is predefined in the form of a directed graph structure; wherein the nodes of the directed graph structure represent the atomic processing nodes of the business, and the edges represent the dependency relationship between the nodes;

[0043] S300, parsing the structured data packet, extracting at least one instantiation condition factor;

[0044] S400, matching the instantiation condition factor with the condition response rule set of the standard process template to obtain the matching rule;

[0045] S500, dynamically reconstructing the directed graph structure of the standard process template according to the matching rule to obtain the instantiation process model matched with the business application.

[0046] In a specific application scenario, in order to facilitate the understanding of the scheme of the embodiments of the present application, the following will combine the process of generating the business process instantiation process model when Mr. Wang, a user, applies for the business of "establishment registration of domestic limited liability company" (i.e. "enterprise opening") through a self-service integrated machine:

[0047] The user selects the "enterprise opening" business type on the integrated machine interface, and according to the prompts, completes the identity information reading, fills in the electronic application form (including enterprise name, registered capital, scope of business, shareholder information, etc.), and uploads the existing proof documents (such as "enterprise name self-reporting notice letter"); the application software of the self-service terminal integrates and standardizes these scattered input information, and encapsulates it into a structured data packet conforming to the predetermined data specification; the data packet at least contains: business type code and structured application content body.

[0048] After receiving the data packet, the background system parses the business type code, and retrieves the "standard process template for opening a domestic enterprise" associated therewith from the process template library. The template is stored and represented in the system in the form of a directed graph data structure, which contains multiple nodes, and each node represents an indivisible minimum business processing unit (atomic processing node), for example:

[0049] N1: name self-declaration;

[0050] N2: acceptance and preliminary examination of establishment registration materials;

[0051] N3: audit and issuance of business license;

[0052] N4: public seal engraving and record (public security);

[0053] N5: bank account opening;

[0054] N6: tax registration and ticket type approval;

[0055] An edge is a directed arrow connecting two nodes, representing the dependency relationship between the nodes, i.e., the execution order constraint. For example, N1→N2 indicates that the acceptance of establishment registration materials N2 can only be performed after the completion of name self-declaration N1; N3→N4 and N3→N5 indicate that seal engraving and account opening can be performed in parallel after obtaining the business license.

[0056] The system performs in-depth analysis on the structured data packet received in step S100, which not only reads the values of each field, but also performs key information identification and semantic association. In this embodiment, the system finds after parsing that:

[0057] The user has uploaded a valid "Enterprise Name Self-Declaration Notice" electronic file, and its status is "passed"; the user has checked "small and micro enterprise" in the "enterprise type" field of the electronic application form; the user has selected a registration address in the "high-tech industrial development zone" of the city; based on these specific and personalized data contents, the system extracts the following instantiated condition factors:

[0058] Preceding link status: name self-declaration, value: completed;

[0059] Enterprise attribute: enterprise type, value: small and micro enterprise;

[0060] Regional attribute: registration address belonging region, value: high-tech industrial development zone.

[0061] The system maintains a condition response rule set bound to the "standard process template for opening an enterprise", and each rule defines "under what conditions, what operation is performed on the process template"; the system matches the instantiated condition factor set extracted in step S300 with the rule set. For example, the rule set contains:

[0062] Rule R1: IF {Preceding Node Status: Name Self-Declaration, Value: Completed} THEN remove node N1.

[0063] Rule R2: IF {Enterprise Attribute: Enterprise Type, Value: Small and Micro Enterprise} AND {Region Attribute: Registered Address Belongs to Region, Value: High-Tech Industrial Development Zone} THEN add attribute label {Label Name: Handling Channel, Value: Green Channel} to node N6 and modify the dependency type of edges N4→N6 and N5→N6 to weak dependency.

[0064] After matching, the system determines that the current instantiation condition factors simultaneously satisfy the triggering conditions of rules R1 and R2, so the matching rules R1 and R2 are obtained.

[0065] Apply rule R1: remove node N1 from the graph, and delete all edges connected to N1; this means that the link is completely removed from the process logic.

[0066] Apply rule R2:

[0067] Add attribute label {Handling Channel: Green Channel} to node N6; this label will be recognized during subsequent task scheduling, allowing the node task to enter the priority queue.

[0068] Modify the attributes of the edges from N4 and N5 to N6, marking them as "weak dependency", indicating that N6 does not need to strictly wait for N4 and N5 to complete, but can attempt to start as long as the preconditions are partially met, increasing the flexibility of the process.

[0069] After completing the above reconstruction operations, the original standard flowchart has changed; the system serializes the reconstructed directed graph structure to generate an instantiated process model, which is specifically tailored for Mr. Wang's application, and clearly indicates that the process starts from "Establishment of Registration Material Acceptance" N2, as the name approval has been skipped; after obtaining the business license N3, stamping N4 and opening an account N5 can be simultaneously promoted; tax registration N6 enjoys a green channel. The system can visualize this model and display it to the user, or convert it into a work order flow and push it to the back-end approval departments for execution.

[0070] It should be noted that the self-service terminal is an intelligent device deployed in government offices, communities, industrial parks, and other places, with human-computer interaction interface, identity recognition module, material collection module, and network communication capabilities; its technical role is to convert the user's offline application into a standard electronic request.

[0071] Structured data package refers to the data unit encapsulated in a unified format (such as JSON, XML or specific protocol buffer format) according to the original data input by the user through various ways such as forms, files, certificate reading, etc., to ensure the standardization and analyzability of data transmission between systems.

[0072] The standard process template and the directed graph structure are formalized and digitized modeling of complex business processes; the nodes represent the smallest business unit (atomicity) that cannot be divided, and the edges represent the logical constraints of the execution order. This representation method enables the computer to accurately understand and process the process logic.

[0073] The instantiation condition factor is an abstraction of the key data segment that can represent the uniqueness of this business application extracted from the structured data package; it is a bridge connecting the original data and the process change rule, and its form is usually a key-value pair or a data object with a semantic label.

[0074] The condition response rule set is a knowledge base that stores condition-action pairs; the condition part is used to match the instantiation condition factor, and the action part defines how to modify the directed graph structure (such as adding / deleting nodes / edges, modifying attributes), which decouples the variable business strategy from the stable process engine.

[0075] The business process instantiation dynamic generation method of the present application intelligently reconstructs the process model according to the real-time analysis of the user's submitted materials, realizes the precise adaptation of one case to one process, eliminates redundant links from the source, optimizes the processing path, and significantly improves the business handling efficiency and the level of personalized service.

[0076] In the above process of realizing personalized generation of business processes, the extraction of instantiation condition factors is the key first step; however, directly extracting factors from structured data packages faces a thorny practical technical problem: uneven data value density and noise interference. A structured data package of a government business usually contains dozens or even hundreds of fields, but not all fields have equal value for process reconstruction decision-making. For example, in the enterprise opening application, "enterprise name", "registered capital" and "scope of business" are key decision-making fields, while "contact phone", "handler email" and other fields will not trigger any process changes in most cases; if all parsed atomic data items are simply converted into instantiation condition factors, a large number of low-value or irrelevant noise factors will be generated. These noise factors will unnecessarily flood into the subsequent rule matching engine, resulting in decreased rule matching performance, reduced decision-making accuracy and poor system maintainability.

[0077] To solve the problem of uneven data value density, the embodiment of the present application proposes an intelligent factor extraction method based on machine learning. The core scheme is to introduce a factor value prediction model trained by historical data, predict a quantitative value score for each atomic data item, and intelligently filter based on the score to ensure that the output instantiated condition factor set has high signal-to-noise ratio and high correlation. Specifically, referring to Figure 2 As shown in the figure, the structured data packet is parsed, and at least one instantiated condition factor is extracted, including: parsing the structured data packet to obtain an atomic data item set containing a plurality of atomic data items; inputting each atomic data item into a factor value prediction model to obtain a factor value prediction value corresponding to each atomic data item; wherein the factor value prediction model is a machine learning model obtained by training based on historical business data, each sample in the training data includes a historical atomic data item and a corresponding process reconstruction association label; the process reconstruction association label represents the influence degree of the historical atomic data item on the historical business instance leading to the reconstruction of the process template; filtering the atomic data item set according to the factor value prediction value to obtain one or more effective atomic data items; and formatting the effective atomic data items into instantiated condition factors.

[0078] Atomic data item refers to the basic data unit obtained by parsing the structured data packet at the minimum granularity; each atomic data item usually contains a field identifier and a data value, which is the smallest object for value evaluation. The factor value prediction model is a machine learning model, which essentially learns a mapping function from data item features to process reconstruction value.

[0079] Each training sample consists of two parts:

[0080] Input: a specific atomic data item in a historical business instance, such as {Is a high-tech enterprise: Yes}.

[0081] Label: a quantitative process reconstruction association label; the construction of this label needs to represent the influence degree of the data item in the historical business instance leading to reconstruction; a feasible technical implementation is: backtracking the process model finally generated by the historical business, calculating the difference degree (such as the number of added and deleted nodes) with the standard template, and then attributing the difference degree to each appearing data item through statistical method to assign a contribution weight as the label. For example, a business skips two links because it is a high-tech enterprise, so the label value of this data item is higher.

[0082] The factor value prediction value is a quantitative estimate of the model on the size of the role that the current atomic data item may play in future process reconstruction, which is a continuous or discrete numerical value, providing an objective and unified measurement standard for subsequent filtering.

[0083] Screening refers to filtering of the atomic data item set according to the factor value prediction value, and the technical means is to set a threshold or sort by score to take Top-N, and the essence is a model-based feature selection process which uses learned knowledge rather than fixed rules to distinguish signal from noise.

[0084] Formatting refers to converting the screened data item carrying the original field information into a standardized instantiated condition factor format meeting the input requirements of the downstream rule matching engine.

[0085] The training method of the factor value pre-evaluation model specifically includes:

[0086] A historical business instance data set is collected, wherein each instance contains a historical structured data packet and a reconstruction operation record of a process model generated by the instance compared with a standard template;

[0087] For each historical instance, the following operations are performed to build a model training sample:

[0088] All atomic data items are extracted from the historical structured data packet;

[0089] For each atomic data item, a process reconstruction association label is labeled for it according to the reconstruction operation record, and the value of the label is determined based on the causal inference strength between the field where the atomic data item is located and its value and each operation in the reconstruction operation record; wherein the causal inference strength is calculated by analyzing the difference in conditional probability of the corresponding process occurring specific reconstruction in a large number of instances where the atomic data item appears and does not appear, or takes different values;

[0090] The field identifier, value and context features of the atomic data item are taken as input features, and the process reconstruction association label is taken as a supervised target to train the factor value pre-evaluation model.

[0091] The embodiment scheme of the application quantitatively evaluates the value of each data item through a machine learning model, and accurately screens based on this, fundamentally solving the problems of uneven value density and serious noise interference in original data. The factor value prediction model is trained based on historical data and can be continuously optimized and updated as the accumulation of business handling data, automatically discovers new important data fields, so that the system has the ability of continuous learning and evolution, and does not need frequent manual adjustment of extraction rules.

[0092] In the above embodiment, the factor value prediction model trained based on historical business data predicts the factor value, which learns the average performance of the data item value in all business scenarios. However, in a real government or business environment, the importance of the same data item in different business types or scenarios is very different. For example, the "registered capital" field in the "enterprise opening" business, which exceeds a certain amount, may trigger the "key enterprise attention" process, and the value is high. However, in the "enterprise information change (capital reduction)" business, its value may be significantly reduced; while in the "certificate loss" business, this field may have no value. The "business scope" field contains specific keywords such as "finance" and "education", which may have an average value in most businesses, but in "industry special license" type businesses, it may be the core factor that determines the process branch.

[0093] How to overcome the scene-insensitive defect of the global value prediction model, so that the evaluation of the factor value can be dynamically and accurately adjusted according to the specific scene where the current business application is located, thereby further improving the accuracy and adaptability of the instantiation condition factor extraction, the embodiment of the present application After obtaining the factor value prediction value corresponding to each atomic data item, it further includes: determining the scene category to which the business application belongs based on the business type; loading the pre-defined scene weight configuration table according to the scene category, and defining a scene adjustment coefficient for each type of atomic data item in the scene weight configuration table; using the scene adjustment coefficient to correct the factor value prediction value to obtain a corrected prediction value; and using the corrected prediction value to replace the factor value prediction value.

[0094] In a specific embodiment scenario, suppose that the factor value prediction model gives the original prediction value for several atomic data items:

[0095] A: {field: registered capital, value: 100 million}->original prediction value: 0.70;

[0096] B: {field: whether it is a high-tech enterprise, value: yes}->original prediction value: 0.85;

[0097] C: {field: construction project EIA reply document number, value: X20240001}->original prediction value: 0.40.

[0098] The system first determines the scene category to which the current application belongs based on the business type, and assumes that the result is:

[0099] Scene S1: Enterprise opening (general);

[0100] Scene S2: Enterprise opening (high-tech enterprise identification special);

[0101] Scene S3: Engineering construction project construction permit.

[0102] The system loads a corresponding scene weight configuration table from storage according to the determined scene category, e.g., S2, which predefines a scene adjustment coefficient for each category of atomic data item. For example:

[0103] For scene S2 (high-tech enterprise special item):

[0104] Registered capital factor: adjustment coefficient = 0.8;

[0105] High-tech enterprise factor: adjustment coefficient = 1.5;

[0106] Construction project EIA approval number factor: adjustment coefficient = 0.1.

[0107] Multiply the original predicted value by the corresponding scene adjustment coefficient to obtain the corrected predicted value:

[0108] A (registered capital): 0.70 x 0.8 = 0.56;

[0109] B (high-tech enterprise): 0.85 x 1.5 = 1.28;

[0110] C (EIA approval): 0.40 x 0.1 = 0.04.

[0111] Replace the factor value prediction with the calculated corrected prediction value (0.56, 1.28, 0.04); the subsequent screening steps will be based on these corrected values that better fit scene S2. The result is that in the "high-tech enterprise identification special item" scene, the importance of the "high-tech enterprise" factor is highlighted, the importance of the "registered capital" is reasonably lowered, and the irrelevant "EIA approval" factor is effectively suppressed.

[0112] The embodiment of the present application introduces a scene-aware value correction scheme, which adds a scene-specific correction layer on the basis of the general prediction model. The correction layer adjusts the prediction value of the general model according to the current business scene, so that the final value evaluation is more consistent with the true importance distribution of the current scene.

[0113] On the basis of the above embodiments, the conditional response rule set can contain hundreds of rules, each of which defines a process adjustment to be performed under a specific business condition; the trigger condition of these rules can itself be a complex expression composed of multiple logical sub-conditions. How to accurately and efficiently find the rule triggered by the current application in the massive and complex business rules, the traditional simple traversal comparison method of instantiating the condition factor and comparing it with the conditions of all rules by string or simple value, has obvious defects: if the complex analysis and comparison of the full condition of each rule in the rule set are performed, when the number of rules is large, the calculation cost is huge, which seriously affects the system response speed and cannot meet the requirements of real-time feedback of self-service terminals.

[0114] To solve the accuracy and efficiency problem of the above rule matching, referring to Figure 3 The matching rule is obtained by matching the instantiated condition factor with the conditional response rule set of the standard process template, including: according to the business type, reading the corresponding conditional response rule subset from the rule storage; for each rule in the conditional response rule subset, analyzing the trigger condition of the rule to obtain a set of sub-condition constraints; comparing each sub-condition constraint with the instantiated condition factor to determine whether there is an instantiated condition factor that satisfies the sub-condition constraint, and if a set of sub-condition constraints are satisfied, it is determined that the rule is matched successfully, and the matching rule is obtained.

[0115] In a specific embodiment, the rule matching process in the business of an enterprise is taken as an example:

[0116] {Pre-approval state: name approval, result: passed};

[0117] {Enterprise qualification attribute: enterprise type, value: small and micro enterprise};

[0118] {Regional policy attribute: registration place, value: high-tech industrial development zone}.

[0119] Step 1, rule subset preloading:

[0120] First, according to the business application type, read the corresponding conditional response rule subset from the central rule storage (such as rule database or rule engine), which avoids traversing all irrelevant business rules and greatly improves the efficiency of subsequent comparison. Assuming that the subset read contains rules R1, R2, R3, etc.

[0121] Step 2, rule trigger condition analysis:

[0122] The system performs structured analysis for each rule in the subset, taking a rule as an example:

[0123] Rule R2: IF {Enterprise Qualification Attribute: Enterprise Type, Value: Small Micro Enterprise} AND {Regional Policy Attribute: Registration Place, Value: High-tech Industrial Development Zone} THEN {Action: Add "Green Channel" Label to Tax Registration Node}.

[0124] The parsing process is to parse the trigger condition of the rule, which is composed of two sub-condition constraints connected by logical "and".

[0125] Sub-condition constraint 1: {Enterprise Qualification Attribute: Enterprise Type, Value: Small Micro Enterprise};

[0126] Sub-condition constraint 2: {Regional Policy Attribute: Registration Place, Value: High-tech Industrial Development Zone}.

[0127] Step three, sub-condition constraint and instantiation condition factor comparison:

[0128] Compare each sub-condition constraint with the current instantiation condition factor set;

[0129] Compare sub-condition constraint 1: Find if there is a factor of type Enterprise Qualification Attribute: Enterprise Type and value Small Micro Enterprise in the instantiation condition factor set; find success.

[0130] Compare sub-condition constraint 2: Find if there is a factor of type Regional Policy Attribute: Registration Place and value High-tech Industrial Development Zone; find success.

[0131] Step four, match determination and result output:

[0132] Since a set of sub-condition constraints (sub-condition constraint 1 AND sub-condition constraint 2) are met, it is determined that the rule R2 matches successfully, and R2 is output as a matching rule; repeat steps two to four for other rules in the rule subset.

[0133] Note: Rule storage refers to a technical component for persistently storing condition response rule sets, which can be a relational database, NoSQL database, or a special rule engine (such as Drools); its role is to separate business rules from program code, enabling rule management and independent updates.

[0134] The condition response rule subset is a rule set associated with a specific business type in the rule storage, indexed and filtered by business type to improve matching efficiency.

[0135] Parsing the trigger condition of the rule is a process of converting a natural language or domain-specific language description of the rule condition into a structured logical representation that can be processed by a computer. For example, the rule text is parsed into a condition syntax tree, where the leaf nodes are atomic sub-condition constraints and the intermediate nodes are logical operators.

[0136] The embodiment of the present application avoids a large number of invalid rule traversals from the source by preloading a rule subset according to the service type; the factor comparison after the structured analysis can realize approximate search with the aid of an efficient data structure, so that the overall matching process can remain efficient even in the face of complex condition combinations.

[0137] Further, the condition response rule subset is stored in a structured manner based on the trigger condition of the rule; wherein each rule is associated with at least one index key, and the index key is determined according to the type of at least one instantiated condition factor that must appear in the trigger condition of the rule. In the embodiment of the present application, when storing the rule, the trigger condition thereof is analyzed in advance, the key features are extracted, and the index key is generated; when matching, a large number of obviously irrelevant rules are first filtered out by using the index keys, and the range of candidate rules that need to perform full condition comparison is greatly reduced.

[0138] Preferably, the dynamic reconstruction operation includes at least one of the following:

[0139] Operation one: if the instantiated condition factor indicates that any atomic processing node has been completed, the corresponding node and all associated edges are removed from the directed graph structure.

[0140] Operation two: if the instantiated condition factor includes user-specified priority information, an attribute label is added to the corresponding node.

[0141] The removal of the node in operation one refers to deleting the vertex object from the vertex set of the graph, and the removal of all associated edges refers to traversing the edge set of the graph to delete all edges whose source vertex or target vertex is the node; the node removal shortens the critical path of the flow, and this simplification occurs in the flow definition stage, which is superior to the scheme of skipping by condition judgment at runtime.

[0142] The addition of the attribute label in operation two converts the priority of the business layer into structured information of the flow model layer, provides clear and machine-readable instructions for downstream scheduling, so that priority handling is no longer a slogan, but a strategy that can be recognized and executed by the system, significantly improving the individualization level of services and user satisfaction.

[0143] The dynamic reconstruction operation of the embodiment of the present application eliminates redundant links from the logical level and directly optimizes the processing path, thereby significantly compressing the overall handling time, reducing the number of tasks and waiting time that must be experienced by the business from the starting point by generating a structurally optimal individualized flow, and realizing a step-by-step improvement in processing efficiency.

[0144] In the operation scheme of the above dynamic reconstruction operation, simply deleting a node and its associated edges from the directed graph removes the node but also cuts off the data flow or control flow path that originally flowed through the node, which can lead to two undesirable consequences: logical breakage, loss of dependency relationship; implicit dependency exposure, X node may not be the only path from A to B; for example, there may be two paths A→X→B and A→C→B in the original process, X is a redundant or secondary path; after removing X, the dependency from A to B still holds through C, and the creation of a new node can introduce redundancy or cycles.

[0145] To solve the problem of dependency repair after node removal, the embodiment scheme of the present application proposes a dependency relationship self-repair scheme. Specifically, referring to FIG. 8, after removing the corresponding node and all associated edges from the directed graph structure, the method further includes: identifying all upstream dependency nodes and downstream dependency nodes of the removed node; for each pair of nodes consisting of an upstream dependency node and a downstream dependency node, performing the following judgment: judging whether there is a direct connection between the node pair. If not, further judging whether the removed node constitutes the only dependency edge of the node pair in the standard process template; if not, creating a new dependency edge between the upstream dependency node and the downstream dependency node of the node pair; if yes, marking the downstream dependency node of the node pair as a node to be manually checked. Figure 4

[0146] In a specific application scheme, after removing a node from the directed graph, the affected node pairs are located, i.e. all nodes originally pointing to the node (upstream dependency node) and the nodes originally pointed to by the node (downstream dependency node). For each such pair of nodes, two levels of judgment are performed: first, check whether there is a direct connection in the current graph; if not, further backtrack to the standard process template to analyze whether the removed node constitutes the only dependency path between the node pair. This backtracking analysis associates the current graph operation with the original business process design intent, and distinguishes whether the node is redundant or the only path by querying the original graph structure or pre-computed path information.

[0147] Based on the first level of judgment, a differentiation processing strategy is implemented: if the node is redundant, a new dependency edge is created between the upstream and downstream nodes to maintain the logical coherence of the process; this self-repair of the graph structure ensures that the process remains a connected and effective workflow after simplification. If the node is a critical unique path, then automatic creation of edges that may convey incorrect semantics is avoided, and instead the downstream node is marked as requiring manual intervention for verification. This decision logic achieves precise risk control: the boundary of automatic processing is clearly defined within the range that does not damage the core business logic, and for complex situations that may introduce uncertainty, a manual supervision process is started.

[0148] ​The embodiment of the present application introduces a two-level judgment dependent relationship processing sub-process, ensures the coexistence of process simplification automation and security, and fundamentally improves the reliability of the generated process model and the robustness of the entire system.

[0149] Further, based on the above embodiment, when scheduling tasks based on the instantiated process model, the attribute label attached to the node is read, and the attribute label is mapped to a scheduling instruction; the scheduling instruction includes raising the priority of the task queue, allocating specific execution resources, setting the processing deadline, or adding task pre-check items.

[0150] In a specific application scenario, a semantic mapping layer is established between the instantiation of the process model and the task scheduling, at the key node of the task scheduling based on the instantiated process model, the attribute label attached to the node is actively read, and through a pre-defined label-instruction mapping rule, these business semantic labels are converted into specific scheduling instructions that can be directly understood and executed by the scheduler. Thus, the personalized information introduced in the process reconstruction phase is transmitted and applied to the execution phase without loss; in addition, the scheduling strategy is no longer fixed in the scheduler code, but is partially externalized to the labels of the process model. By modifying the label-instruction mapping table, the scheduling strategy for different business demands can be flexibly adjusted without modifying the core logic of the scheduler, achieving configurable and business-oriented scheduling strategies.

[0151] In the above scheme, when dynamically reconstructing according to the matched rules, there may be a logical consistency problem: different rules may indicate mutually contradictory or conflicting reconstruction operations for the same element (node or edge) of the process graph. For example, rule A may indicate adding an urgent label to node N, and rule B may indicate adding a manual review label to node N, which may not conflict in business. However, if rule C indicates removing node N, and rule D indicates adding a label to node N, these two operations are logically mutually exclusive and cannot be executed simultaneously. If all matching rules are simply stacked and executed, the process model may be in a contradictory or invalid state.

[0152] To solve the above rule conflict problem, the embodiment of the present application also introduces a reconstruction operation conflict pre-check and resolution. Specifically, before dynamically reconstructing the directed graph structure of the standard process template according to the matching rules, it also includes: analyzing the reconstruction operations indicated by the matching rules, and if there are mutually exclusive operation pairs for the same graph element, triggering conflict resolution. It should be noted that the same graph element refers to the same node or the same edge in the process directed graph; the mutually exclusive operation pair refers to two operations that cannot coexist on the same graph element; the mutually exclusive relationship can be predefined in an operation exclusion relationship table, such as the exclusion of removing a node and adding an attribute to a node; the exclusion of deleting an edge and modifying an edge weight, etc.

[0153] Trigger conflict resolution is the processing action after detecting the conflict, the resolution process involves decision logic, and the result is to select one from the conflicting operations, or merge according to certain rules, or reject all and alarm; the resolution strategy is an important part of the scheme, which can be rule priority comparison, operation type weight comparison or calling more complex conflict resolver.

[0154] By reconstructing operation conflict pre-check and resolution, the generation of process models containing contradictory instructions due to rule conflicts is avoided from the source, ensuring that each generated instantiated process model is internally logically self-consistent and interpretable, greatly improving the reliability of the output results.

[0155] The second aspect is to solve the above technical problems, the application provides a business process instantiation dynamic generation device for executing the business process instantiation dynamic generation method of the first embodiment, as shown in Figure 5 The device comprises:

[0156] The application receiving module receives the business application of the self-service terminal, obtains the business type and the structured data packet of the business application;

[0157] The template management module retrieves the corresponding standard process template according to the business type, and the standard process template is predefined in a directed graph structure; wherein the nodes of the directed graph structure represent atomic processing nodes of the business, and the edges represent the dependency relationship between the nodes;

[0158] The factor extraction module parses the structured data packet and extracts at least one instantiation condition factor;

[0159] The rule matching module matches the instantiation condition factor with the condition response rule set of the standard process template to obtain a matching rule;

[0160] The process reconstruction engine dynamically reconstructs the directed graph structure of the standard process template according to the matching rule to obtain an instantiated process model matched with the business application.

[0161] The device of the embodiment of the application is used to execute the method of the first embodiment, both of which are based on the same inventive concept and have the same technical effect, which will not be described here.

[0162] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0163] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart or flowsheet block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 one or more flowcharts and / or blocks

[0164] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart or flowsheet block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 one or more flowcharts and / or blocks

[0165] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart or flowsheet block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 one or more flowcharts and / or blocks

[0166] Obviously, the above-described embodiments are only examples for clarity and are not intended to limit the implementation. Based on the above description, other different forms of changes or variations can be made by those skilled in the art. Here, it is not necessary and impossible to enumerate all the implementations. The obvious changes or variations derived therefrom are still within the protection scope of the present application.

Claims

1. A method for dynamically generating business process instantiation, characterized in that, include: In response to a service request from a self-service terminal, obtain the service type and the structured data packet of the service request; The corresponding standard process template is retrieved according to the business type. The standard process template is predefined in a directed graph structure. The nodes of the directed graph structure represent the atomic processing nodes of the business, and the edges represent the dependencies between the nodes. Parse the structured data packet and extract at least one instantiation condition factor; The instantiated condition factor is matched with the condition response rule set of the standard process template to obtain the matching rule; The directed graph structure of the standard process template is dynamically reconstructed according to the matching rules to obtain an instantiated process model that matches the business application. Parse the structured data packet and extract at least one instantiation condition factor, including: Parse the structured data packet to obtain an atomic data item set containing multiple atomic data items; Each of the atomic data items is input into the factor value prediction model to obtain the factor value prediction value corresponding to each atomic data item; wherein, the factor value prediction model is a machine learning model trained based on historical business data, and each sample in its training data includes a historical atomic data item and its corresponding process reconstruction association label; the process reconstruction association label characterizes the degree of influence of the historical atomic data item on the reconstruction process template in historical business instances; Based on the predicted factor value, filter the atomic data item set to obtain one or more valid atomic data items; The valid atomic data item is formatted as the instantiation condition factor.

2. The method for dynamically generating business process instantiation according to claim 1, characterized in that, After obtaining the predicted factor value for each of the atomic data items, the process further includes: The scenario category to which the service application belongs is determined based on the service type; According to the scene category, a predefined scene weight configuration table is loaded, and a scene adjustment coefficient is defined for each type of atomic data item in the scene weight configuration table; The predicted value of the factor is corrected using the scenario adjustment coefficient to obtain the corrected predicted value; Replace the factor value prediction with the corrected prediction.

3. The method for dynamically generating business process instantiation according to claim 1, characterized in that, The instantiated condition factor is matched with the condition response rule set of the standard process template to obtain matching rules, including: Based on the business type, read the corresponding subset of conditional response rules from the rule storage; For each rule in the subset of conditional response rules, the triggering condition of the rule is parsed to obtain a set of sub-conditional constraints; Each sub-condition constraint is compared with the instantiation condition factor to determine whether there is an instantiation condition factor that satisfies the sub-condition constraint. If all sub-condition constraints are satisfied, the rule is determined to be a successful match, and the matching rule is obtained.

4. The method for dynamically generating business process instantiation according to claim 3, characterized in that, The subset of conditional response rules is stored in a structured manner based on the triggering conditions of the rules; wherein each rule is associated with at least one index key, the index key being determined according to the type of at least one instantiated condition factor that must appear in the triggering conditions of the rule.

5. The method for dynamically generating business process instantiation according to claim 1, characterized in that, The dynamic reconstruction operation includes at least one of the following: If the instantiation condition factor indicates that any of the atomic processing nodes has been completed, then the corresponding node and all its associated edges are removed from the directed graph structure. If the instantiation condition factor includes user-specified priority information, then an attribute label is added to the corresponding node.

6. The method for dynamically generating business process instantiation according to claim 5, characterized in that, After removing the corresponding node and all its associated edges from the directed graph structure, the process further includes: Identify all upstream and downstream dependent nodes of the removed node; For each node pair consisting of an upstream dependent node and a downstream dependent node, perform the following judgment: Determine whether there is a direct dependency edge between the node pairs. If not, further determine whether the removed node constitutes the only dependency edge for the node pair in the standard process template. If not, a new dependency edge is created between the upstream dependent node and the downstream dependent node that constitute the node pair; If so, then mark the downstream dependent node of the node pair as a node to be manually verified.

7. The method for dynamically generating business process instantiation according to claim 5, characterized in that, The method further includes: when scheduling tasks based on the instantiated process model, reading the attribute tags attached to the nodes and mapping the attribute tags to scheduling instructions; the scheduling instructions include increasing the priority of the task queue, allocating specific execution resources, setting processing deadlines, or adding pre-task checks.

8. The method for dynamically generating business process instantiation according to claim 5, characterized in that, Before dynamically reconstructing the directed graph structure of the standard process template according to the matching rules, the method further includes: analyzing the reconstruction operation indicated by the successfully matched rules, and if there are mutually exclusive operation pairs for the same graph element, then conflict resolution is triggered.

9. A business process instantiation dynamic generation apparatus, used to execute the business process instantiation dynamic generation method according to any one of claims 1-8, characterized in that, include: The application receiving module receives service applications from self-service terminals and obtains the service type and the structured data packet of the service application. The template management module retrieves the corresponding standard process template based on the business type. The standard process template is predefined in a directed graph structure. The nodes of the directed graph structure represent the atomic processing nodes of the business, and the edges represent the dependencies between the nodes. The factor extraction module parses the structured data packet and extracts at least one instantiated condition factor; The rule matching module matches the instantiated condition factor with the condition response rule set of the standard process template to obtain matching rules; The process refactoring engine dynamically refactors the directed graph structure of the standard process template according to the matching rules to obtain an instantiated process model that matches the business application.

Images