Business process optimization strategy generation device and method of financial institution, terminal and medium

Abstract

The embodiment of the invention relates to the field of data processing, and provides a business process optimization strategy generation device and method of a financial institution, a terminal and a medium, and the method comprises the steps: obtaining business process node data, business interaction support data and business result feedback data of a to-be-processed financial institution; based on a dynamic condition response mechanism, performing dynamic bottleneck process mining processing according to the business process node data and the business interaction support data to obtain bottleneck process analysis data; performing bottleneck process influence quantification processing according to the bottleneck process analysis data and the service result feedback data to obtain bottleneck process priority data; performing bottleneck process association rule mining processing according to the bottleneck process priority data, the business process node data and the business interaction support data to obtain bottleneck process association relationship data; according to the bottleneck process analysis data, the bottleneck process priority data and the bottleneck process association relationship data, business process optimization processing is carried out, and a more accurate business process optimization strategy can be obtained.

Description

Technical Field

[0001] This application relates to the field of data processing technology, specifically to a business process optimization strategy generation device, method, terminal, and medium for financial institutions. Background Technology

[0002] In the digital transformation of financial institutions, business processes such as credit approval and cross-border payments are becoming increasingly complex, involving multi-system interactions and multi-role collaboration. They also need to respond to dynamic regulatory and customer demands, requiring extremely high precision in process optimization. However, current optimization practices have significant limitations, often focusing on localized data in single stages, with analysis remaining at the level of phenomenological description. Furthermore, existing systems lack the ability to quickly adapt to dynamically changing business objectives, leading to wasted resources and limited efficiency gains, failing to meet the core needs of financial businesses for precise and efficient optimization. Therefore, how to achieve more precise and efficient optimization of financial institutions' business processes has become an urgent problem to be solved. Summary of the Invention

[0003] This application provides a business process optimization strategy generation device, method, terminal, and medium for financial institutions, which can achieve more accurate and efficient optimization of financial institutions' business processes.

[0004] The first aspect of this application provides a method for intelligent optimization of business processes in financial institutions, which includes: Acquire business process node data, business interaction support data, and business result feedback data of the financial institutions to be processed; Based on the dynamic condition response mechanism, dynamic bottleneck process mining is performed on business process node data and business interaction support data to obtain bottleneck process analysis data. Based on bottleneck process analysis data and business result feedback data, the impact of bottleneck processes is quantified to obtain bottleneck process priority data. Based on bottleneck process priority data, business process node data, and business interaction support data, bottleneck process association rule mining is performed to obtain bottleneck process association relationship data. Based on bottleneck process analysis data, bottleneck process priority data, and bottleneck process correlation data, business process optimization is performed to obtain business process optimization strategies.

[0005] A second aspect of this application provides a smart optimization device for financial institution business processes, the smart optimization device for financial institution business processes includes: The acquisition unit is used to acquire business process node data, business interaction support data, and business result feedback data of the financial institution to be processed. The first processing unit is used to perform dynamic bottleneck process mining based on the business process node data and the business interaction support data, according to the dynamic condition response mechanism, to obtain bottleneck process analysis data. The second processing unit is used to perform bottleneck process impact quantification processing based on the bottleneck process analysis data and the business result feedback data to obtain bottleneck process priority data. The third processing unit is used to perform bottleneck process association rule mining processing based on the bottleneck process priority data, the business process node data and the business interaction support data to obtain bottleneck process association relationship data. The fourth processing unit is used to perform business process optimization processing based on the bottleneck process analysis data, the bottleneck process priority data, and the bottleneck process correlation data to obtain a business process optimization strategy.

[0006] A third aspect of this application provides a terminal including a processor, an input device, an output device, and a memory, wherein the processor, input device, output device, and memory are interconnected, wherein the memory is used to store a computer program, the computer program including program instructions, and the processor is configured to invoke the program instructions to execute the step instructions as described in the first aspect of this application.

[0007] A fourth aspect of this application provides a computer-readable storage medium storing a computer program for electronic data interchange, wherein the computer program causes a computer to perform some or all of the steps described in the first aspect of this application.

[0008] A fifth aspect of this application provides a computer program product, wherein the computer program product includes a non-transitory computer-readable storage medium storing a computer program operable to cause a computer to perform some or all of the steps described in the first aspect of this application. The computer program product may be a software installation package.

[0009] Implementing the embodiments of this application has the following beneficial effects: By acquiring business process node data, business interaction support data, and business result feedback data of the financial institution to be processed, a dynamic bottleneck process mining process can be performed based on a dynamic conditional response mechanism. This process yields bottleneck process analysis data, which is then used to quantify the impact of bottleneck processes, resulting in bottleneck process priority data. Furthermore, bottleneck process association rule mining can be performed based on this priority data, business process node data, and business interaction support data to obtain bottleneck process association relationship data. Finally, business process optimization can be performed based on this bottleneck process analysis data, priority data, and association relationship data, leading to more accurate business process optimization strategies. This enables more precise and efficient optimization of financial institution business processes. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This application provides a schematic diagram of the structure of an intelligent optimization method for business processes in financial institutions. Figure 2 This application provides a flowchart illustrating a method for intelligent optimization of business processes in financial institutions. Figure 3 This is a schematic diagram of the structure of a terminal provided in an embodiment of this application; Figure 4 This application provides a schematic diagram of the structure of a smart optimization device for business processes in financial institutions. Detailed Implementation

[0012] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0013] The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0014] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application can be combined with other embodiments.

[0015] To better understand the intelligent optimization method for financial institution business processes provided in this application, a brief introduction to existing financial institution business process optimization solutions is given below. Existing solutions often rely on single indicators (such as processing time and rework count) to identify bottlenecks, failing to quantify and integrate the multi-dimensional impact characteristics (efficiency, quality, experience) of bottleneck nodes, and further failing to establish feature associations with business interaction support data (system logs, interface status, etc.). This results in bottlenecks being identified only as superficial phenomena, unable to distinguish the actual impact of different nodes on the process, and struggling to identify low-frequency but high-impact key bottlenecks, leading to misallocation of optimization resources. In the bottleneck analysis and optimization implementation phase, existing technologies suffer from vague priorities and weak decision support. Bottleneck priority division often relies on subjective settings based on human experience, lacking objective quantification and dynamic calibration based on data. This leads to problems such as urgent bottlenecks not being prioritized and secondary bottlenecks consuming core resources. Simultaneously, the correlation rules between bottlenecks, process nodes, and supporting data are not deeply explored. Optimization solutions only make local adjustments to individual bottleneck nodes, failing to address root causes such as interface failures and abnormal data flow, ultimately falling into an inefficient cycle of "optimization-problem recurrence-re-optimization," making it difficult to achieve precise process upgrades.

[0016] To address the aforementioned issues, this application provides an intelligent optimization method for financial institution business processes. This method can accurately locate bottlenecks through multi-dimensional feature quantification and integration, clarify core optimization objects by combining data-driven priority division, and then delve into the root causes of bottlenecks through association rule mining. Finally, it outputs targeted optimization solutions, which not only avoids resource misallocation but also breaks the cycle of "partial adjustment - problem recurrence," effectively improving the accuracy and efficiency of financial business process optimization.

[0017] Please see Figure 1 , Figure 1 A schematic diagram of an intelligent optimization system for business processes in financial institutions is shown. Figure 1 As shown, the intelligent optimization system for financial institution business processes can include a data acquisition module, a dynamic bottleneck process mining module, a bottleneck process impact quantification module, a bottleneck process association rule mining module, and a business process optimization module. The data acquisition module collects multi-source data from the financial institution's business processes, including business process node data, business interaction support data (such as system logs and interface records), and business result feedback data (such as complaints and rework records), providing a data foundation for subsequent analysis. The dynamic bottleneck process mining module, combined with a dynamic conditional response mechanism, mines and locates bottleneck nodes such as redundancy, delays, backtracking, and blockages in the business processes from the collected business data, forming bottleneck process analysis data. The bottleneck process impact quantification module quantifies the multi-dimensional impact characteristics (such as efficiency, quality, and experience) of the mined bottleneck nodes, generating dynamically weighted bottleneck feature data, and prioritizing bottlenecks based on clustering algorithms to obtain bottleneck process priority data. The bottleneck process association rule mining module, based on bottleneck process priority data, business process node data, and business interaction support data, mines association rules between bottleneck nodes and underlying causes and business results, forming bottleneck process association relationship data. The business process optimization module can be used to generate targeted optimization solutions (such as interface optimization and node logic adjustment) for core bottlenecks based on bottleneck process priority data and correlation data, and output optimization suggestions to guide the implementation and improvement of business processes.

[0018] Please see Figure 2 , Figure 2 This application provides a flowchart illustrating a method for intelligent optimization of business processes in financial institutions. For example... Figure 2 As shown, the intelligent optimization methods for financial institution business processes include: S10: Obtain business process node data, business interaction support data, and business result feedback data from the financial institution to be processed.

[0019] The business process node data can refer to the node information of each stage of a financial institution's business process. For example, the node information of the initial review of documents and risk control review in a bank's loan approval business process. Optionally, the business process node data may include, but is not limited to, the process identifier (ID), processing rules, upstream and downstream relationships of each business process, etc., and this application does not impose any restrictions on this.

[0020] Business interaction support data can refer to the associated data that supports the operation of each process node, such as system interface response logs, material flow records, server resource utilization, etc. Business result feedback data can refer to the effect data of the final business output, such as customer complaint volume, business rework rate, process time statistics, etc., and this application does not impose any restrictions on this.

[0021] Specifically, data on the aforementioned business process nodes, business interaction support data, and business result feedback data can be collected through the business system interfaces of financial institutions, such as office automation (OA) systems, customer relationship management (CRM) systems, and core transaction systems. This application does not impose any restrictions on this. Optionally, the collected data can be cleaned and deduplicated to form a more standardized dataset that is easier to process subsequently. This application does not impose any restrictions on this.

[0022] S20: Based on the dynamic condition response mechanism, dynamic bottleneck process mining is performed according to the business process node data and the business interaction support data to obtain bottleneck process analysis data.

[0023] The Dynamic Condition Response (DCR) mechanism can be understood as a dynamic process modeling method based on event, condition, and response logic. For example, in a loan application scenario, after a loan applicant submits their documents, they can enter the initial document review process node. At this time, the system receives the loan documents submitted by the applicant (event). When the system detects that the documents are incomplete (condition), it can trigger a rejection (response). This application does not impose any restrictions on this.

[0024] Bottleneck process analysis data can refer to the relevant analysis data about bottleneck processes obtained after dynamic bottleneck process mining processing based on the DCR mechanism, according to business process node data and the business interaction support data. This bottleneck process analysis data may include various types of bottleneck nodes (such as redundant bottleneck nodes, delayed bottleneck nodes, turnaround bottleneck nodes, blocking bottleneck nodes, etc.) and their positions and relationships in the process, etc., and this application does not impose any restrictions on this.

[0025] Specifically, the DCR mechanism can be used to dynamically map business process node data to financial business events, dynamically map business interaction support data to financial business conditions and / or financial business responses, and further construct a dynamic process graph based on the mapped data. This can be compared with a preset standard process graph to identify differences (such as node redundancy and processing delays). From these differences, bottleneck nodes such as redundant bottleneck nodes (i.e., duplicate nodes), delayed bottleneck nodes (i.e., timeout nodes), backtracking bottleneck nodes (i.e., repeatedly backtracking nodes), and blocking bottleneck nodes (i.e., process interruption nodes) can be identified. The process association information of each bottleneck node can be integrated to generate the bottleneck process analysis data mentioned above.

[0026] S30: Based on the bottleneck process analysis data and the business result feedback data, perform bottleneck process impact quantification processing to obtain bottleneck process priority data.

[0027] The bottleneck process priority data can refer to relevant data that is prioritized according to the degree of impact of the bottleneck nodes. This bottleneck process priority data may include priority labels (such as urgent, high, medium, low) and confidence scores, etc., which are not limited in this application.

[0028] Specifically, bottleneck feature indicators can be extracted from bottleneck process analysis data and business result feedback data. The extracted bottleneck feature indicators can be enhanced with time series and dynamically weighted with business scenario weights. This allows for further clustering and priority classification. Furthermore, it can be calibrated with preset constraints (such as compliance constraints, resource constraints, etc.) to obtain the aforementioned bottleneck process priority data.

[0029] S40: Based on the bottleneck process priority data, the business process node data, and the business interaction support data, perform bottleneck process association rule mining to obtain bottleneck process association relationship data.

[0030] Among them, bottleneck process correlation data can refer to rule-related data used to describe the causal relationship between bottleneck nodes and business interaction support data.

[0031] Specifically, bottleneck nodes can be filtered, such as selecting high-priority bottleneck nodes (e.g., those with an urgent priority and a confidence score ≥ 80) as target bottleneck nodes. This allows for the extraction of business interaction support data (e.g., interface logs, resource parameters) associated with the target bottleneck nodes. Then, association rule mining algorithms (e.g., Association Rule Mining Algorithm Based on Apriori Principle, Apriori) can be used to mine and obtain an initial set of association rules. These rules can then be validated and filtered using historical data (e.g., relevant data of historical bottleneck nodes in business process node data) to generate and obtain the bottleneck process association relationship data. This application does not impose any limitations on this process.

[0032] S50: Based on the bottleneck process analysis data, the bottleneck process priority data, and the bottleneck process correlation data, perform business process optimization processing to obtain a business process optimization strategy.

[0033] Among them, the business process optimization strategy can refer to specific improvement solutions for each bottleneck node, including but not limited to node merging, interface upgrades, rule adjustments, etc. This application does not impose any restrictions on this.

[0034] Specifically, bottleneck process analysis data (including bottleneck types), bottleneck process priority data (including the processing order of each bottleneck node), and bottleneck process correlation data (including the root causes of each bottleneck node) can be combined to generate precise and targeted strategies. For example, for risk control review delays with urgent priority, a strategy can be generated to improve the response speed of the credit investigation interface to within 15 seconds. Optionally, the business process optimization strategy may include the specific implementation steps and expected effects of each strategy, which is not limited in this application.

[0035] By accurately identifying bottleneck nodes through a dynamic conditional response mechanism and combining them with characteristic indicators for quantitative processing and time-series analysis, the system can intelligently prioritize each bottleneck node. Then, by mining association rules, the system can pinpoint the root causes of each bottleneck node and generate targeted optimization strategies, i.e., business process optimization strategies. This solves the problems of crude bottleneck identification, subjective priority determination, and ambiguous root cause location in traditional process optimization. It can significantly improve the efficiency and quality of business process optimization in financial institutions and provide a feasible and explainable intelligent optimization solution for dynamically changing financial business scenarios.

[0036] In this embodiment, by acquiring the business process node data, business interaction support data, and business result feedback data of the financial institution to be processed, a dynamic bottleneck process mining process can be performed based on a dynamic conditional response mechanism. This process yields bottleneck process analysis data, and further, the bottleneck process impact is quantified based on the bottleneck process analysis data and the business result feedback data to obtain bottleneck process priority data. This allows for bottleneck process association rule mining based on the bottleneck process priority data, the business process node data, and the business interaction support data, resulting in bottleneck process association relationship data. Finally, business process optimization is performed based on the bottleneck process analysis data, the bottleneck process priority data, and the bottleneck process association relationship data, leading to more accurate business process optimization strategies. This enables more precise and efficient optimization of the financial institution's business processes.

[0037] In one possible implementation, during dynamic bottleneck process mining, a dynamic condition response mechanism can be used to map business process node data to dynamic business events and business interaction support data to dynamic business conditions and responses. A dynamic flowchart is then constructed based on the mapped data to present the actual business process's operational logic. This dynamic flowchart is compared with a preset standard flowchart to identify differences in node quantity, time consumption, paths, etc., generating differential process data. Bottleneck nodes can then be located from this differential process data, and their relationships within the process can be integrated to generate bottleneck process analysis data. Specifically, a method for obtaining bottleneck process analysis data by performing dynamic bottleneck process mining based on the business process node data and the business interaction support data using a dynamic condition response mechanism may include: A1. Based on the dynamic condition response mechanism, the business process node data is processed by event mapping to obtain dynamic business event data; A2. Based on the dynamic condition response mechanism, the business interaction support data is processed by condition mapping to obtain dynamic business condition data and dynamic business response data. A3. Construct a dynamic flowchart based on the dynamic business event data, the dynamic business condition data, and the dynamic business response data; A4. Perform difference identification processing based on the dynamic flowchart and the preset standard flowchart to generate difference process data; A5. Based on the difference process data, bottleneck node determination processing is performed to obtain bottleneck process analysis data.

[0038] Event mapping can be understood as the process of transforming business process node data into actions or scenarios (i.e., events) that trigger state changes within the process. Dynamic business event data refers to the set of events corresponding to the business process obtained after mapping processing.

[0039] Specifically, the node triggering rules in the business process node data can be parsed, such as extracting the start / end action of each node. For example, when a customer submits materials, the initial review of materials is triggered. The format can be: event ID + event name + triggering node (physical node of the business process) + timestamp, such as: 71 + initial review of materials node start + initial review of materials node (which can also be a node in the system, such as the “loan_check_001” node) + 2025-XX-XX 10:25:06. This application does not impose any restrictions on this.

[0040] Condition mapping can be understood as the process of extracting constraints (dynamic business condition data) and actions (dynamic business response data) from business interaction support data to determine the flow of the process. Dynamic business condition data refers to the conditional data obtained after mapping, such as data completeness ≥ 100% or interface response time ≤ 5s. Dynamic business response data refers to the response data obtained after mapping, such as the process progressing to the next node, triggering a rollback notification, or a system alarm.

[0041] Specifically, fields related to process rules (such as material verification results and interface response time) can be filtered from business interaction support data and converted into conditions to obtain the above dynamic business condition data (such as material recognition success rate < 80%). At the same time, system actions when the conditions are met / not met (such as triggering manual review if material recognition fails) are extracted to form and obtain the above dynamic business response data. This application does not impose any restrictions on this.

[0042] A dynamic flowchart can refer to a dynamic model that presents the logical relationship between the aforementioned event-condition-response in a graphical manner (it can include the association information of each node, the triggering path of the rule, etc.). Specifically, the dynamic flowchart can be constructed using visualization tools (such as a flowchart engine) based on dynamic business event data (starting point), dynamic business condition data (judgment), and dynamic business response data (result). For example, taking the submission of materials as an example, the corresponding dynamic flowchart can be: Submission of materials event → Verification of material completeness (condition) → If complete, proceed to risk control review (response) / If incomplete, return (response), to intuitively show the actual operation path of the business process. This application does not impose any restrictions on this.

[0043] A preset standard flowchart can refer to an ideal process model pre-set by various financial institutions. For example, a standard loan approval process includes three nodes, and the entire loan approval process takes ≤24 hours. Difference identification processing can be understood as comparing the dynamic flowchart with the preset standard flowchart to identify differences in process operation. Difference process data can refer to the specific information recorded after difference identification processing. This difference process data may include specific information such as node redundancy, processing delays, and path anomalies; this application does not impose any limitations on this.

[0044] Specifically, the difference can be identified by comparing the number of nodes, their order, triggering rules, and processing time of the two flowcharts. For example, if the preset standard flowchart contains three nodes ("preliminary review - risk control - final review"), while the dynamic flowchart has an additional "duplicate verification" node, this difference can be identified as a redundancy difference. If the preset standard flowchart indicates that node A takes ≤1 hour, while the dynamic flowchart shows that node A actually takes 3 hours, this difference can be identified as a delay difference. Optionally, a structured list of differences can be output based on the identified redundancy and delay differences to obtain the aforementioned difference process data; this application does not impose any restrictions on this.

[0045] In one possible implementation, the bottleneck node determination process based on the differential process data to obtain bottleneck process analysis data may include: determining redundant bottleneck nodes based on the differential process data to obtain redundant bottleneck node data; determining delayed bottleneck nodes based on the differential process data to obtain delayed bottleneck node data; determining turnaround bottleneck nodes based on the differential process data to obtain turnaround bottleneck node data; determining blocking bottleneck nodes based on the differential process data to obtain blocking bottleneck node data; and obtaining bottleneck process analysis data based on the redundant bottleneck node data, the delayed bottleneck node data, the turnaround bottleneck node data, and the blocking bottleneck node data.

[0046] Redundant bottleneck node data can refer to information about unnecessary duplicate nodes (such as the same document being verified repeatedly by two nodes). Delayed bottleneck node data can refer to information about nodes whose processing time exceeds the standard (such as risk control review taking more than 2 hours). Returned bottleneck node data can refer to information about nodes that are repeatedly returned (such as the document submission node being returned multiple times due to material issues). Blocking bottleneck node data can refer to information about nodes that cause process interruptions (such as interface failure causing the review node to stagnate).

[0047] Specifically, judgment conditions for various types of bottleneck nodes can be preset. For example, setting the number of nodes exceeding the standard as the judgment condition for redundant bottleneck nodes, setting the time exceeding the standard as the judgment condition for delayed bottleneck nodes, setting the number of backtrackings exceeding the threshold as the judgment condition for reversal bottleneck nodes, and setting the process interruption exceeding 30 minutes as the judgment condition for blocking bottleneck nodes. This application does not impose any restrictions on this.

[0048] In other words, based on the differential process data, process redundancy caused by repeated execution of unnecessary responses (such as repeated verification of the same material) can be filtered to identify redundant bottleneck node data. This redundant bottleneck node data is associated with the corresponding DCR repeated responses and triggering conditions. Similarly, based on the differential process data, process delays caused by response execution times exceeding preset thresholds (such as credit data retrieval responses exceeding 20 minutes) can be filtered to identify delay bottleneck node data. This delay bottleneck node data is associated with the corresponding DCR responses and time consumption data. Furthermore, based on the differential process data, processes that repeatedly trigger return responses at the same node due to unmet conditions (such as material identification failures) can be filtered. Process reversals caused by power consumption <80% (resulting in multiple rejections of data submission nodes) can be identified to determine bottleneck node data. This bottleneck node data is associated with the corresponding DCR rejection response, triggering condition (material identification failure), and reversal count statistics. Based on the differential process data, process blockages caused by missing key conditions or system anomalies preventing response execution (e.g., cross-system interface failure preventing risk control review nodes from obtaining credit data, thus hindering process progress) can be identified to determine blocking bottleneck node data. This blocking bottleneck node data is associated with the corresponding DCR non-execution response, the reason for missing triggering condition (interface failure), and the duration of the blockage.

[0049] Furthermore, the redundant bottleneck node data (including repeated responses and triggering conditions), delayed bottleneck node data (including response time and associated events), backtracking bottleneck node data (including returned responses, triggering conditions and backtracking times), and blocked bottleneck node data (including unexecuted responses, reasons for missing conditions, and blocking duration) can be integrated to generate bottleneck process analysis data with DCR dynamic relationship annotations.

[0050] By accurately screening and quantifying four types of bottleneck nodes—redundancy, delay, backtracking, and blockage—and linking each node's data to corresponding DCR dynamic logic (response type, triggering conditions, and unique indicators), analytical data with DCR annotations is generated. This not only achieves precise location of node-level bottlenecks but also reveals the condition-response root causes of bottlenecks through the dynamic relationship of DCRs (such as backtracking due to repeated triggering of the "material identification failure" condition). This solves the problem in traditional process analysis where the existence of bottlenecks is known but the cause of their occurrence is unknown. It provides a dual basis for subsequent priority quantification and optimization strategy generation, combining specific nodes and dynamic logic, significantly improving the depth and practicality of bottleneck analysis.

[0051] It should be noted that this application uses bottleneck nodes, including redundant bottleneck nodes, delayed bottleneck nodes, turnaround bottleneck nodes, and blocking bottleneck nodes, as examples for illustration, and does not constitute a limitation on this application. Optionally, other different types of bottleneck nodes can be determined according to different business process requirements, and this application does not impose any restrictions on this.

[0052] In this embodiment, by transforming business data into an event-condition-response logical structure through a dynamic condition response mechanism, the limitations of traditional static process analysis are overcome. By mapping business process node data with business interaction support data, the dynamic rules of process operation (such as node triggering conditions and abnormal response paths) can be accurately captured, avoiding a coarse analysis of the surface process. By comparing the constructed dynamic process graph with the preset standard process graph, four types of bottleneck nodes—redundancy, delay, backtracking, and blockage—can be systematically identified, and their process association information can be integrated to achieve refined mining from discovering deviations to locating specific bottleneck nodes. Compared with traditional methods that can only identify macro-level problems such as excessively long overall process time, the method provided in this application can clearly identify which node is redundant, why it is delayed, and why it is blocked. This can provide accurate problem targets for subsequent quantitative analysis and optimization strategy generation, significantly improving the accuracy and operability of bottleneck mining in financial business processes and laying a data foundation for subsequent optimization.

[0053] In one possible implementation, when quantifying the impact of bottleneck processes, efficiency, quality, and experience—three types of impact indicators—can be extracted from bottleneck process analysis data and business result feedback data to form feature data. These three types of feature data are then further integrated into three-dimensional bottleneck feature data. This enhances the feature dimensions by incorporating temporal patterns, allowing for the initial determination of reference priorities based on the enhanced feature data. Furthermore, these reference priorities can be calibrated using preset constraints (such as compliance and resources) to obtain more accurate bottleneck process priority data. Specifically, a method for quantifying the impact of bottleneck processes based on the bottleneck process analysis data and the business result feedback data to obtain bottleneck process priority data may include: B1. Based on the bottleneck process analysis data and the business result feedback data, efficiency impact indicators are extracted and processed to obtain efficiency impact feature data; B2. Based on the bottleneck process analysis data and the business result feedback data, quality impact indicators are extracted and processed to obtain quality impact characteristic data; B3. Based on the bottleneck process analysis data and the business result feedback data, perform experience impact index extraction and processing to obtain experience impact feature data; B4. Based on the efficiency impact feature data, the quality impact feature data, and the experience impact feature data, perform fusion processing to obtain three-dimensional bottleneck feature data; B5. Perform temporal feature enhancement processing on the three-dimensional bottleneck feature data to obtain temporally enhanced bottleneck feature data; B6. Based on the bottleneck process analysis data and the time-series enhanced bottleneck feature data, bottleneck priorities are divided to obtain reference bottleneck process priority data. B7. Based on preset constraints, the reference bottleneck process priority data is calibrated to obtain bottleneck process priority data.

[0054] The efficiency impact indicator can also be understood as the bottleneck delay contribution, which measures the proportion of delays caused by a single bottleneck node in the overall business process. It is a core indicator for judging the degree of impact of a bottleneck on process efficiency. For example, if a bottleneck causes 60% of the total delay in loan approval, the delay contribution of that bottleneck is 60%. Efficiency impact characteristic data can refer to structured data containing the delay contribution of each bottleneck node. This efficiency impact characteristic data may include, but is not limited to, the bottleneck node ID, the corresponding business process, and the delay contribution value. This application does not impose any restrictions on this.

[0055] Specifically, the delay time of each bottleneck node can be obtained from the bottleneck process analysis data, such as the difference between the actual time and the standard time of the bottleneck node. The total delay time of the overall process can be obtained from the business result feedback data. Then, the delay contribution of each bottleneck can be calculated by: single bottleneck delay time ÷ total delay time of the overall process × 100%. The data can be organized in the format of "node ID - delay contribution" to obtain real-time efficiency impact characteristic data.

[0056] Quality impact metrics can also be understood as bottleneck-related rework rates, indicating the probability of business process rework due to bottleneck node issues and reflecting the degree of damage bottleneck nodes cause to business quality. For example, in a bottleneck-related loan application, if 20% require rework, its associated rework rate is 20%. Quality impact characteristic data can refer to structured data containing the associated rework rates of each bottleneck node. This quality impact characteristic data may include, but is not limited to, bottleneck node ID, number of reworks, total business volume, and rework rate values, etc., and this application does not impose any restrictions on this.

[0057] Specifically, bottleneck nodes with problems can be located from bottleneck process analysis data, such as turnaround bottleneck nodes. Business records associated with the node can be filtered from business result feedback data, such as loan applications processed by the node. This allows for further statistics on the number of reworks of the node's associated business and the total number of associated business transactions. The associated rework rate can be calculated by dividing the number of reworks by the total number of transactions by 100%, thereby forming and obtaining the aforementioned quality impact characteristic data.

[0058] Experience impact metrics can also be understood as bottleneck complaint correlation, which indicates the probability of a customer complaint being associated with a bottleneck node, reflecting the negative impact of the bottleneck on customer experience. For example, if 30% of customer complaints can be traced back to a bottleneck, then the complaint correlation for that bottleneck is 30%. Experience impact feature data can refer to structured data containing the complaint correlation of each bottleneck node. This experience impact feature data may include, but is not limited to, bottleneck node ID, number of associated complaints, total number of complaints, and complaint correlation value, etc., and this application does not impose any restrictions on this.

[0059] Specifically, customer complaint records can be extracted from business result feedback data, which may include the reason for the complaint and the business process involved. Bottleneck nodes corresponding to the complaints can be matched from bottleneck process analysis data to further count the number of complaints associated with a certain bottleneck node and the total number of customer complaints. The complaint correlation degree can be calculated by: number of associated complaints ÷ total number of complaints × 100%, and then the above-mentioned experience impact characteristic data can be generated and obtained.

[0060] Three-dimensional bottleneck feature data refers to multi-dimensional bottleneck evaluation data formed by integrating three types of influencing feature data: efficiency, quality, and experience. It can be understood that in three-dimensional bottleneck feature data, each bottleneck node can include three sets of core indicators: delay contribution, associated rework rate, and complaint correlation.

[0061] Specifically, the bottleneck node ID can be used as a unique identifier. Fields such as the delay contribution in the efficiency impact feature data, the associated rework rate in the quality impact feature data, and the complaint correlation in the experience impact feature data can be concatenated to form three-dimensional bottleneck feature data in which each data point contains node ID, delay contribution, associated rework rate, and complaint correlation, ensuring that the data dimensions are complete and correspond one-to-one.

[0062] Time-series feature enhancement processing can be understood as the process of combining time dimensions (such as daily / weekly / monthly, peak / off-peak periods) to analyze the changing patterns of three-dimensional bottleneck feature data and supplement time-series information. For example, a bottleneck may contribute more to delays during peak loan approval periods than during off-peak periods. Time-series enhanced bottleneck feature data can refer to three-dimensional bottleneck feature data after supplementing with time-series patterns. This time-series enhanced bottleneck feature data may include, but is not limited to, the original three-dimensional indicators, time-period features, and indicator time-series fluctuation values; this application does not impose any restrictions on this.

[0063] Specifically, time stamps, such as business processing time periods and dates, can be extracted from business result feedback data and further correlated with three-dimensional bottleneck feature data. For example, data can be grouped by peak periods (e.g., 9:00-11:00) and off-peak periods (e.g., 14:00-16:00) to further calculate the mean, maximum, and other fluctuation characteristics of the three-dimensional indicators of bottleneck nodes within each period. This allows the time stamps and fluctuation characteristics to be added to the three-dimensional bottleneck feature data, resulting in the aforementioned time-series enhanced bottleneck feature data. This application does not impose any limitations on this. Reference bottleneck process priority data can refer to the preliminary bottleneck priority results based on the time-series enhanced bottleneck feature data. For example, it can be initially divided into four priorities: urgent, high, medium, and low. It should be noted that this reference bottleneck process priority does not yet consider actual business constraints and needs to be further calibrated based on actual business constraints in subsequent steps.

[0064] Specifically, based on financial business objectives, such as the current priority of improving efficiency, dynamic weights can be set for the three-dimensional indicators of time-series enhanced bottleneck characteristic data, such as efficiency weight of 40%, quality weight of 30%, and experience weight of 30%. The comprehensive score of each bottleneck can be calculated by multiplying the indicator value by the corresponding weight. Priority intervals can be initially divided according to the comprehensive score, such as a score ≥80 being urgent, 60-79 being high, 40-59 being medium, and a score ≤39 being low. The aforementioned reference bottleneck process priority data can then be generated and obtained. This application does not impose any restrictions on this.

[0065] Pre-set constraints refer to the limitations that financial institutions pre-define based on business rules, regulatory requirements, and current resource availability, affecting the priority of bottleneck optimization. It is understood that these pre-set constraints are not to be arbitrarily violated and directly determine the feasibility of the optimization plan. These pre-set constraints may include, but are not limited to, compliance constraints and resource constraints. Compliance constraints can refer to insurmountable limitations imposed by regulatory requirements, such as prioritizing bottlenecks involving customer privacy; resource constraints can refer to limitations imposed by limited human, technological, and financial resources, such as only being able to optimize two system-related bottlenecks in the short term. This application does not impose such limitations. It is understood that the priority results that can be directly used for business optimization after calibration with pre-set constraints are the aforementioned bottleneck process priority data.

[0066] Specifically, the priority data of the reference bottleneck process can be adjusted according to preset constraints. For example, if a certain medium-priority bottleneck node involves compliance risks, such as a customer information verification bottleneck node violating data security regulations, the medium-priority bottleneck node can be upgraded to an emergency priority bottleneck node. If an emergency priority bottleneck node requires a lot of technical resources, such as the need to modify the core trading system, but resources are insufficient in the short term, the emergency priority bottleneck node can be temporarily downgraded to a high priority bottleneck node, thereby achieving calibration based on preset constraints and obtaining the aforementioned bottleneck process priority data.

[0067] In this embodiment, the process design, which involves three-dimensional indicator extraction, multi-dimensional fusion, temporal enhancement, and constraint calibration, overcomes the limitations of traditional single-dimensional and subjective prioritization. It not only quantifies the impact of bottleneck nodes from three core dimensions—efficiency, quality, and experience—avoiding evaluation biases caused by single indicators (such as focusing solely on efficiency while ignoring customer complaints), but also captures the fluctuation patterns of bottlenecks at different times by combining temporal characteristics (such as the greater impact of delay bottlenecks during peak hours), making priority allocation more aligned with actual business operation scenarios. Furthermore, through calibration using preset constraints (such as compliance constraints and resource constraints), it ensures that the priority results are not only reasonable but also feasible, effectively solving the problem of priorities being disconnected from actual business operations. This allows the generated bottleneck process priority data to accurately guide financial institutions in prioritizing process optimization, ensuring that the optimization direction aligns with business objectives while avoiding resource waste, significantly improving the efficiency and feasibility of business process optimization.

[0068] In one possible implementation, when prioritizing bottlenecks, the time-series enhanced bottleneck feature data can be dynamically weighted by combining preset base weights with current business objectives to obtain dynamically weighted bottleneck feature data. Then, based on this dynamically weighted bottleneck feature data, K financial business cluster centers are determined, allowing the calculation of the Euclidean distance and corresponding confidence level between each bottleneck node and the cluster center. Finally, the bottleneck priorities can be determined by combining the Euclidean distance and confidence level to obtain reference bottleneck process priority data. Specifically, a method for prioritizing bottlenecks based on the bottleneck process analysis data and the time-series enhanced bottleneck feature data to obtain reference bottleneck process priority data may include: C1. Based on the preset basic weights and the current business objectives, the time-series enhanced bottleneck feature data is dynamically weighted to obtain dynamically weighted bottleneck feature data. C2. Based on the dynamically weighted bottleneck feature data, perform cluster center determination processing to obtain K financial business cluster centers; C3. Based on the bottleneck nodes in the bottleneck process analysis data, perform Euclidean distance calculation processing according to the dynamic weighted bottleneck feature data and the K financial business cluster centers to obtain the business Euclidean distance set. C4. Calculate the confidence level for each service Euclidean distance in the service Euclidean distance set to obtain the service confidence level set. C5. Based on the Euclidean distance set and the confidence set of the business, bottleneck priorities are divided to obtain reference bottleneck process priority data.

[0069] The preset base weights can refer to the benchmark weights of characteristic indicators for bottleneck nodes set by financial institutions based on historical experience. For example, the three-dimensional (efficiency, quality, experience) characteristic indicators for bottleneck nodes can be set based on historical experience, such as a benchmark weight of 40% for efficiency characteristic indicators, 30% for quality characteristic indicators, and 30% for experience characteristic indicators. This application does not impose any restrictions on this. Current business objectives can refer to the phased key optimization directions of business processes. For example, the key optimization direction for this quarter is to improve customer experience, and the focus for this year is to improve business process efficiency. This application does not impose any restrictions on this.

[0070] Dynamic weighting can be understood as adjusting preset base weights based on current business objectives, and then assigning new weights to various indicators of time-series enhanced bottleneck feature data based on these adjusted weights. Dynamically weighted bottleneck feature data refers to the bottleneck feature data that reflects scenario priority after dynamic weighting. For example, if the current business objective emphasizes user experience, the weight of the experience indicator can be increased from the original base weight of 30% to 50%, and this application does not impose any restrictions on this. It is understood that this dynamically weighted bottleneck feature data can be bottleneck feature data that integrates three-dimensional indicator features, time-series patterns, and scenario weights, and can be the feature data that best reflects bottleneck differences.

[0071] For example, if the current business objective is to reduce customer complaints, the weight of the experience impact indicator (complaint correlation) can be increased, such as from 30% to 50%, while the weight of the efficiency indicator can be decreased, such as from 40% to 20%. This will multiply the values ​​of each indicator in the time-series enhanced bottleneck feature data by the corresponding adjusted weights to obtain the aforementioned dynamically weighted bottleneck feature data.

[0072] Cluster center determination can be understood as the process of extracting core data points representing different priority categories from dynamically weighted bottleneck feature data using a clustering algorithm. The clustering algorithm can be the K-means clustering algorithm (K-means). K-means can quickly divide high-dimensional data (such as dynamically weighted bottleneck feature data) into K clusters with similar features, and the number of clusters can be objectively determined through methods such as the elbow rule. This better suits the need for batch classification of bottleneck nodes based on their impact features in financial scenarios, and this application does not impose any restrictions on this.

[0073] In the context of financial business, a cluster center refers to a data center representing bottleneck nodes with similar characteristics (such as similar priorities) calculated from a dataset reflecting the impact characteristics of bottleneck nodes (such as the dynamically weighted bottleneck feature data mentioned above) using a clustering algorithm (such as K-means adopted in this application). K is the number of clusters determined by combining the actual needs of financial business and data characteristics, and K is a positive integer.

[0074] Specifically, the elbow rule can be used to analyze the clustering error of dynamically weighted bottleneck feature data to determine the optimal number of clusters K (e.g., K=4). Then, the K-means algorithm can be used to cluster the dynamically weighted bottleneck feature data to obtain k financial business cluster centers. Each financial business cluster center can correspond to a set of dynamically weighted bottleneck features.

[0075] For example, the value of K can be determined through a dual logic of business goal anchoring and algorithm verification optimization to ensure a high degree of fit with the actual application requirements of bottleneck priority division. For instance, a reference value for K can be initially anchored based on the business optimization goals and management habits of financial institutions. If the business routinely divides bottleneck priorities into four levels—urgent, high, medium, and low—to match the degree of refinement in resource allocation (e.g., prioritizing core resources for urgent bottlenecks and delaying the processing of low-priority bottlenecks), then K can be initially set to 4. This reference value can then be further verified and adjusted using the elbow rule, i.e., performing multiple clustering operations on dynamically weighted bottleneck feature data, plotting a curve with clustering error (e.g., sum of squares within clusters) on the vertical axis and K value on the horizontal axis. The K value corresponding to the inflection point where the rate of error decreases from fast to slow is the optimal value that balances clustering effect and computational efficiency. For example, if the error decreases significantly when K increases from 3 to 4, and the error decreases more gradually when K increases from 4 to 5, then the inflection point K=4 can be considered optimal.

[0076] Optionally, historical optimization data from financial operations can be used for auxiliary calibration. For example, if a 5-level priority division in historical data easily leads to resource dispersion, and a 3-level division is too coarse, then the rationality of K=4 can be further confirmed. It should be noted that the determined K value can ensure a one-to-one correspondence between cluster centers and priority levels in the business (e.g., K=4 corresponds to 4 priority cluster centers), and can also ensure through algorithms that the bottleneck nodes represented by each cluster center have the highest similarity in features and the most significant differences between categories, thus laying the foundation for subsequent accurate classification of bottleneck process priority data.

[0077] Bottleneck nodes in bottleneck process analysis data refer to specific, identified abnormal nodes (such as redundant nodes, delayed nodes, backtracking nodes, and blocked nodes) extracted from bottleneck process analysis data with DCR dynamic relationship annotations. It should be noted that each bottleneck node is associated with a unique identifier (such as a node ID) and basic attributes (such as the business process it belongs to and the corresponding DCR response type). This bottleneck node can be viewed as the target object for Euclidean distance calculation; that is, which node's Euclidean distance is being calculated.

[0078] The Euclidean distance calculation can be understood as the process of calculating the spatial distance between the dynamically weighted features of each bottleneck node and each financial business cluster center based on the Euclidean distance formula (applicable to distance calculation of high-dimensional data). It should be noted that a smaller distance indicates a greater similarity between the features of the bottleneck node and the features of its corresponding cluster center, meaning the corresponding cluster center can be the priority category to which the bottleneck node belongs; a larger distance indicates a less similarity between the features of the bottleneck node and the features of its corresponding cluster center. This application does not impose any restrictions on this.

[0079] The set of business Euclidean distances may include one or more business Euclidean distances, which may refer to the Euclidean distance results between the bottleneck node and K financial business cluster centers. Optionally, the business Euclidean distance may include: bottleneck node ID, the distance between the bottleneck node and cluster center 1, the distance between the bottleneck node and cluster center 2, and so on up to the distance between the bottleneck node and cluster center K. This application does not impose any restrictions on this.

[0080] Confidence score calculation can be understood as measuring the probability that a bottleneck node belongs to a certain financial business cluster center (priority level). In other words, the smaller the distance between the bottleneck node and financial business cluster center 1, the higher the confidence level that the bottleneck node belongs to cluster center 1; conversely, the larger the distance, the lower the confidence level. It's important to note that confidence score calculation eliminates the absolute differences in the magnitude of the Euclidean distance values, transforming them into a more easily understood probability indicator, thus providing a more intuitive view of the financial business cluster center to which the bottleneck node belongs.

[0081] The business confidence set can include one or more business confidence scores, which can refer to the probability data of each bottleneck node belonging to each cluster center. For example, taking K=4 (urgent, high, medium, low) as an example, the confidence score of bottleneck node A belonging to the urgent category is 92%, the confidence score of belonging to the high category is 65%, the confidence score of belonging to the medium category is 45%, and the confidence score of belonging to the low category is 15%. Thus, the reliability of the category assignment of bottleneck node A can be intuitively judged.

[0082] Optionally, the process of calculating the confidence level for each service Euclidean distance in the service Euclidean distance set to obtain the service confidence level set can be found in the following formula: Where conf(i,j) can represent the business confidence that the i-th bottleneck node in the business confidence set belongs to the j-th cluster center (corresponding to priority), and the value range can be 0-100%. The higher the confidence value, the more reliable the attribution; i can represent the index of the bottleneck node; j can represent the index of the cluster center when calculating for a certain target cluster center, such as when calculating the business confidence of the j-th cluster center; `dist(i,k)` can represent the maximum distance between the i-th bottleneck node and all K cluster centers; `k` can represent the index corresponding to traversing all K financial business cluster centers, such as when calculating the Euclidean distance between the i-th bottleneck node and each financial business cluster center; `K` can represent the number of financial business cluster centers, such as K=4, which can correspond to 4 priorities: urgent, high, medium, and low; `dist(i,k)` can represent the Euclidean distance between the dynamically weighted feature data of the i-th bottleneck node and the feature data of the k-th financial business cluster center. The smaller the distance, the more similar the bottleneck node is to the priority category represented by the cluster center; `dist(i,j)` can represent the Euclidean distance between the dynamically weighted feature data of the i-th bottleneck node and the feature data of the j-th financial business cluster center. It can represent the minimum distance between the i-th bottleneck node and all K cluster centers.

[0083] It should be noted that the calculation of the aforementioned Euclidean distance relies on two types of core data. For example, when calculating dist(i,k), it is necessary to clarify the dynamically weighted bottleneck feature data of the i-th bottleneck node. This can be obtained by weighting the time-series enhanced bottleneck feature data (including efficiency, quality, and experience three-dimensional indicators) based on preset basic weights and current business objectives. The feature data of the k-th financial business cluster center can be obtained by clustering the dynamically weighted bottleneck feature data and obtaining the three-dimensional feature mean of each center, such as the weighted delay contribution mean and weighted complaint correlation mean of the emergency priority cluster center, to provide a quantitative basis for the bottleneck node to belong to the priority category. In other words, by calculating the Euclidean distance (i.e., dist(i,1), dist(i,2)...dist(i,K)) between the bottleneck node i and all K cluster centers (k=1 to K), the cluster center with the smallest distance can be selected. The priority category (e.g., urgency) corresponding to this center is the reference priority of the bottleneck node. At the same time, the business confidence level is further calculated by combining these distances to finally determine the bottleneck process priority data, ensuring the objectivity and accuracy of priority division. This application does not impose any restrictions on this.

[0084] The bottleneck priority results further determined based on the Euclidean distance and business confidence level are the aforementioned reference bottleneck process priority data. This reference bottleneck process priority data may include the bottleneck node ID, reference priority level, and business confidence level; this application does not impose any limitations on these.

[0085] In this embodiment, a multi-layered logic of dynamic weighting, cluster analysis, and distance and confidence verification solves the problems of fixed weights and reliance on subjective judgment in traditional priority allocation. The dynamic weighting mechanism allows feature data to flexibly adapt to current business objectives (such as focusing on efficiency or experience), avoiding a one-size-fits-all weight setting. Furthermore, data-driven objective classification is achieved through cluster centers and Euclidean distance, and confidence screening ensures the reliability of priority allocation (such as retaining only high-confidence classification results). Compared with traditional methods, the method provided in this application reflects the dynamism of business scenarios while ensuring the objectivity and accuracy of priorities. It provides a basis and interpretable priority ranking for subsequent bottleneck optimization, enabling financial institutions to focus on core bottlenecks and efficiently allocate optimization resources.

[0086] In one possible implementation, when performing bottleneck process association rule mining, core target bottleneck nodes can be filtered based on bottleneck process priority data, and supporting data associated with the target nodes can be extracted from business interaction support data. This supporting data is then discretized to meet the requirements of the mining algorithm, allowing for further mining of initial association rules based on the discretized data. These rules can then be combined with business process node data to filter and verify the rules, resulting in usable bottleneck process association data. Specifically, a method for performing bottleneck process association rule mining based on the bottleneck process priority data, the business process node data, and the business interaction support data to obtain bottleneck process association data may include: D1. Based on the bottleneck process priority data, perform bottleneck node filtering to obtain target bottleneck node data; D2. Extract the supporting data associated with the target bottleneck node data from the business interaction supporting data to obtain the target bottleneck node supporting data set; D3. Discretize the support data of each target bottleneck node in the target bottleneck node support data set to obtain a discrete correlation data set. D4. Perform association rule mining processing on each discrete association data in the discrete association data set to obtain an initial association rule set; D5. Based on the business process node data and the initial set of association rules, perform rule filtering and verification to obtain bottleneck process association data.

[0087] The bottleneck node screening process can be understood as the process of selecting core nodes from all bottleneck nodes that require focused association rule mining, based on business optimization goals (such as prioritizing the resolution of high-impact bottlenecks). Target bottleneck node data can include data corresponding to one or more target bottleneck nodes. This target bottleneck node data refers to the selected bottleneck nodes that better align with the current business optimization goals (and may include the bottleneck node ID, the associated business process, and the corresponding DCR response type, etc.). Target bottleneck nodes can be considered the core objects for subsequent association mining.

[0088] Specifically, priority filtering thresholds can be set according to business optimization goals. For example, only urgent and high-priority bottleneck nodes can be retained. Bottleneck nodes that meet the priority filtering thresholds can be extracted from the bottleneck process priority data. For example, bottleneck nodes such as risk control review blockage (urgent) and data submission return (high) can be filtered out to obtain the above-mentioned target bottleneck node data.

[0089] Related supporting data extraction can be understood as the process of filtering and extracting data from business interaction supporting data that is temporally and logically related to the target bottleneck node. For example, interface call data contemporaneous with the risk control review blocking bottleneck node. The target bottleneck node supporting data set may include one or more target bottleneck node supporting data, which can be understood as supporting data directly related to the target bottleneck node. For example, credit reporting interface call logs and cross-system data transmission records corresponding to the risk control review blocking bottleneck node; this application does not impose any limitations on this.

[0090] Specifically, the business processing timestamp and node ID of the target bottleneck node can be used as association keys to match supporting data from the same period and business link in the business interaction supporting data. For example, for a document submission return bottleneck node, material identification result logs and manual return operation records during the processing period of the bottleneck node can be extracted to further summarize and form a supporting data set for the document submission return bottleneck node.

[0091] Discrete processing can be understood as transforming continuous supporting data (such as an interface response time of 20 minutes or a material identification time of 5 seconds) into discrete intervals or classification labels, making it suitable for the processing of association rule mining algorithms (such as the Apriori algorithm, which requires discrete data). The discrete association data set can include one or more discrete association data sets, which can be understood as supporting data presented in a discrete form after discrete processing. Optionally, the discrete form can be similar to a "transaction-itemset" format, such as Transaction 1: Material identification time > 3 seconds (item 1), Material identification failed (item 2), Data submission returned (item 3). This application does not impose any restrictions on this.

[0092] Specifically, business thresholds can be set for continuous indicators in the business interaction support data, such as interface response time ≤ 10 minutes (normal), 10 minutes < duration ≤ 30 minutes (mild delay), and duration > 30 minutes (severe delay). This allows continuous values ​​to be converted into discrete labels, and classification labels can be directly retained for categorized data, such as material identification results: success / failure. Thus, discrete indicators and target bottleneck nodes can be integrated on a per-business-transaction basis (such as a loan application), forming and obtaining the aforementioned discrete related data set.

[0093] Association rule mining can be understood as the process of using association mining algorithms to extract potential patterns of simultaneous occurrence or causal relationship between bottleneck-related elements from a discrete association data set (core data after screening, association, and discretization), and transforming them into association rules to form an initial set of association rules that can explain the causes of bottlenecks.

[0094] Optionally, the association rules in this application can be presented in the form of "antecedent → consequent (support, confidence)". The antecedent is the triggering element of the rule, specifically referring to the preconditions or abnormal states extracted from the discrete association data set that may induce a bottleneck. For example, for a blocked risk control review node, the antecedent could be a credit reporting interface response time > 20 minutes. This element comes from the discretized results of the credit reporting interface call logs and is a potential cause of the bottleneck. The consequent is the result element of the rule, directly corresponding to the specific bottleneck phenomenon in the target bottleneck node data, i.e., the core bottleneck that needs to be analyzed after priority screening. For example, again taking the blocked risk control review node as an example, this bottleneck itself is the consequent in the association rule, a business result that is highly likely to occur after the antecedent is triggered, directly anchoring the bottleneck problem that needs to be focused on in the analysis.

[0095] Support reflects the prevalence of a rule and can be calculated based on the frequency with which the antecedent and consequent occur simultaneously in a discrete, correlated dataset. For example, in 1000 discrete business records, if a credit reporting interface response time > 20 minutes and a risk control review node blockage occur simultaneously 80 times, then the support of this rule is 8%, indicating that this type of association is not accidental in business and has a certain degree of prevalence. Confidence reflects the reliability of a rule and can refer to the probability that the consequent occurs when the antecedent occurs. For example, in discrete, correlated data, if a credit reporting interface response time > 20 minutes occurs 100 times, and 90 of these times cause a risk control review node blockage, then the confidence of this rule is 90%, indicating that the triggering effect of the antecedent on the consequent has high reliability and can serve as a key basis for analyzing the causes of bottlenecks.

[0096] The initial association rule set may include one or more initial association rules, which can refer to rules generated through association rule mining. It should be noted that initial association rules can be selected through mathematical statistics, that is, association rules that meet the minimum support and minimum confidence requirements, but have not yet been verified by business logic.

[0097] Specifically, a minimum support level can be set, such as 5%, meaning the association rule occurs at a frequency of ≥5% in all transactions. A minimum confidence level can also be set, such as 70%, meaning the probability of the consequent occurring is ≥70% when the antecedent occurs. Then, an association rule mining algorithm can be used to filter frequent itemsets layer by layer, such as material identification failures and data submission returns, to generate candidate rules based on frequent itemsets. Rules that meet the threshold are then retained to form and obtain the above-mentioned initial set of association rules.

[0098] Rule filtering and validation can refer to combining business process logic to eliminate statistically significant but business-meaningless rules, ensuring that the rules conform to the actual business process. The set of related rules that, after verification by business process logic, can be directly used to guide process optimization constitutes the bottleneck process correlation data. This bottleneck process correlation data may include correlations such as bottleneck cause → bottleneck node, and bottleneck node → business result; this application does not impose any restrictions on this.

[0099] Specifically, by comparing the business process node data, the rationality of the business logic of the initial association rules can be judged. For example, rules that exclude non-adjacent node associations, such as account opening node delay → risk control review blockage (the two have no direct business dependency), can be verified to ensure consistency between the rules and DCR logic. For example, if the credit scoring interface is severely delayed → risk control review blockage, DCR rules that rely on credit scoring data for risk control review need to be matched. Finally, rules with high statistical indicators and business interpretability are retained to form and obtain the above-mentioned bottleneck process association data.

[0100] In this embodiment, the process design of priority screening, related data extraction, discretization, algorithm mining, and business verification ensures both the core nature of the mining objects (i.e., focusing on high-priority bottlenecks) and the reliability of the rules (i.e., combining business logic verification and prioritization to avoid ineffective mining of low-impact bottlenecks, focusing on the core issues that financial businesses need to address most (such as urgent blocking nodes), and improving mining efficiency). From supporting data extraction to business verification, the process is always anchored to the actual business scenario (e.g., discretizing data based on business thresholds and verifying rules against process logic), effectively avoiding the problem of statistical rules being disconnected from business in traditional mining (e.g., eliminating false rules with no business relevance). As a result, the generated bottleneck process correlation data can accurately reveal the root causes of bottlenecks (e.g., severe delays in the credit reporting interface are the core reason for risk control review blockage) and their chain effects (e.g., material recognition failure → document submission return → customer complaints). This provides data support for financial institutions to formulate targeted optimization solutions (e.g., optimizing the stability of the credit reporting interface and improving the accuracy of material recognition algorithms), significantly improving the accuracy and effectiveness of bottleneck optimization.

[0101] For examples consistent with the above embodiments, please refer to... Figure 3 , Figure 3 This is a schematic diagram of the structure of a terminal provided in an embodiment of this application, such as... Figure 3 As shown, it includes a processor, an input device, an output device, and a memory, which are interconnected. The memory is used to store a computer program, which includes program instructions. The processor is configured to call the program instructions. The program includes instructions for performing the following steps. Acquire business process node data, business interaction support data, and business result feedback data of the financial institutions to be processed; Based on the dynamic condition response mechanism, dynamic bottleneck process mining is performed according to the business process node data and the business interaction support data to obtain bottleneck process analysis data. Based on the bottleneck process analysis data and the business result feedback data, the impact of the bottleneck process is quantified to obtain the bottleneck process priority data. Based on the bottleneck process priority data, the business process node data, and the business interaction support data, bottleneck process association rule mining is performed to obtain bottleneck process association relationship data. Based on the bottleneck process analysis data, the bottleneck process priority data, and the bottleneck process correlation data, business process optimization processing is performed to obtain a business process optimization strategy.

[0102] The above mainly describes the solutions of the embodiments of this application from the perspective of the method execution process. It is understood that, in order to achieve the above functions, the terminal includes the corresponding hardware structure and / or software modules for executing each function. Those skilled in the art should readily recognize that, in conjunction with the units and algorithm steps of the various examples described in the embodiments provided herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0103] This application embodiment can divide the terminal into functional units according to the above method example. For example, each function can be divided into a separate functional unit, or two or more functions can be integrated into one processing unit. The integrated unit can be implemented in hardware or as a software functional unit. It should be noted that the unit division in this application embodiment is illustrative and only represents one logical functional division. In actual implementation, there may be other division methods.

[0104] For those consistent with the above, please refer to Figure 4 , Figure 4 This application provides a schematic diagram of the structure of an intelligent optimization device for financial institution business processes. For example... Figure 4 As shown, the device includes: The acquisition unit 101 is used to acquire business process node data, business interaction support data and business result feedback data of the financial institution to be processed. The first processing unit 102 is used to perform dynamic bottleneck process mining processing based on the business process node data and the business interaction support data according to the dynamic condition response mechanism, and to obtain bottleneck process analysis data. The second processing unit 103 is used to perform bottleneck process impact quantification processing based on the bottleneck process analysis data and the business result feedback data to obtain bottleneck process priority data. The third processing unit 104 is used to perform bottleneck process association rule mining processing based on the bottleneck process priority data, the business process node data and the business interaction support data to obtain bottleneck process association relationship data. The fourth processing unit 105 is used to perform business process optimization processing based on the bottleneck process analysis data, the bottleneck process priority data, and the bottleneck process correlation data to obtain a business process optimization strategy.

[0105] In one possible implementation, the first processing unit 102 is used to perform dynamic bottleneck process mining based on the business process node data and the business interaction support data according to the dynamic condition response mechanism to obtain bottleneck process analysis data, specifically for: Based on the dynamic condition response mechanism, the business process node data is processed by event mapping to obtain dynamic business event data. Based on the dynamic condition response mechanism, the business interaction support data is processed by condition mapping to obtain dynamic business condition data and dynamic business response data. A dynamic flowchart is constructed based on the dynamic business event data, the dynamic business condition data, and the dynamic business response data. Based on the dynamic flowchart and the preset standard flowchart, difference identification processing is performed to generate difference flow data; Bottleneck node identification is performed based on the differential process data to obtain bottleneck process analysis data.

[0106] In one possible implementation, the second processing unit 103 is configured to perform bottleneck process impact quantification processing based on the bottleneck process analysis data and the business result feedback data to obtain bottleneck process priority data, specifically for: Based on the bottleneck process analysis data and the business result feedback data, efficiency impact indicators are extracted and processed to obtain efficiency impact feature data. Based on the bottleneck process analysis data and the business result feedback data, quality impact indicators are extracted and processed to obtain quality impact characteristic data. Based on the bottleneck process analysis data and the business result feedback data, experience impact indicators are extracted and processed to obtain experience impact feature data. The efficiency impact feature data, the quality impact feature data, and the experience impact feature data are fused together to obtain three-dimensional bottleneck feature data. Temporal feature enhancement processing is performed on the three-dimensional bottleneck feature data to obtain temporally enhanced bottleneck feature data. Based on the bottleneck process analysis data and the time-series enhanced bottleneck feature data, bottleneck priorities are divided to obtain reference bottleneck process priority data. Based on preset constraints, the reference bottleneck process priority data is calibrated to obtain bottleneck process priority data.

[0107] In one possible implementation, the second processing unit 103 is configured to perform bottleneck priority division based on the bottleneck process analysis data and the timing-enhanced bottleneck feature data to obtain reference bottleneck process priority data, specifically for: Based on preset base weights and current business objectives, the time-series enhanced bottleneck feature data is dynamically weighted to obtain dynamically weighted bottleneck feature data. Based on the dynamically weighted bottleneck feature data, cluster center determination is performed to obtain K financial business cluster centers; Based on the bottleneck nodes in the bottleneck process analysis data, Euclidean distance calculation is performed according to the dynamic weighted bottleneck feature data and the K financial business cluster centers to obtain the business Euclidean distance set. The confidence level is calculated for each service Euclidean distance in the service Euclidean distance set to obtain the service confidence level set. Bottleneck priority is determined based on the set of Euclidean distances and the set of confidence levels of the business processes, resulting in reference bottleneck process priority data.

[0108] In one possible implementation, the third processing unit 104 is used to perform bottleneck process association rule mining processing based on the bottleneck process priority data, the business process node data, and the business interaction support data to obtain bottleneck process association relationship data, specifically for: Based on the bottleneck process priority data, bottleneck node filtering is performed to obtain target bottleneck node data. Extract supporting data associated with the target bottleneck node data from the business interaction supporting data to obtain the target bottleneck node supporting data set; Discretize the support data of each target bottleneck node in the target bottleneck node support data set to obtain a discrete associated data set. Based on each discrete association data in the discrete association data set, association rule mining is performed to obtain an initial association rule set; Based on the business process node data and the initial set of association rules, rule filtering and verification are performed to obtain bottleneck process association data.

[0109] This application also provides a computer storage medium storing a computer program for electronic data interchange, which causes a computer to perform some or all of the steps of any of the intelligent optimization methods for financial institution business processes described in the above method embodiments.

[0110] This application also provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program that causes a computer to perform some or all of the steps of any of the intelligent optimization methods for financial institution business processes described in the above method embodiments.

[0111] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0112] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0113] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical or other forms.

[0114] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0115] Furthermore, the functional units in the various embodiments of the application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software program module.

[0116] If the integrated unit is implemented as a software program module and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0117] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage device, which may include: a flash drive, a read-only memory, a random access memory, a magnetic disk, or an optical disk, etc.

[0118] The embodiments of this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A smart optimization device for business processes in financial institutions, characterized in that, The device includes: The acquisition unit is used to acquire business process node data, business interaction support data, and business result feedback data of the financial institution to be processed. The first processing unit is used to perform dynamic bottleneck process mining based on the business process node data and the business interaction support data, according to the dynamic condition response mechanism, to obtain bottleneck process analysis data. The second processing unit is used to perform bottleneck process impact quantification processing based on the bottleneck process analysis data and the business result feedback data to obtain bottleneck process priority data. The third processing unit is used to perform bottleneck process association rule mining processing based on the bottleneck process priority data, the business process node data and the business interaction support data to obtain bottleneck process association relationship data. The fourth processing unit is used to perform business process optimization processing based on the bottleneck process analysis data, the bottleneck process priority data, and the bottleneck process correlation data to obtain a business process optimization strategy.

2. The intelligent optimization device for financial institution business processes according to claim 1, characterized in that, The first processing unit is used to perform dynamic bottleneck process mining based on the business process node data and the business interaction support data according to the dynamic condition response mechanism, to obtain bottleneck process analysis data, specifically for: Based on the dynamic condition response mechanism, the business process node data is processed by event mapping to obtain dynamic business event data. Based on the dynamic condition response mechanism, the business interaction support data is processed by condition mapping to obtain dynamic business condition data and dynamic business response data. A dynamic flowchart is constructed based on the dynamic business event data, the dynamic business condition data, and the dynamic business response data. Based on the dynamic flowchart and the preset standard flowchart, difference identification processing is performed to generate difference flow data; Bottleneck node identification is performed based on the differential process data to obtain bottleneck process analysis data.

3. The intelligent optimization device for financial institution business processes according to claim 2, characterized in that, The second processing unit is used to perform bottleneck process impact quantification processing based on the bottleneck process analysis data and the business result feedback data to obtain bottleneck process priority data, specifically for: Based on the bottleneck process analysis data and the business result feedback data, efficiency impact indicators are extracted and processed to obtain efficiency impact feature data. Based on the bottleneck process analysis data and the business result feedback data, quality impact indicators are extracted and processed to obtain quality impact characteristic data. Based on the bottleneck process analysis data and the business result feedback data, experience impact indicators are extracted and processed to obtain experience impact feature data. The efficiency impact feature data, the quality impact feature data, and the experience impact feature data are fused together to obtain three-dimensional bottleneck feature data. Temporal feature enhancement processing is performed on the three-dimensional bottleneck feature data to obtain temporally enhanced bottleneck feature data. Based on the bottleneck process analysis data and the time-series enhanced bottleneck feature data, bottleneck priorities are divided to obtain reference bottleneck process priority data. Based on preset constraints, the reference bottleneck process priority data is calibrated to obtain bottleneck process priority data.

4. The intelligent optimization device for financial institution business processes according to claim 2, characterized in that, The second processing unit is used to perform bottleneck priority division based on the bottleneck process analysis data and the timing enhancement bottleneck feature data to obtain reference bottleneck process priority data, specifically for: Based on preset base weights and current business objectives, the time-series enhanced bottleneck feature data is dynamically weighted to obtain dynamically weighted bottleneck feature data. Based on the dynamically weighted bottleneck feature data, cluster center determination is performed to obtain K financial business cluster centers; Based on the bottleneck nodes in the bottleneck process analysis data, Euclidean distance calculation is performed according to the dynamic weighted bottleneck feature data and the K financial business cluster centers to obtain the business Euclidean distance set. The confidence level is calculated for each service Euclidean distance in the service Euclidean distance set to obtain the service confidence level set. Bottleneck priority is determined based on the set of Euclidean distances and the set of confidence levels of the business processes, resulting in reference bottleneck process priority data.

5. The intelligent optimization device for financial institution business processes according to any one of claims 1-4, characterized in that, The third processing unit is used to perform bottleneck process association rule mining processing based on the bottleneck process priority data, the business process node data, and the business interaction support data to obtain bottleneck process association relationship data, specifically for: Based on the bottleneck process priority data, bottleneck node filtering is performed to obtain target bottleneck node data. Extract the supporting data associated with the target bottleneck node data from the business interaction supporting data to obtain the target bottleneck node supporting data set; Discretize the support data of each target bottleneck node in the target bottleneck node support data set to obtain a discrete associated data set. Based on each discrete association data in the discrete association data set, association rule mining is performed to obtain an initial association rule set; Based on the business process node data and the initial set of association rules, rule filtering and verification are performed to obtain bottleneck process association data.

6. A method for intelligent optimization of business processes in financial institutions, characterized in that, The intelligent optimization method for financial institution business processes includes: Acquire business process node data, business interaction support data, and business result feedback data of the financial institutions to be processed; Based on the dynamic condition response mechanism, dynamic bottleneck process mining is performed according to the business process node data and the business interaction support data to obtain bottleneck process analysis data. Based on the bottleneck process analysis data and the business result feedback data, the impact of the bottleneck process is quantified to obtain the bottleneck process priority data. Based on the bottleneck process priority data, the business process node data, and the business interaction support data, bottleneck process association rule mining is performed to obtain bottleneck process association relationship data. Based on the bottleneck process analysis data, the bottleneck process priority data, and the bottleneck process correlation data, business process optimization processing is performed to obtain a business process optimization strategy.

7. The intelligent optimization method for financial institution business processes according to claim 6, characterized in that, The dynamic condition response mechanism, based on the business process node data and the business interaction support data, performs dynamic bottleneck process mining to obtain bottleneck process analysis data, including: Based on the dynamic condition response mechanism, the business process node data is processed by event mapping to obtain dynamic business event data. Based on the dynamic condition response mechanism, the business interaction support data is processed by condition mapping to obtain dynamic business condition data and dynamic business response data. A dynamic flowchart is constructed based on the dynamic business event data, the dynamic business condition data, and the dynamic business response data. Based on the dynamic flowchart and the preset standard flowchart, difference identification processing is performed to generate difference flow data; Bottleneck node identification is performed based on the differential process data to obtain bottleneck process analysis data.

8. The intelligent optimization method for financial institution business processes according to claim 7, characterized in that, The step of quantifying the impact of the bottleneck process based on the bottleneck process analysis data and the business result feedback data to obtain bottleneck process priority data includes: Based on the bottleneck process analysis data and the business result feedback data, efficiency impact indicators are extracted and processed to obtain efficiency impact feature data. Based on the bottleneck process analysis data and the business result feedback data, quality impact indicators are extracted and processed to obtain quality impact characteristic data. Based on the bottleneck process analysis data and the business result feedback data, experience impact indicators are extracted and processed to obtain experience impact feature data. The efficiency impact feature data, the quality impact feature data, and the experience impact feature data are fused together to obtain three-dimensional bottleneck feature data. Temporal feature enhancement processing is performed on the three-dimensional bottleneck feature data to obtain temporally enhanced bottleneck feature data. Based on the bottleneck process analysis data and the time-series enhanced bottleneck feature data, bottleneck priorities are divided to obtain reference bottleneck process priority data. Based on preset constraints, the reference bottleneck process priority data is calibrated to obtain bottleneck process priority data.

9. A terminal, characterized in that, The system includes a processor, an input device, an output device, and a memory, which are interconnected. The memory stores a computer program, which includes program instructions. The processor is configured to invoke the program instructions to execute the intelligent optimization method for financial institution business processes as described in any one of claims 6-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, the computer program including program instructions, which, when executed by a processor, cause the processor to perform the intelligent optimization method for financial institution business processes as described in any one of claims 6-8.

Images