Workflow processing method, system and equipment based on routing algorithm and medium

By combining the construction of a semantic graph and the DQN optimization layer, the problems of format incompatibility and implicit association in cross-departmental approval are solved, and efficient and stable approval path optimization and visual report generation are achieved.

CN120746254AActive Publication Date: 2025-10-03INSPUR GENERSOFT CO LTD

Patent Information

Application Number
CN202511140683.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-08-15
Publication Date
2025-10-03
Estimated Expiration
2045-08-15

AI Technical Summary

Technical Problem

In cross-departmental approval scenarios within enterprises, existing technologies struggle to process approval requests from different systems, leading to incompatible formats, data parsing errors, waste of resources, and incomplete approval processes. Traditional routing decisions also ignore implicit associations, resulting in low approval efficiency and easily blocked processes.

Method used

By constructing a semantic graph to extract entity relationships, combining the real-time status and historical information of the approval terminal, using collaborative analysis to calculate dynamic weights, combining the rule engine and DQN optimization layer to make routing decisions, optimize the approval path, and generate a visual report.

Benefits of technology

It improves the accuracy and efficiency of approval routing, avoids task backlogs, balances the load on approval terminals, enhances process stability and transparency, and realizes intelligent and adaptive workflow.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120746254A_ABST
    Figure CN120746254A_ABST
Patent Text Reader

Abstract

The invention provides a workflow processing method, system and device based on a routing algorithm and a medium, and belongs to the technical field of workflow approval, and the method specifically comprises the steps: obtaining an approval request transmitted by an API gateway, constructing a semantic map, extracting an entity relationship, calculating a dynamic weight in combination with the real-time state and historical information of an approval terminal, and generating a feature vector; analyzing the request to obtain basic data, re-constructing a semantic map based on the basic data, and fusing the terminal state and the historical record to generate a feature vector; processing the basic data and the feature vectors through a rule engine priority layer to obtain preliminary routing judgment; adjusting a routing value through a DQN optimization layer and then determining an optimal path; and finally, executing an approval process and pushing a task, receiving feedback and generating a visual report, and forming closed-loop optimization. Through multi-layer processing and optimization, the accuracy and efficiency of approval routing are improved, blocking is reduced, the process stability is enhanced, transparency is improved through a visual report, and workflow processing is integrally optimized.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of workflow approval, and in particular relates to a workflow processing method, system, device and medium based on a routing algorithm. Background Art

[0002] In some financial and manufacturing industries, such as banks, insurance companies, and securities firms, credit approval, claims settlement, and risk management processes often involve multiple steps, including client information, account managers, risk management, legal affairs, and loan / compensation disbursement. Completing each approval process is time-consuming, creates a poor user experience, and requires full traceability. In the manufacturing industry, equipment procurement, spare parts procurement, and quality issue handling require cross-factory collaboration, spanning multiple departments including quality, logistics, and finance, disrupting normal production.

[0003] Overall, in cross-departmental enterprise approval scenarios, existing technologies struggle to process approval requests from disparate systems. For example, the formats for reimbursement approvals in the finance system and leave approvals in the HR system differ significantly. Format incompatibility often leads to data parsing errors after receiving them, reducing overall work efficiency. Traditional routing decisions rely solely on surface-level data, determining routes based on the approval amount and ignoring implicit relationships. For example, in project approvals involving multiple subcontracts, existing technologies may route each subcontract approval individually, failing to consider the overall project interdependencies. This results in an incomplete approval process, a lack of comprehensive understanding of the project's status during approval, and prone to biased approval decisions. When assigning approval tasks, the status of approval terminals is not dynamically monitored. For example, even if one approval terminal is already backlogged with high-priority tasks, the system still assigns new tasks, leaving other, less-loaded approval terminals idle. This wastes resources, delays the processing of new tasks, and even leads to task blockages, hindering business progress. Summary of the Invention

[0004] The present invention provides a workflow processing method based on routing algorithm, which improves the accuracy and efficiency of approval routing, reduces blockage, enhances process stability, realizes visual reporting to improve transparency, and optimizes workflow processing as a whole through multi-layer processing and optimization.

[0005] Methods include: S101: Obtain the approval request transmitted via the API gateway, construct a semantic graph of the request content to extract entity relationships, obtain the real-time working status of the approval terminal, combine the request history information, calculate the dynamic weight through collaborative analysis, and obtain a feature vector; S102: Parse the approval request to obtain basic data; S103: Based on the basic data, a semantic graph of the request text content is constructed to extract entity relationships. The real-time working status of the approval terminal is obtained based on the associated approval terminal information. Combined with historical approval records, the weight is calculated through collaborative analysis to obtain a feature vector. S104: The basic data obtained by parsing and the extracted feature vectors are processed by the rule engine priority layer to obtain the approval routing judgment; S105: In hybrid decision routing, the DQN optimization layer, combined with the blocking risk prediction, calculates the risk-adjusted routing value on the content processed by the rule engine priority layer or directly on the basic data and feature vectors obtained by parsing, and determines the optimized routing path; S106: Based on the path determined by the hybrid decision routing, execute the approval process and push the task, receive operation feedback, and generate a workflow visualization report.

[0006] It should be further explained that step S102 specifically includes: Parse the approval request data to extract the approval amount field, application reason description, request initiation time, and source system identifier to form the original data set; Based on the predefined approval data rule library, the parsed structured data is verified for context; Perform intent recognition on the original data set content, extract key business elements, and combine them with the department affiliation information of the associated approval terminal to generate urgency labels and a list of involved departments; Verify the basic data obtained from the analysis. When it is detected that key fields are missing, execute the correction process to fill in the missing information by default or associating historical data.

[0007] It should be further explained that step S103 specifically includes: Based on the basic data obtained through analysis, combined with the approval data rule library and semantic graph, explicit entity relationships and implicit business associations in the text content are extracted; Based on the department information in the basic data, collect the real-time working status of the related approval terminals and associate the department collaboration mode data; Filter historical cases that match the current request from historical approval records, calculate the average processing time, abnormal interruption ratio, and user feedback score of the approval node in each case, and generate a historical efficiency feature set; The entity relationships, real-time work status indicator set, and historical efficiency feature set of the semantic graph are input into the collaborative analysis module, the weight distribution of the features is adjusted, and a feature vector that integrates content relevance, resource availability, and historical reliability is generated.

[0008] It should be further explained that step S104 specifically includes: Based on the basic data parsed in step S102 and the feature vector extracted in step S103, an index of the approval data rule base is constructed, classified and stored by business type and approval level, and main rules and auxiliary rules are marked; Perform rule matching preprocessing on basic data and feature vectors to extract key parameters related to the rules in the approval data rule base; Perform a preliminary match of features using the main rules to filter out candidate routes that meet the main conditions; The matching results are checked for contradictions. If there are multiple routes that meet the conditions, the final approval route is determined based on the rule priority.

[0009] It should be further explained that step S1044 specifically includes: Extract conflict features from the candidate route set after rule matching, identify the conflict type, and mark the conflict dimension of each candidate route; Re-rank the weights of conflicting rules based on historical decision-making information and current business scenarios; For candidate routes with conflicting priorities, a weighted scoring method is used to calculate the comprehensive score, and the route with the highest comprehensive score is selected as the final approval path; For unmatched requests, the abnormal features in the basic data and the key information in the feature vector are extracted and marked as abnormal feature request marking codes, with additional explanations of the conflict reasons and associated parameters, and the approval routing judgment results containing the marking information are output.

[0010] It should be further explained that step S105 specifically includes: Receive the approval routing judgment result outputted in step S104, obtain the basic data parsed in step S102 and the feature vector extracted in step S103, integrate the three types of information into a decision input data set, and mark the key influencing information in the decision input data set; Based on the input decision input data set, the blocking risk assessment dimension is constructed to form a risk assessment matrix; In the DQN optimization layer, risk matching is performed on the routes processed by the rule engine priority layer, and the blocking probability of the route under the current risk assessment matrix is ​​calculated. For requests that have not been processed by the rule engine, the blocking probability of the route is calculated based on the basic data and feature vectors. Adjust each candidate route based on the blocking probability, giving priority to routing paths with low blocking risk and that meet business objectives. If multiple equivalent paths exist, determine the final optimized routing path based on historical execution efficiency and output the path details.

[0011] It should be further explained that step S1053 specifically includes: A dual-channel risk mapper is established inside the DQN optimization layer. The first channel receives marked routing paths, and the second channel receives special requests that are not defined by rules, and generates path state snapshots for each channel. Inject real-time terminal heartbeat sequences and cross-department delay fingerprints into each path status snapshot to form a dynamic risk fingerprint; Call the adaptive risk metric based on the risk fingerprint and output the blocking probability vector of each candidate path; The blocking probability vector is coupled with the path service weight to generate a risk-adjusted routing value indication and cache it in the decision log.

[0012] The present application also provides a workflow processing system based on a routing algorithm, the system comprising: The feature extraction module is used to obtain approval requests transmitted through the API gateway, build a semantic graph of the request content to extract entity relationships, obtain the real-time working status of the approval terminal, and calculate dynamic weights through collaborative analysis based on the request history information to obtain the feature vector; Request parsing module, used to parse approval requests and extract basic data; The semantic fusion module builds a semantic graph of the request text content based on basic data to extract entity relationships. It obtains the real-time working status of the approval terminal based on the associated approval terminal information, combines historical approval records, calculates weights through collaborative analysis, and obtains a feature vector. The priority judgment module is used to process the basic data obtained through analysis and the extracted feature vectors using the rule engine priority layer to obtain the approval routing judgment; The routing decision module is used in hybrid decision routing to calculate the risk-adjusted routing value based on the content processed by the rule engine priority layer or directly on the basic data and feature vectors obtained through the analysis, through the DQN optimization layer and combined with the blocking risk prediction, to determine the optimized routing path; The execution feedback module is used to execute the approval process and push tasks according to the path determined by the hybrid decision routing, receive operation feedback, and generate a workflow visualization report.

[0013] According to another embodiment of the present application, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the workflow processing method based on the routing algorithm when executing the program.

[0014] According to another embodiment of the present application, a storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the workflow processing method based on the routing algorithm are implemented.

[0015] It can be seen from the above technical solutions that the present invention has the following advantages: The method of the present application constructs a semantic graph to extract entity relationships and integrates industry relationships. In project approval, it can explore the relationship between sub-contracts and projects, and reasonably associate and route related approval tasks, so that the approval terminal can obtain comprehensive information and shorten the approval cycle. While obtaining the real-time status of the terminal, optimize the route in combination with risk prediction. Balance the load of the approval terminal, avoid concentrated backlog of tasks, improve resource utilization, and improve overall approval efficiency. Use SHAP analysis to generate a decision report, receive user ratings and feedback to the model. Make users clear about the basis for approval routing, continuously optimize through feedback, reduce unreasonable routing, and improve decision accuracy. Directly calculate the route through the DQN optimization layer. Realize adaptive and rapid processing of special requests such as emergency projects, respond in a timely manner, realize intelligent, efficient and adaptive workflow approval, and improve approval efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for the description. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0017] Figure 1 It is a flow chart of the workflow processing method based on routing algorithm; Figure 2 This is a flow chart of an embodiment of a workflow processing method based on a routing algorithm; Figure 3 This is a flow chart of another embodiment of a workflow processing method based on a routing algorithm; Figure 4 This is a flow chart of another embodiment of a workflow processing method based on a routing algorithm; Figure 5 This is a schematic diagram of a workflow processing system based on routing algorithm; Figure 6 Schematic diagram of an embodiment of a workflow processing system based on a routing algorithm; Figure 7 Schematic diagram of an electronic device. DETAILED DESCRIPTION

[0018] This application's method addresses the issue of rigid approval paths by employing a dynamic routing algorithm based on reinforcement learning to optimize approval paths in real time. Intelligent load balancing is employed, combining the real-time availability and historical response speed of approvers to make decisions, addressing issues of low approval efficiency and process blockage. An adaptive relationship graph is used to learn and update approval paths. This eliminates the need for manual adjustments to approval nodes and rules when organizational structures change, providing a basis for routing decisions and visual analysis.

[0019] The workflow processing method based on the routing algorithm involved in this application will be described in detail below. For the purpose of illustration rather than limitation, specific details such as specific system structures and technologies are provided to facilitate a thorough understanding of the embodiments of this application. However, it should be clear to those skilled in the art that this application can also be implemented in other embodiments without these specific details.

[0020] It should be understood that when used in this specification, the term "comprising" indicates the presence of the described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or their collections. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0021] The phrases "one embodiment" or "some embodiments" described in this application mean that the specific features, structures, or characteristics described in the embodiment are included in one or more embodiments of the application. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in other embodiments," etc. that appear in different places in this application do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized.

[0022] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0023] See also Figure 1 FIG. 1 is a flowchart of a workflow processing method based on a routing algorithm in a specific embodiment, the method comprising: Step S101: Obtain the approval request transmitted through the API gateway, build a semantic graph of the request content to extract entity relationships, obtain the real-time working status of the approval terminal, combine the request history information, calculate the dynamic weight through collaborative analysis, and obtain the feature vector.

[0024] In some embodiments, after obtaining the approval request transmitted through the API gateway, the format and integrity of the request are first verified to eliminate invalid or abnormal requests. Subsequently, for the request content, a semantic graph containing multi-level entities and the relationship between entities is constructed through semantic analysis technology to accurately extract the entity relationship. At the same time, by connecting with the approval terminal system, the terminal's current task volume, response speed, online status and other working status data are collected in real time, and the historical processing records of such requests are retrieved, including historical approval paths, processing time, approval results and other information. The entity relationship features, terminal real-time status features and historical information features in the semantic graph are collaboratively associated, and the weight ratio of each feature in the analysis is dynamically adjusted according to the business type of the request. Finally, a fusion is generated to generate a feature vector that can fully reflect the request characteristics and processing environment.

[0025] It can be seen that this embodiment ensures the reliability of the data source by verifying the validity of the approval request, uses semantic graph technology to mine the intrinsic entity association of the request content, combines the real-time processing capabilities and historical processing rules of the approval terminal, realizes feature fusion through weight distribution, and provides input data for the routing algorithm.

[0026] Step S102: Parse the approval request to obtain basic data.

[0027] Optionally, the basic data includes but is not limited to: text content, approval amount, urgency label, and involved departments.

[0028] This embodiment parses received approval requests, identifying their source, which could be a financial system, a human resources system, or other systems. It then adapts the format to the specific system and unifies the data structure. It extracts explicit basic data from the request, including text content, approval amount, urgency tag, and involved departments. It also uses semantic analysis to mine implicit business attributes from the text, such as project type, contract duration, and associated business processes.

[0029] In some specific embodiments, step S102 specifically includes the following steps: S1021: Analyze the approval request data to extract the approval amount field, application reason description, request initiation time, and source system identifier to form the original data set.

[0030] It should be noted that data parsing can be performed using the BERT-Base model to perform sequence annotation and extract business elements. Relevant data from the original dataset can also be obtained through API gateway log parsing, ultimately merging into a raw dataset containing structured fields, text elements, and system identifiers.

[0031] S1022: Based on the predefined approval data rule library, the parsed structured data is context-checked. For example, the verification is performed to see if the approved amount is within the department's authority and if the department involved has business relevance to the application content, thereby generating basic data.

[0032] Optionally, the approval data rule library includes a department authority table and business relevance rules. During verification, SQL is used to query whether the matching data complies with the rules and generate a pass / fail mark.

[0033] S1023: Perform intent recognition on the content of the original data set, extract key business elements, and combine them with the department affiliation information of the associated approval terminal to generate urgency labels and a list of involved departments.

[0034] It should be noted that the urgency label is generated using a text classification model. The model input is the text vector of the application reason, and the output is three types of labels: high / medium / low. The classification threshold is adjusted according to the text features of historical emergency requests; the list of departments involved is generated by matching business elements through the department relationship map.

[0035] S1024: Verify the basic data obtained by parsing. When it is detected that a key field is missing, trigger a correction mechanism to fill in the missing information by using preset default values ​​or associating historical data.

[0036] Completeness checks are performed on a set of key fields, such as amount, involved department, and urgency. When a field is missing, a default value is first extracted from the statistical model of historical requests of the same type. If this cannot be done, the request is marked as pending and sent to the manual review interface. After correction, the data is synchronized with the original dataset.

[0037] It can be seen that step S102 achieves compatible processing of approval requests from different source systems, improving the versatility of the system. The extracted basic data contains both explicit information and implicit attributes, enriching the data dimension and improving the efficiency of the entire workflow processing.

[0038] Step S103: Based on the basic data, a semantic graph of the request text content is constructed to extract entity relationships. The real-time working status of the approval terminal is obtained according to the associated approval terminal information. Combined with historical approval records, the weight is calculated through collaborative analysis to obtain a feature vector.

[0039] In some embodiments, based on the basic data parsed in step S102, the data is first pre-processed to filter out business-related data such as the project's field, budget category, etc., and mark the business-related points in the data. Based on the text content in the basic data, combined with the industry, such as the supplier-material-order association in procurement approval, the scope of semantic graph construction is expanded, the entity relationship network is enriched, and entity relationships are extracted. According to the information such as the departments involved in the basic data, the associated approval terminals are determined, and the real-time working status of these terminals is obtained through multi-channel data collection. At the same time, the historical approval records related to the approval request are retrieved, and features such as the historical approval cycle, approval pass rate, and cross-departmental collaboration efficiency are extracted. These features are deeply associated with the basic data and the real-time status of the terminal, and the dimensions and weights of the collaborative analysis are adjusted to calculate the feature vector. Improve analysis efficiency and enhance the comprehensiveness of entity relationship extraction; collect terminal status to more accurately reflect the terminal processing capabilities; combine historical approval records so that the feature vector contains historical experience information.

[0040] Step S104: The basic data obtained by parsing and the extracted feature vectors are processed by the rule engine priority layer to obtain the approval routing judgment.

[0041] In some embodiments, based on the basic data analyzed in step S102 and the feature vector extracted in step S103, a rule library is constructed, and the rules are classified and stored according to financial approval, personnel appointment and dismissal approval, or department level and company level, and marked, for example, large-scale fund approval requires approval by the general manager and cross-departmental approval requires confirmation by the department coordinator.

[0042] Rule matching preprocessing is performed on the basic data and feature vectors to extract the approval amount range and the authority level of the relevant departments related to the rules in the rule base, converting the unstructured data into structured parameters that can be directly matched. A hierarchical matching strategy is implemented, with the structured parameters initially matched against the primary rules to select candidate routes that meet the primary criteria. A secondary screening of the candidate routes is then performed using auxiliary rules to eliminate routes that do not meet the rule requirements. Finally, conflict detection is performed on the matching results. If multiple routes meet the criteria, the final approval route is determined based on the rule priority. If no matching rules exist, the request is marked as a special request and forwarded to the subsequent optimization layer for processing.

[0043] It can be seen that the construction of the classification rule library makes rule management clearer and improves the efficiency of rule calling and matching; the hierarchical matching and conflict detection mechanism ensures the accuracy of routing judgment, avoids decision-making errors caused by rule conflicts, and improves the efficiency of workflow processing.

[0044] Step S105: In the hybrid decision routing, the DQN optimization layer is combined with the blocking risk prediction to calculate the risk-adjusted routing value of the content processed by the rule engine priority layer or directly on the basic data and feature vectors obtained by the analysis to determine the optimized routing path.

[0045] In this embodiment, step S105 receives the approval route determination result output from step S104, simultaneously obtains the basic data from step S102 and the feature vector from step S103, and integrates these three types of information into a unified decision input dataset, noting the urgency and terminal load. Based on this dataset, a blocking risk assessment dimension is constructed, including the historical blocking probability of the approval terminal, the current task queue length, and cross-departmental collaboration delay records, forming a risk assessment matrix. In the DQN optimization layer, the blocking probability of routes processed by the rule engine is calculated under the current risk assessment matrix. For special requests not processed by the rule engine, the blocking probability of potential routes is calculated directly based on the basic data and feature vectors. The value of each candidate route is adjusted based on the blocking probability, prioritizing routing paths with low blocking risk and consistent processing speed and pass rate. If multiple equivalent paths exist, the final optimized routing path is determined based on historical execution efficiency.

[0046] It can be seen that through the integration of multi-source data, comprehensive input is provided for risk assessment and routing optimization; blocking risk assessment enables routing decisions to take potential blocking problems into account, improving the stability of the workflow; the optimized routing path can reduce blocking risks while ensuring efficiency, thereby improving the reliability of the approval process.

[0047] Step S106: Based on the path determined by the hybrid decision routing, the approval process is executed and tasks are pushed, operation feedback is received, and a workflow visualization report is generated.

[0048] In some embodiments, based on the optimized routing path determined in step S105, the approval process execution engine is started, the approval task is pushed to the corresponding approval terminal, and the execution status of the task is tracked in real time. Feedback from the approval terminal indicating approval, rejection, return for revision, etc. is received, and information such as the approval time and approver is recorded.

[0049] This embodiment generates a workflow visualization report based on the approval process's execution and operational feedback, displaying approval progress and the status of each node. SHAP analysis is also used to analyze the routing decision process, generating a visualization report explaining the decision's rationale and key influencing factors. User rating feedback on the decision explanation is received, and the SHAP analysis results and user ratings are fed back to the user terminal. This SHAP analysis enhances decision explainability and increases user trust in the system. User rating feedback is used to optimize the model, improving the accuracy and adaptability of routing decisions and achieving continuous optimization.

[0050] In one embodiment of the present invention, based on step S103, a possible embodiment will be given below to illustrate its specific implementation scheme in a non-limiting manner. Figure 2 As shown, step S103 specifically includes: S1031: Based on the parsed basic data, combined with the approval data rule library and semantic graph, the explicit entity relationships and implicit business associations in the text content are extracted. Optionally, the implicit business associations can be based on the financial review requirements implicit in the purchase application.

[0051] It should be noted that the construction of semantic graphs can be based on the combination of syntactic analysis and approval data rule base. The approval data rule base can have approved contracts, invoices, approval roles and relationship templates. The relationships can involve contracts, associations, legal departments, etc. Approval relationships are generated through pattern matching and semantic role labeling, and support is provided for expanding library data based on new approval types.

[0052] S1032: Based on the department information involved in the basic data, collect the real-time working status of the associated approval terminal, including the current task queue length, the average processing time in the preset time period, the number of suspended tasks, and associate the department collaboration mode data.

[0053] The real-time working status of the collection-related approval terminal in this embodiment can call the load data interface of the corresponding approval terminal according to the information of the relevant departments. The collection content includes the current task queue length, the average processing time within the preset time, the number of suspended tasks, and the associated department collaboration mode data to form a real-time status indicator set to assist subsequent approval.

[0054] S1033: Filter historical cases that match the current request in terms of approval type, involved departments, and amount range from historical approval records, calculate the average processing time, abnormal interruption ratio, and user feedback score of the approval node in each case, and generate a historical efficiency feature set that includes time efficiency, stability, and satisfaction.

[0055] The historical efficiency feature generation in this embodiment adopts multi-condition matching rules to screen historical cases that match the current request in terms of approval type, involved department, and amount range, and calculates the average processing time, abnormal interruption ratio, and user feedback score of the approval node in each case to generate a historical efficiency feature set.

[0056] S1034: Input the entity relationships, real-time work status indicator set, and historical efficiency feature set of the semantic graph into the collaborative analysis module, adopt a cross-dimensional association strategy driven by business rules, adjust the weight distribution of features, and generate a feature vector that integrates content relevance, resource availability, and historical reliability.

[0057] This embodiment transforms basic data into quantifiable feature vectors through data fusion and adjustment, providing input for hybrid decision routing. By prioritizing emergency requests to low-load terminals, feature weights are adjusted. Features are first standardized, and then global weights are adjusted using association rules to generate feature vectors that meet user needs.

[0058] In one embodiment of the present invention, based on step S104, a possible embodiment will be given below to illustrate its specific implementation scheme in a non-limiting manner. Figure 3 As shown, step S104 specifically includes: Step S1041: Based on the basic data parsed in step S102 and the feature vector extracted in step S103, an index of the approval data rule library is constructed, classified and stored by business type and approval level, and main rules and auxiliary rules are marked.

[0059] This embodiment can create an index by calculating the correlation between rules, business types, and approval levels; correlation = number of rule applications / total number of rule applications. A hash table is used to store rule classification results, enabling fast query and call of rules.

[0060] Step S1042: Perform rule matching preprocessing on the basic data and feature vectors to extract key parameters related to the rules in the approval data rule base.

[0061] Step S1043: Preliminary matching of features is performed through the main rules to screen out candidate routes that meet the main conditions; candidate routes are sorted by matching degree = number of rule clauses met / number of total rule clauses; auxiliary rules retain candidate routes that meet the conditions.

[0062] Step S1044: Detect contradictions in the matching results. If there are multiple routes that meet the conditions, determine the final approval route based on the rule priority.

[0063] If no matching rule exists, the request is marked as requiring processing by the optimization layer, and the approval routing judgment result is output. Optionally, the rule priority can be such that the amount rule takes precedence over the department rule.

[0064] In this embodiment, the rule priority is determined by the hierarchical analysis method; when no matching rule is found, the request type is marked by the exception marking algorithm to ensure process continuity. Optionally, a mark value of 1 indicates a special request, and 0 indicates a normal request.

[0065] This embodiment uses basic data and feature vectors as input for indexing to achieve rapid positioning of rules; adopts a hierarchical matching strategy to accurately screen candidate routes, and then resolves multi-rule matching conflicts through a conflict detection mechanism, and finally outputs a clear approval route judgment result, while ensuring the accuracy of route judgment under clear rules.

[0066] Step S1044 of this embodiment further specifically includes the following steps: Step S10441: extracting conflict features from the candidate route set after rule matching, identifying the conflict type, and marking the conflict dimension of each candidate route; Optionally, the conflict type may include rule type conflict, business scenario conflict, or priority conflict. The conflict dimension may include simultaneous triggering of amount rule and department rule, overlapping time ranges of emergency rule and regular rule, etc.

[0067] Step S10442: Based on historical decision information and current business scenarios, re-order the weights of conflicting rules.

[0068] It's important to note that historical decision-making information can be used to consider the approval rate and average time required for rule hits over a period of time. Current business scenarios can be combined with actual approval status, such as during holidays and unexpected business peaks. Re-weighting conflicting rules can prioritize rules with high approval rates and prioritize rules with low time consumption in emergency situations.

[0069] In this embodiment, the priority adjustment method can adjust the priority in combination with historical performance and scene characteristics.

[0070] Optionally, priority score = 0.6 × historical priority score + 0.3 × hit pass rate in a preset time period + 0.1 × current scenario adaptation coefficient. Here, emergency scenario coefficient = 1.2, daily scenario coefficient = 1.0.

[0071] Among them, the historical priority score is assigned according to the preset priority of the rule; the hit pass rate within the preset time period = the number of passes after hitting / the total number of hits; the scenario adaptation coefficient is adjusted according to the current business scenario.

[0072] Step S10443: For candidate routes with conflicting priorities, a weighted scoring method is used to calculate a comprehensive score, and the route with the highest comprehensive score is selected as the final approval path.

[0073] Optionally, the weighted scoring method may use a linear weighted model to calculate the comprehensive score of candidate routes with conflicting priorities.

[0074] The specific configuration method is: comprehensive score = 0.5 × dynamic priority score + 0.3 × (historical pass rate / 100) + 0.2 × (1-average time / benchmark time).

[0075] The historical pass rate and average time are based on data from a preset time period. For example, a historical pass rate of 90% would result in a score of 0.9, an average time of 2 days would result in a score of 0.8, and a baseline time of 3 days. This embodiment can also be adjusted based on business objectives. For example, if efficiency is emphasized, the time coefficient can be increased to 0.3.

[0076] Step S10444: For unmatched requests, extract abnormal features in the basic data and key information in the feature vector, mark them as abnormal feature request marking codes, attach explanations of the conflict reasons and associated parameters, and output the approval routing judgment results containing the marking information.

[0077] Optionally, an abnormal feature request marker code is generated for requests that do not match the rules, including the abnormality type, timestamp, associated parameters and conflicting rule ID. The marker code format can be AAAA-abnormality type-timestamp-associated parameter hash value, which facilitates rapid location of problem scenarios during subsequent model training.

[0078] This embodiment solves the problem of rigid routing decisions caused by rule conflicts or incomplete coverage in the rule engine by extracting conflicting features, adjusting priorities, and scoring. It accurately marks requests that do not match the rules, implements the process of rule matching, conflict resolution, and problem marking, and improves the flexibility and accuracy of routing decisions.

[0079] In one embodiment of the present invention, based on step S105, a possible embodiment will be given below to illustrate its specific implementation scheme in a non-limiting manner. Figure 4 As shown, step S105 specifically includes: Step S1051: Receive the approval routing judgment result output by step S104, and simultaneously obtain the basic data parsed by step S102 and the feature vector extracted by step S103, integrate the three types of information into a unified decision input data set, and mark the key influencing information in the data.

[0080] Key impact information may include urgency, terminal load, etc.

[0081] This embodiment integrates information by calculating the field relevance of data from different sources. Using a key information extraction algorithm, key information is labeled based on the weight of the field's impact on routing decisions. Field relevance is configured as relevance = number of common fields / total number of fields.

[0082] Step S1052: Based on the input data set, a blocking risk assessment dimension is constructed, including the historical blocking probability of the approval terminal, the current task queue length, and the cross-departmental collaboration delay record, to form a risk assessment matrix.

[0083] This embodiment can construct a risk assessment method based on the analytic hierarchy process. The risk weight of each dimension is determined by combining historical data verification. Indicators such as the terminal's historical blocking probability and task queue length are converted into risk values ​​between 0 and 1. For example, the queue length exceeds the threshold ratio × the basic risk coefficient.

[0084] Step S1053: In the DQN optimization layer, risk matching is first performed on the routes processed by the rule engine priority layer, and the blocking probability of the routes under the current risk assessment matrix is ​​calculated; for special requests that have not been processed by the rule engine, the blocking probability of potential routes is calculated based on the basic data and feature vectors.

[0085] Alternatively, a blocking probability calculation method can be used to obtain a comprehensive probability by weighted summation of each risk dimension; for example, blocking probability = Σ(dimension risk value × dimension weight). Routing characteristics are compared with the risk assessment matrix to achieve rapid risk matching.

[0086] Step S1054: Adjust each candidate route based on the blocking probability, giving priority to routing paths with low blocking risk and that meet business objectives; if multiple equivalent paths exist, determine the final optimized routing path based on historical execution efficiency, and output the path details.

[0087] This embodiment adopts a route adjustment method based on risk impact coefficients set according to the sensitivity of services to congestion; a path sorting algorithm is used to sort paths based on the weighted sum of adjusted routes and historical execution efficiencies to determine the optimal route.

[0088] As can be seen, this embodiment receives and associates rule results, basic data, and feature vectors; extracts risk information from the terminal status library and historical database to construct an assessment matrix; configures a DQN optimization layer processing unit with built-in risk matching and value adjustment logic to calculate and rank the risk of candidate routes; and establishes a path decision module to determine and output the final route based on risk value and historical efficiency. Furthermore, within the DQN optimization layer, routing is adjusted based on blocking risk, prioritizing low-risk, high-efficiency paths. Adaptive routing decisions are also implemented for special requests, ultimately outputting an optimized routing path, improving the stability and efficiency of workflow approvals.

[0089] Step S1053 of this embodiment further specifically includes the following steps: Step S10531: A dual-channel risk mapper is established within the DQN optimization layer. The first channel receives marked routing paths, and the second channel receives special requests that are not defined by rules, and generates path state snapshots for each channel. Step S10532: injecting the real-time terminal heartbeat sequence and cross-department delay fingerprint into each path status snapshot to form a dynamic risk fingerprint; Step S10533: calling the adaptive risk metric according to the risk fingerprint and outputting the blocking probability vector of each candidate path; In this embodiment, the adaptive risk measurer can receive the risk fingerprint generated by step S10532. The risk fingerprint is a multidimensional vector containing real-time terminal heartbeat features and cross-department delay features. The measurer compares the vector with each template in the pre-built historical risk template library. The comparison method uses vector similarity calculation, specifically calculating the Euclidean distance between the current fingerprint and each template. The template with higher similarity is considered to be closer to the historical blocking scenario. Based on this, the measurer assigns a blocking probability value to each candidate path: the probability corresponding to the template with the highest similarity is assigned the largest weight, and the remaining templates contribute smaller weights in descending order of similarity. Finally, the weighted probability value constitutes the blocking probability vector of the path for subsequent risk adjustment.

[0090] Step S10534: Couple the blocking probability vector with the path service weight to generate a risk-adjusted routing value indication and cache it in the decision log.

[0091] As can be seen, step S1053 first feeds the rule-defined paths and special requests into the dual-channel risk mapper, where each path is instantly snapshotted. The system then extracts the approval terminal's heartbeat sequence and the latency fingerprint of cross-departmental collaboration in real time, overlaying them onto the corresponding snapshot to form a unique, non-replayable dynamic risk fingerprint. After reading this fingerprint, the adaptive risk metric quickly generates a blocking probability vector, which is then coupled with the service weight to generate a route value indicator that can be directly used for ranking. The entire process is completed in milliseconds, and all intermediate results are written to the decision log for subsequent auditing and visualization.

[0092] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0093] The following is an embodiment of a workflow processing system based on a routing algorithm provided by an embodiment of the present disclosure. This system and the workflow processing method based on a routing algorithm of the above-mentioned embodiments belong to the same inventive concept. For details not fully described in the embodiment of the workflow processing system based on a routing algorithm, please refer to the embodiment of the workflow processing method based on a routing algorithm.

[0094] like Figure 5 As shown, the system includes: Feature extraction module 201 is used to obtain approval requests transmitted via the API gateway, construct a semantic graph of the request content to extract entity relationships, obtain the real-time working status of the approval terminal, and calculate dynamic weights through collaborative analysis based on request history information to obtain a feature vector; Request parsing module 202, used to parse the approval request and extract basic data; Semantic fusion module 203, based on basic data, constructs a semantic graph of the request text content to extract entity relationships, obtains the real-time working status of the approval terminal based on the associated approval terminal information, combines historical approval records, calculates weights through collaborative analysis, and obtains a feature vector; Priority judgment module 204 is used to process the basic data obtained by parsing and the extracted feature vectors using the rule engine priority layer to obtain approval routing judgment; Routing decision module 205 is used to calculate the risk-adjusted routing value of the content processed by the rule engine priority layer or directly on the basic data and feature vectors obtained by parsing in the hybrid decision routing through the DQN optimization layer combined with the blocking risk prediction, and determine the optimized routing path; The execution feedback module 206 is used to execute the approval process and push tasks according to the path determined by the hybrid decision routing, receive operation feedback, and generate a workflow visualization report. Figure 6 As shown in the figure, SHAP analysis is used to analyze the routing decision process and generate a visual report explaining the decision and key factors influencing the routing decision. User feedback on the decision explanation is received, and the SHAP analysis results and user ratings are fed back to the user terminal. This SHAP analysis enhances the explainability of decisions and increases user trust in the system. User rating feedback is used to optimize the model, improving the accuracy and adaptability of routing decisions and achieving continuous optimization.

[0095] like Figure 7 As shown, the present application also provides an electronic device, including a display module 103, a memory 102, a processor 101, and a computer program stored in the memory and executable on the processor 101, wherein the processor 101 implements the steps of a workflow processing method based on a routing algorithm when executing the program.

[0096] In the embodiments of the present invention, electronic devices include, but are not limited to, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the embodiments of the present application described and / or claimed herein.

[0097] In the embodiment of the present application, the processor 101 can be implemented by using at least one of a special purpose integrated circuit, a programmable logic device, a field programmable gate array, a processor, a controller, a microcontroller, a microprocessor, and an electronic unit designed to perform the functions described herein. In some cases, such an embodiment can be implemented in a controller. For software implementation, an embodiment such as a process or function can be implemented with a separate software module that allows the execution of at least one function or operation. The software code can be implemented by a software application (or program) written in any appropriate programming language, and the software code can be stored in a memory and executed by a controller.

[0098] The display module 103 is used to display information input by the user or information provided to the user. The display module 103 may include a display panel, which may be configured in the form of a liquid crystal display, an organic light emitting diode, etc.

[0099] The memory 102 can be used to store software programs and various data. The memory 102 can include high-speed random access memory and non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0100] The present application also provides a storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the workflow processing method based on the routing algorithm are implemented.

[0101] The storage medium can be any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or component, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0102] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A workflow processing method based on routing algorithm, characterized in that: Methods include: S101: Obtain the approval request transmitted via the API gateway, construct a semantic graph of the request content to extract entity relationships, obtain the real-time working status of the approval terminal, combine the request history information, calculate the dynamic weight through collaborative analysis, and obtain a feature vector; S102: Parse the approval request to obtain basic data; S103: Based on the basic data, a semantic graph of the request text content is constructed to extract entity relationships. The real-time working status of the approval terminal is obtained based on the associated approval terminal information. Combined with historical approval records, the weight is calculated through collaborative analysis to obtain a feature vector. S104: The basic data obtained by parsing and the extracted feature vectors are processed by the rule engine priority layer to obtain the approval routing judgment; S105: In hybrid decision routing, the DQN optimization layer, combined with the blocking risk prediction, calculates the risk-adjusted routing value on the content processed by the rule engine priority layer or directly on the basic data and feature vectors obtained by parsing, and determines the optimized routing path; S106: Based on the path determined by the hybrid decision routing, execute the approval process and push the task, receive operation feedback, and generate a workflow visualization report.

2. The workflow processing method based on routing algorithm according to claim 1, characterized in that: Step S102 specifically includes: Parse the approval request data to extract the approval amount field, application reason description, request initiation time, and source system identifier to form the original data set; Based on the predefined approval data rule library, the parsed structured data is verified for context; Perform intent recognition on the original data set content, extract key business elements, and combine them with the department affiliation information of the associated approval terminal to generate urgency labels and a list of involved departments; Verify the basic data obtained from the analysis. When it is detected that key fields are missing, execute the correction process to fill in the missing information by default or associating historical data.

3. The workflow processing method based on routing algorithm according to claim 1, characterized in that: Step S103 specifically includes: Based on the basic data obtained through analysis, combined with the approval data rule library and semantic graph, explicit entity relationships and implicit business associations in the text content are extracted; Based on the department information in the basic data, collect the real-time working status of the related approval terminals and associate the department collaboration mode data; Filter historical cases that match the current request from historical approval records, calculate the average processing time, abnormal interruption ratio, and user feedback score of the approval node in each case, and generate a historical efficiency feature set; The entity relationships, real-time work status indicator set, and historical efficiency feature set of the semantic graph are input into the collaborative analysis module, the weight distribution of the features is adjusted, and a feature vector that integrates content relevance, resource availability, and historical reliability is generated.

4. The workflow processing method based on routing algorithm according to claim 1, characterized in that: Step S104 specifically includes: Based on the basic data parsed in step S102 and the feature vector extracted in step S103, an index of the approval data rule base is constructed, classified and stored by business type and approval level, and main rules and auxiliary rules are marked; Perform rule matching preprocessing on basic data and feature vectors to extract key parameters related to the rules in the approval data rule base; Perform a preliminary match of features using the main rules to filter out candidate routes that meet the main conditions; The matching results are checked for contradictions. If there are multiple routes that meet the conditions, the final approval route is determined based on the rule priority.

5. The workflow processing method based on routing algorithm according to claim 4, characterized in that: Step S1044 specifically includes: Extract conflict features from the candidate route set after rule matching, identify the conflict type, and mark the conflict dimension of each candidate route; Re-rank the weights of conflicting rules based on historical decision-making information and current business scenarios; For candidate routes with conflicting priorities, a weighted scoring method is used to calculate the comprehensive score, and the route with the highest comprehensive score is selected as the final approval path; For unmatched requests, the abnormal features in the basic data and the key information in the feature vector are extracted and marked as abnormal feature request marking codes, with additional explanations of the conflict reasons and associated parameters, and the approval routing judgment results containing the marking information are output.

6. The workflow processing method based on routing algorithm according to claim 1, characterized in that: Step S105 specifically includes: Receive the approval routing judgment result outputted in step S104, obtain the basic data parsed in step S102 and the feature vector extracted in step S103, integrate the three types of information into a decision input data set, and mark the key influencing information in the decision input data set; Based on the input decision input data set, the blocking risk assessment dimension is constructed to form a risk assessment matrix; In the DQN optimization layer, risk matching is performed on the routes processed by the rule engine priority layer, and the blocking probability of the route under the current risk assessment matrix is ​​calculated. For requests that have not been processed by the rule engine, the blocking probability of the route is calculated based on the basic data and feature vectors. Adjust each candidate route based on the blocking probability, giving priority to routing paths with low blocking risk and that meet business objectives. If multiple equivalent paths exist, determine the final optimized routing path based on historical execution efficiency and output the path details.

7. The workflow processing method based on routing algorithm according to claim 6, characterized in that: Step S1053 specifically includes: A dual-channel risk mapper is established inside the DQN optimization layer. The first channel receives marked routing paths, and the second channel receives special requests that are not defined by rules, and generates path state snapshots for each channel. Inject real-time terminal heartbeat sequences and cross-department delay fingerprints into each path status snapshot to form a dynamic risk fingerprint; Call the adaptive risk metric based on the risk fingerprint and output the blocking probability vector of each candidate path; The blocking probability vector is coupled with the path service weight to generate a risk-adjusted routing value indication and cache it in the decision log.

8. A workflow processing system based on routing algorithm, characterized in that: The system is used to implement the workflow processing method based on routing algorithm according to any one of claims 1 to 7; The system includes: The feature extraction module is used to obtain approval requests transmitted through the API gateway, build a semantic graph of the request content to extract entity relationships, obtain the real-time working status of the approval terminal, and calculate dynamic weights through collaborative analysis based on the request history information to obtain the feature vector; Request parsing module, used to parse approval requests and extract basic data; The semantic fusion module builds a semantic graph of the request text content based on basic data to extract entity relationships. It obtains the real-time working status of the approval terminal based on the associated approval terminal information, combines historical approval records, calculates weights through collaborative analysis, and obtains a feature vector. The priority judgment module is used to process the basic data obtained through analysis and the extracted feature vectors using the rule engine priority layer to obtain the approval routing judgment; The routing decision module is used in hybrid decision routing to calculate the risk-adjusted routing value based on the content processed by the rule engine priority layer or directly on the basic data and feature vectors obtained through the analysis, through the DQN optimization layer and combined with the blocking risk prediction, to determine the optimized routing path; The execution feedback module is used to execute the approval process and push tasks according to the path determined by the hybrid decision routing, receive operation feedback, and generate a workflow visualization report.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the workflow processing method based on the routing algorithm as claimed in any one of claims 1 to 7 are implemented.

10. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the workflow processing method based on the routing algorithm according to any one of claims 1 to 7 are implemented.

Citation Information

Patent Citations

  • Examination and approval task distribution method and device, computer equipment and storage medium

    CN109726983A

  • Contract approval method and device, equipment and storage medium

    CN116109277A

  • Enterprise big data analysis processing method and system based on cloud computing

    CN119396889A

  • Electronic government affair platform management method and system based on cloud data

    CN120234427A

  • Workflow condition node management method based on dynamic rule engine

    CN120235592A

Cited By

  • Fund dynamic routing decision-making system based on AI large model

    CN122175586A