Intelligent matching system for authentication review and reviewer ability portrait based on big data

Abstract

This invention provides an intelligent matching system for certification audits and auditor competency profiles based on big data. It relates to the field of intelligent matching technology for auditor competency profiles, and includes: a big data resource pool and auditor competency profile construction module, an intelligent matching and dynamic scheduling engine module, and a scheduling execution and closed-loop management module. This invention enables multi-source data governance and dynamic profile construction, project requirement analysis and precise matching, re-matching compliance verification and full-process evidence storage, closed-loop management of scheduling execution, and iterative optimization of the profile. It effectively solves the problems of inefficient matching, insufficient compliance, and scheduling lag in existing technologies, improves the utilization rate of audit resources and matching accuracy, and ensures the compliance and traceability of certification audits.

Description

Technical Field

[0001] This invention relates to the field of intelligent matching technology for auditor competency profiles, and particularly to an intelligent matching system for certification audits and auditor competency profiles based on big data. Background Technology

[0002] In dynamic matching scenarios for audit resource scheduling, existing technologies lack effective real-time data collection and event-driven mechanisms at the dynamic event perception level. Regarding changes in the status of auditors, existing systems only support manual reporting of "available / unavailable," and cannot automatically collect key dynamic information such as auditors' real-time travel trajectories, changes in qualification status, and task overload. Subsequent processing can only proceed after auditors actively report or managers manually discover the issue, resulting in scheduling response delays and failing to meet the business needs for rapid scheduling. For sudden changes on the audit task side, such as companies applying to adjust audit times, increase or decrease audit scope, or the inability to conduct audits on-site, existing technologies still rely on offline communication, email, or instant messaging tools to transmit information, lacking automatic collection and event triggering mechanisms. This makes it impossible to capture task-side dynamics in real time, further exacerbating the lag in scheduling instructions. At the dynamic rematching algorithm level, existing technologies only support simple static matching strategies and lack dynamic rematching and intelligent scheduling capabilities. When dynamic events such as individual auditor abnormalities occur, existing systems can only perform manual replacement operations and cannot perform multi-task joint optimization, itinerary merging, or time slice rescheduling based on the global audit resource pool, resulting in low overall resource utilization. In addition, existing matching algorithms usually use fixed static rules and cannot perceive real-time changes in constraints (such as urgency, travel costs, customer level, and risk level) in dynamic scenarios, nor can they adjust constraint weights and objective functions in real time according to dynamic events, resulting in a disconnect between matching results and actual needs. For urgent order insertion audit scenarios, such as overdue supervisory audits, complaint verification, or special regulatory inspections, existing systems lack priority-based preemptive scheduling or incremental insertion capabilities and can only handle them by manually and crudely squeezing out existing tasks, causing a large number of scheduling conflicts and resource waste. At the spatiotemporal joint scheduling level, the existing scheduling model has a single dimension and lacks spatiotemporal joint scheduling capabilities. The auditor resources of each branch or regional office are independent of each other, forming regional resource silos. When local audit resources are insufficient, the system cannot automatically schedule external cross-regional resources and can only rely on manual coordination by management personnel, resulting in limited scheduling scope and extremely low efficiency. Regarding the compliance aspect of dynamic scheduling, existing technologies pose a risk of losing control over compliance during real-time adjustments. When performing emergency replacements or dynamic re-matching, existing technologies only check the immediate availability of auditors and do not re-trigger and execute the verification process for legally mandated compliance requirements such as qualification scope, qualification validity, conflict of interest avoidance, and personnel levels for high-risk projects. This can easily lead to situations where auditors are improperly assigned after dynamic replacements. When an anomaly occurs in a certain link (such as flight cancellation or on-site delay), the anomaly information cannot be automatically transmitted to subsequent related tasks and trigger linkage adjustments. There is a problem of low reliability in the intelligent matching of certification audits and auditor capability profiles due to slow dynamic scheduling response and low efficiency. Summary of the Invention

[0003] In view of this, the embodiments of the present invention provide an intelligent matching system for certification audit and auditor capability profile based on big data. It can realize multi-source data governance and dynamic profile construction, project requirement analysis and accurate matching, re-matching compliance verification and full-process evidence storage, closed-loop management of scheduling and execution, and iterative optimization of profile. It effectively solves the problems of inefficient matching, insufficient compliance, and scheduling lag in existing technologies, improves the utilization rate of audit resources and matching accuracy, and ensures the compliance and traceability of certification audit.

[0004] This invention provides an intelligent matching system for certification auditing and auditor competency profiling based on big data. The system includes: a big data resource pool and auditor competency profiling construction module, an intelligent matching and dynamic scheduling engine module, and a scheduling execution and closed-loop management module. The big data resource pool and auditor competency profiling construction module is used for the collection, cleaning, fusion, and standardized storage of multi-source data. Based on the standardized multi-source data, it constructs a dynamically iterative digital competency profile of auditors. The intelligent matching and dynamic scheduling engine module is used to structurally decompose the target certification project, generate a standardized requirement feature vector for the project, and then... The system performs similarity calculations and multi-objective constraint optimization matching with the auditor's competency profile to generate a compliance matching list, and performs real-time re-matching and conflict resolution for dynamic and sudden scheduling events; the compliance verification and process evidence preservation module is used to perform full-process legal compliance verification for real-time re-matching, perform credible evidence preservation and traceability of the entire matching scheduling process, and generate and execute matching scheduling instructions; the scheduling execution and closed-loop management module is used to push matching scheduling instructions to the target terminal in real time, collect auditor execution feedback data and on-site status data to form a full closed-loop management of scheduling-execution-feedback, and feed back execution data to the big data resource pool module to achieve iterative optimization of the profile.

[0005] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: Through the synergistic interaction of four core modules, this system effectively addresses technical pain points in existing certification audit resource scheduling and dynamic matching scenarios, such as lagging dynamic event perception, insufficient intelligent matching accuracy, lack of compliance control, and broken execution loops. It achieves precise, intelligent, compliant, and efficient scheduling of certification audit resources. By standardizing the governance of multi-source data and constructing dynamically iterative profiles, it breaks down data silos, generating digital capability profiles of auditors containing five core tags. These profiles accurately represent auditors' actual qualifications, professional abilities, spatiotemporal adaptability, and dynamic status, completely resolving the static and distorted nature of traditional profiles and providing high-quality data support for subsequent intelligent matching. Through structured decomposition of project requirements, similarity calculation, and multi-objective constraint optimization matching, combined with dynamic adjustment of audit type matching weights and similarity weight correction driven by matching failure rates, it significantly improves the matching accuracy between auditors and certification projects, effectively reducing matching failures. This system improves efficiency and addresses the pain points of traditional scheduling, such as delayed response, low resource utilization, and frequent conflicts, by enabling real-time re-matching and conflict resolution for dynamic and sudden scheduling events. Through full-process legal compliance verification, it strictly controls compliance requirements such as qualification validity, scope of practice, and avoidance relationships, eliminating unauthorized delegation. Furthermore, through trusted evidence storage, traceability, and dynamic adjustment of on-chain latency thresholds, it ensures traceability and data immutability throughout the matching and scheduling process, meeting regulatory verification requirements and reducing compliance risks for certification bodies. Encrypted communication enables real-time push of scheduling instructions and accurate collection of multi-dimensional execution data, constructing a closed-loop management system of scheduling-execution-feedback. This effectively avoids issues such as missed instruction execution, progress delays, and untimely handling of anomalies. The feedback of execution data enables continuous iterative optimization of auditor capability profiles, forming a virtuous cycle of "data-profile-matching-execution-iteration," further improving the accuracy of subsequent matching and scheduling efficiency. Attached Figure Description

[0006] Figure 1 This is a schematic diagram of the structure of the intelligent matching system for certification audit and auditor capability profile based on big data provided in an embodiment of the present invention; Figure 2 This is a flowchart of the process for constructing a digital competency profile of auditors in an intelligent matching system for certification auditing and auditor competency profiles based on big data, provided in an embodiment of the present invention. Figure 3 This is a flowchart of the full-process legal compliance verification process of the intelligent matching system for certification audit and auditor competence profiling based on big data, provided in this embodiment of the invention. Detailed Implementation

[0007] This invention provides an intelligent matching system for certification auditing and auditor competency profiling based on big data. For example... Figure 1The diagram shows the structure of an intelligent matching system for certification auditing and auditor competency profiling based on big data. The system includes: a big data resource pool and auditor competency profiling construction module, an intelligent matching and dynamic scheduling engine module, and a scheduling execution and closed-loop management module.

[0008] As the first module of this system, the Big Data Resource Pool and Auditor Capability Profile Construction Module is used for the collection, cleaning, fusion and standardized storage of multi-source data. Based on the standardized multi-source data, a dynamically iterative digital capability profile of auditors is constructed.

[0009] It needs to be explained that the specific steps for building a dynamically iterative digital capability profile for auditors are as follows: Multi-source data includes audit performance data, real-time status data, certification project basic data, regulatory compliance data, and spatiotemporal travel data. The digital capability profile of auditors includes, but is not limited to, legal qualification tags, professional capability tags, performance compliance tags, spatiotemporal resource tags, and dynamic status tags. By accessing a pre-processed standardized dataset from a big data resource pool and based on a pre-defined tagging system, feature extraction and classification labeling are performed on multi-source data to generate legal qualification tags, professional competence tags, performance compliance tags, spatiotemporal resource tags, and dynamic status tags. These tags are then linked and integrated to construct an initial digital profile of the auditor's capabilities. Specifically, the legal qualification tag corresponds to regulatory compliance information such as the auditor's qualification validity period, registration level, and scope of practice; the professional competence tag corresponds to professional information such as the auditor's familiar certification standards, industry sub-sectors, and technical background; the performance compliance tag corresponds to performance information such as the auditor's audit report pass rate, violation records, and customer feedback; the spatiotemporal resource tag corresponds to spatiotemporal information such as the auditor's permanent location, transportation adaptability, and travel range; and the dynamic status tag corresponds to dynamic information such as the auditor's real-time availability, workload, and travel arrangements.

[0010] In this embodiment, firstly, multi-source data such as audit performance data, real-time status data, certification project basic data, regulatory compliance data, and spatiotemporal travel data are collected and integrated. After cleaning, fusion, and standardization, a high-quality standardized dataset is formed, effectively breaking down data silos and ensuring data availability. Subsequently, based on a preset tagging system, feature extraction and classification labeling are performed on the standardized data to generate legal qualification tags corresponding to regulatory compliance information, professional competence tags corresponding to professional competence, performance compliance tags corresponding to professional practice quality, spatiotemporal resource tags corresponding to geographical adaptability, and dynamic status tags reflecting real-time work status, thereby achieving a multi-dimensional and accurate characterization of auditors' capabilities. Then, the above five types of tags are associated and integrated to construct an initial digital capability profile of the auditors, enabling a comprehensive and quantitative presentation of the auditors' qualifications, capabilities, performance, spatiotemporal resources, and dynamic status. This provides a real, complete, and quantifiable data foundation for subsequent intelligent matching and scheduling, significantly improving the reliability and accuracy of the matching criteria.

[0011] It should be understood that Figure 2 This invention provides a flowchart of the process for constructing a digital competency profile for auditors using an intelligent matching system for certification auditing and auditor competency profiles based on big data. The specific process is as follows: First, a new round of multi-source data is collected by initiating a real-time data interaction mechanism, and it is determined whether incremental data exists. If so, preprocessing is performed to generate incremental standardized data, which is used to update and correct the corresponding tags of the auditor competency profile, achieving real-time dynamic updates of the profile. Then, profile quality verification is conducted by combining historical data and execution feedback data. If tag consistency is not up to standard, biased tags are removed and erroneous tags are corrected. Next, it is determined whether there are changes in industry standards, regulatory requirements, or business scenarios. If so, the tag system and feature extraction rules are optimized; if not, the profile is directly retained. Finally, the competency profile is iteratively optimized to ensure a high degree of matching between the profile and the auditor's actual capabilities, business needs, and regulatory requirements.

[0012] It should be further explained that building a dynamically iterative digital capability profile for auditors also includes: Establish a real-time data interaction mechanism. When the big data resource pool collects a new round of multi-source data, including but not limited to auditors' new audit performance, qualification status changes, real-time schedule adjustments, task load changes, and regulatory compliance information updates, the data preprocessing process is automatically triggered to generate incremental standardized data. This updates, supplements, and corrects the corresponding tags of the auditor's initial digital capability profile, enabling real-time dynamic updates of the auditor's digital capability profile. The quality of the constructed digital capability profiles of auditors is verified. By combining historical data retained in the big data resource pool and execution feedback data returned from the scheduling and closed-loop management modules, the consistency between the digital capability profile labels of auditors and the actual profile labels is compared, so as to eliminate deviation labels and correct erroneous labels. Based on updates to certification industry standards, adjustments to regulatory requirements, and changes in business scenarios, we optimize the labeling system and feature extraction rules to achieve continuous iterative optimization of auditors' digital capability profiles, ensuring that the profiles are highly aligned with auditors' actual capabilities, business needs, and regulatory requirements.

[0013] In this embodiment, by establishing a real-time data interaction mechanism, when a new round of multi-source data is collected from the big data resource pool, such as new audit performance data, changes in qualification status, real-time schedule adjustments, changes in task load, and updates to regulatory compliance information, the data preprocessing process is automatically triggered to generate incremental standardized data. This allows for real-time updates, supplementation, and correction of the corresponding tags in the initial digital capability profile of the auditors, ensuring that the profile synchronously reflects the latest status of the auditors and avoiding matching deviations caused by the disconnect between static profiles and actual conditions. Simultaneously, the quality of the constructed profile is verified by combining historical data and execution feedback data from scheduling execution. By comparing tag consistency, deviations are eliminated and errors are corrected, effectively ensuring the authenticity and accuracy of the profile data. Furthermore, the tag system and feature extraction rules are continuously optimized based on updates to certification industry standards, adjustments to regulatory requirements, and changes in business scenarios, achieving full lifecycle iterative optimization of the profile. This ensures that the profile is always highly matched with the auditors' actual capabilities, business needs, and regulatory requirements, providing continuous, reliable, and accurate data support for subsequent intelligent matching and dynamic scheduling.

[0014] As the second module of this system, the intelligent matching and dynamic scheduling engine module is used to structurally decompose the audit type, industry field, standard scope, risk level, time constraints, geographical constraints, and urgency of the target certification project, generate a standardized requirement feature vector for the project, perform similarity calculation and multi-objective constraint optimization matching between the standardized requirement feature vector of the project and the auditor's capability profile, generate a compliance matching list, and perform real-time re-matching and conflict resolution for dynamic and sudden scheduling events.

[0015] Furthermore, the specific steps for similarity calculation and multi-objective constraint optimization matching are as follows: Step 1: Obtain the similarity weight and the review type matching weight. The similarity weight is used to adjust the proportion of each dimension's matching degree in the overall similarity calculation, and the review type matching weight is used to represent the weight proportion of the review type matching degree in the comprehensive similarity calculation. Step 2: When the audit type of the target certification project is a special audit, the audit type matching weight adjustment mechanism is triggered; when the audit type of the target certification project is a regular surveillance audit, the audit type matching weight adjustment mechanism is triggered. Step 3: Determine whether the matching failure rate is within the preset failure rate threshold range. If so, maintain the similarity baseline weight unchanged. If not, determine whether the matching failure rate is greater than the preset failure rate threshold upper limit. A mapping relationship between audit and audit type matching weight adjustment factors is constructed. This mapping relationship is used to represent the pre-established one-to-one correspondence between the audit type of the target certification project and the corresponding weight adjustment factor. The mapping of special audit type outputs the matching weight increase factor, and the mapping of regular supervision audit type outputs the matching weight decrease factor, so that the intelligent matching and dynamic scheduling engine module can directly query the corresponding weight adjustment factor based on the input audit type.

[0016] The specific steps to trigger the review type matching weight adjustment mechanism are as follows: If the review type matching weight adjustment mechanism is triggered, the corresponding review type is input into the review-review type matching weight adjustment factor mapping relationship, the corresponding matching weight adjustment factor is output, and the result of doubling the review type matching baseline weight and the matching weight adjustment factor and then rounding down is used as the target review type matching weight. The specific steps to trigger the review type matching weight reduction mechanism are as follows: If the review type matching weight reduction mechanism is triggered, the corresponding review type will be input into the review-review type matching weight adjustment factor mapping relationship, and the corresponding matching weight reduction factor will be output. The result of multiplying the review type matching baseline weight and the matching weight reduction factor and then rounding up will be used as the target review type matching weight.

[0017] In this embodiment, step one is first executed, using the analytic hierarchy process (AHP) to construct a similarity weight judgment matrix. Seven matching dimensions—review type, industry sector, standard scope, risk level, time constraint, geographical constraint, and urgency—are used as evaluation indicators for the judgment matrix. The relative importance of each dimension is determined using a pairwise comparison method. The eigenvectors of the judgment matrix are calculated and consistency checks are performed (consistency check index CI < 0.1 and consistency ratio CR < 0.1 are considered passing). After passing the check, the eigenvectors are normalized to obtain the similarity weights corresponding to each dimension (used to adjust the proportion of each dimension's matching degree in the overall similarity calculation). Simultaneously, a baseline weight for review type matching is preset, clarifying that the review type matching weight is used to characterize the weight proportion of review type matching degree in the comprehensive similarity calculation. Then, step two is executed, pre-constructing the review- The audit type matching weight adjustment factor mapping relationship is used to characterize the pre-established one-to-one correspondence between the audit type of the target certification project and the corresponding weight adjustment factor. Specifically, the special audit type mapping outputs a matching weight increase factor (value range 1.2-1.5), and the regular supervision audit type mapping outputs a matching weight decrease factor (value range 0.7-0).9) This allows the intelligent matching and dynamic scheduling engine module to directly query the corresponding weight adjustment factor based on the input audit type. When the audit type of the target certification project is a special audit, the audit type matching weight increase mechanism is triggered; when the audit type of the target certification project is a regular supervision audit, the audit type matching weight decrease mechanism is triggered. The specific steps for triggering the audit type matching weight increase mechanism are as follows: If the increase mechanism is triggered, the corresponding special audit type is input into the above mapping relationship, the corresponding matching weight increase factor is queried and output, and the rounding down algorithm is used to multiply the audit type matching baseline weight and the matching weight increase factor, then rounded down to obtain the target audit type matching weight, ensuring the accuracy and rationality of the weight adjustment. The specific steps for triggering the audit type matching weight decrease mechanism are as follows: If the decrease mechanism is triggered, the corresponding regular supervision audit type is input into the... In the mapping relationship, the query output corresponding matching weight adjustment factor is also rounded up using the rounding algorithm. The matching baseline weight of the review type and the matching weight adjustment factor are multiplied and then rounded up to obtain the target review type matching weight, avoiding decimal deviations in weight that could affect matching accuracy. Finally, step three is executed: a preset matching failure rate critical interval is established (this interval is a closed interval formed by the preset lower limit and the preset upper limit of the failure rate). Recent matching data is collected in real time and the matching failure rate is calculated. The calculated matching failure rate is compared with the preset failure rate critical interval. If the matching failure rate is within the critical interval, the similarity baseline weight determined by the analytic hierarchy process remains unchanged to ensure the stability of the weight. If the matching failure rate is not within the critical interval, it is further determined whether the matching failure rate is greater than the preset upper limit of the failure rate, laying the foundation for subsequent dynamic correction of the similarity weight. The entire process utilizes the analytic hierarchy process (AHP) to scientifically assign weights. Combined with a precise weight adjustment algorithm and matching failure rate judgment logic, it effectively solves the problems of subjective and coarse-grained traditional weight assignment, significantly improving the accuracy of comprehensive similarity calculation. This provides scientific and reliable weight support for efficient matching between auditors and certification projects. Furthermore, strict rounding rules and consistency checks ensure the standardization and rationality of weight adjustments, further reducing the matching failure rate.

[0018] It should be further explained that the specific steps for determining whether the matching failure rate is within the preset failure rate threshold range are as follows: A failure rate-similarity weight correction factor mapping table is pre-constructed to establish a one-to-one correspondence between the matching failure offset, the matching failure deviation and the corresponding similarity weight correction factor, providing a clear basis for factor query for the dynamic correction of the similarity benchmark weight. The target similarity weight obtained after the doubling process meets the normalization constraint requirements of the similarity weight. If the matching failure rate is within the preset failure rate critical interval, the similarity benchmark weight remains unchanged. The preset failure rate critical interval represents the closed interval formed by the preset lower limit and the preset upper limit of the failure rate.

[0019] Determining whether the matching failure rate is within the preset failure rate threshold also includes: If the matching failure rate is greater than the preset failure rate threshold, the matching failure offset is input into the matching failure rate-similarity weight correction factor mapping table, and the similarity weight gain factor is output. The current similarity baseline weight and the similarity weight gain factor are multiplied to obtain the target similarity weight. The matching failure offset represents the positive difference between the matching failure rate and the preset failure rate threshold. If the matching failure rate is less than the preset failure rate threshold, the matching failure deviation is input into the matching failure rate-similarity weight correction factor mapping table, and the similarity weight reduction factor is output. The current similarity benchmark weight and the similarity weight reduction factor are multiplied to obtain the target similarity weight. The matching failure deviation represents the negative difference between the matching failure rate and the preset failure rate threshold.

[0020] In this embodiment, the dynamic matching failure rate is first calculated in real time based on the sliding window statistical algorithm. The sliding time window is formed by the N most recent valid matching scheduling results. After removing abnormal samples and interference data, a smoothed target matching failure rate is obtained. Simultaneously, a failure rate-similarity weight correction factor mapping table is pre-constructed. This mapping table establishes a one-to-one quantitative relationship between the matching failure offset, matching failure deviation, and the corresponding similarity weight correction factor, providing a clear factor query basis for the dynamic correction of the similarity benchmark weight. Furthermore, the target similarity weight obtained after the doubling process must satisfy the L2 normalization constraint algorithm verification, ensuring that the sum of all dimension weights is always 1 and meets the normalization constraint requirements. Then, the target matching failure rate output by the sliding window algorithm is compared with a closed interval formed by a preset lower and upper limit of the failure rate threshold. If the target matching failure rate is within this preset failure rate threshold interval, the normalized and calibrated similarity benchmark weight remains unchanged to ensure the stability and continuity of the matching calculation. If the target matching failure rate is greater than the preset upper limit of the failure rate threshold, the matching failure offset representing the positive difference between the two is calculated and input into the failure rate-similarity weight correction factor mapping table. The similarity weight correction factor mapping table is used to query and output the corresponding similarity weight gain factor. The current similarity baseline weight is multiplied by this gain factor, and then normalized and corrected using the L2 normalization constraint algorithm to obtain the target similarity weight that meets the constraint conditions. If the target matching failure rate is less than the preset failure rate threshold, the matching failure deviation representing the negative difference between the two is calculated. This deviation is input into the mapping table to query and output the corresponding similarity weight reduction factor. The current similarity baseline weight is multiplied by this reduction factor, and then normalized and calibrated using the L2 normalization constraint algorithm to obtain a compliant and accurate target similarity weight. By introducing sliding window statistics and normalization constraint algorithms in synergy, the robustness of failure rate calculation is improved, and the consistency and effectiveness of the system calculation after weight adjustment are strictly guaranteed.

[0021] As the third module of this system, the compliance verification and process evidence preservation module is used to perform full-process legal compliance verification of real-time rematching, including verification of qualification validity, scope of practice, avoidance relationship, and personnel level of high-risk projects. It also performs credible evidence preservation and traceability of the entire matching and scheduling process, and generates and executes matching and scheduling instructions.

[0022] It needs to be explained that, Figure 3This is a flowchart of the full-process legal compliance verification process of the intelligent matching system for certification audit and auditor capability profiling based on big data, provided in this embodiment of the invention. The specific process is as follows: First, the actual instruction receipt delay is obtained, and a receipt delay range is preset. Then, the scenario is distinguished by two condition judgments: If the actual receipt delay is less than the lower limit, it is determined that the instruction is immediately received, and the initial on-chain delay threshold remains unchanged; If the actual receipt delay is within the range (not less than the lower limit and not greater than the upper limit), it is determined to be a first-level delayed receipt. The actual receipt delay is input into the mapping relationship to obtain the first-level gain coefficient, which is multiplied by the initial threshold to obtain the target on-chain delay threshold; If the actual receipt delay is greater than the upper limit, it is determined to be a second-level delayed receipt. First, the receipt delay offset (the positive difference between the actual delay and the upper limit) is calculated, and then the offset is input into the mapping relationship to obtain the second-level gain coefficient, which is multiplied by the initial threshold to obtain the target on-chain delay threshold. Finally, the dynamic adjustment of the on-chain delay threshold is completed.

[0023] It should be understood that the specific process for performing full-process legal compliance verification on real-time re-matching is as follows: The initial on-chain latency threshold is preset. The initial on-chain latency threshold is the baseline time interval between the generation time of the scheduling instruction and the on-chain operation of the full-process evidence storage data. Obtain the scheduling instruction generation time parameter and the instruction receipt feedback time parameter, and calculate the actual instruction receipt delay. The actual instruction receipt delay is the time difference between the instruction receipt feedback time and the scheduling instruction generation time. The actual instruction receipt delay is compared with the preset receipt delay range, and the threshold for the on-chain delay of the evidence storage data throughout the entire process is dynamically adjusted based on the comparison results. A pre-built mapping relationship between instruction receipt delay and on-chain delay threshold is established. This mapping relationship is used to characterize the pre-established segmented quantitative correspondence between the actual instruction receipt delay, the receipt delay offset, and the corresponding on-chain delay threshold gain coefficient. This allows the compliance verification and process evidence preservation module to directly query the corresponding gain coefficient based on the input actual instruction receipt delay or receipt delay offset, and then dynamically adjust the initial on-chain delay threshold to obtain the target full-process evidence preservation data on-chain delay threshold that is adapted to the current instruction receipt status.

[0024] The full-process legal compliance verification for real-time rematching also includes: If the actual instruction receipt delay is within the preset receipt delay interval, it is determined to be an instruction delayed receipt. The current actual instruction receipt delay is input into the instruction receipt delay-on-chain delay threshold mapping relationship, and the first-level gain coefficient of the on-chain delay threshold is output. The preset initial on-chain delay threshold and the first-level gain coefficient of the on-chain delay threshold are multiplied to obtain the target full-process evidence storage data on-chain delay threshold. The preset receipt delay interval represents the closed interval formed by the preset lower limit of receipt delay and the preset upper limit of receipt delay. If the actual instruction receipt delay is less than the preset receipt delay lower limit, it is determined that the instruction is received immediately, and the initial on-chain delay threshold remains unchanged. If the actual instruction receipt delay is greater than the preset receipt delay upper limit, it is determined that the instruction receipt is delayed. The receipt delay offset is input into the instruction receipt delay-on-chain delay threshold mapping relationship, and the second-level gain coefficient of the on-chain delay threshold is output. The preset initial on-chain delay threshold and the second-level gain coefficient of the on-chain delay threshold are multiplied to obtain the target full-process evidence storage data on-chain delay threshold. The receipt delay offset represents the positive difference between the actual instruction receipt delay and the preset receipt delay upper limit.

[0025] In this embodiment, firstly, an initial on-chain latency threshold is preset. This initial on-chain latency threshold is the baseline time interval between the generation time of the scheduling instruction and the on-chain operation initiated by the full-process evidence storage data. Simultaneously, an exponential smoothing algorithm is used to preprocess and calibrate the initial on-chain latency threshold, introducing a smoothing coefficient α (range 0.1-0.3). Combining historical on-chain latency data with the initial threshold, the calibrated initial on-chain latency threshold is calculated using the formula S0 = α × initial on-chain latency threshold + (1-α) × historical average on-chain latency threshold. This effectively reduces the impact of historical on-chain fluctuations on the initial threshold and improves the rationality of the threshold setting. Subsequently, the compliance verification and process evidence storage module accurately obtains the scheduling instruction generation time parameter and instruction receipt feedback time parameter through a standardized interface, calculating the actual instruction receipt latency. This actual instruction receipt latency is the time difference between the instruction receipt feedback time and the scheduling instruction generation time. After calculation, an outlier removal algorithm is used to remove abnormal latency data caused by terminal failure and network latency, ensuring the accuracy of the actual instruction receipt latency. Next, an instruction receipt latency- The on-chain latency threshold mapping relationship is used to characterize the pre-established segmented quantization correspondence between the actual instruction receipt latency, the receipt latency offset, and the corresponding on-chain latency threshold gain coefficient. A linear interpolation algorithm is embedded in the mapping relationship. For different values ​​of the actual instruction receipt latency within the preset receipt latency interval and different positive deviations of the receipt latency offset, a precise gain coefficient is calculated through linear interpolation. This allows the compliance verification and process evidence preservation module to quickly query and output the corresponding gain coefficient based on the input actual instruction receipt latency or receipt latency offset, thereby dynamically adjusting the initial on-chain latency threshold to obtain the target full-process evidence preservation data on-chain latency threshold adapted to the current instruction receipt status. Afterwards, outliers are removed... The actual instruction receipt delay is compared with the preset receipt delay interval (which is a closed interval formed by the preset lower limit and the preset upper limit of the receipt delay). Based on the comparison result, the on-chain delay threshold of the evidence storage data is dynamically adjusted according to different scenarios. Specifically: if the actual instruction receipt delay is less than the preset lower limit of the receipt delay, it is determined that the instruction is immediately received, and the initial on-chain delay threshold calibrated by the exponential smoothing algorithm remains unchanged to ensure the timeliness of evidence storage on-chain; if the actual instruction receipt delay is within the preset receipt delay interval, it is determined that the instruction is delayed, and the current actual instruction receipt delay is input into the above mapping relationship. The corresponding on-chain delay threshold first-level gain coefficient (within the range of 1.1-1) is queried and output through the linear interpolation algorithm.3) The calibrated initial on-chain latency threshold is multiplied by the first-level gain coefficient to obtain the target full-process evidence storage data on-chain latency threshold, which is suitable for evidence storage requirements in mild delay signing scenarios; if the actual instruction signing latency is greater than the preset signing latency upper limit, it is determined to be a severe instruction signing latency. First, the signing latency offset is calculated (this offset represents the degree of positive deviation between the actual instruction signing latency and the preset signing latency upper limit), and this signing latency offset is input into the instruction signing latency - In the on-chain latency threshold mapping relationship, the corresponding on-chain latency threshold secondary gain coefficient (range 1.4-1.6) is queried and output through a linear interpolation algorithm. The larger the deviation, the larger the value of the secondary gain coefficient. Then, the calibrated initial on-chain latency threshold is multiplied by this secondary gain coefficient to obtain the target full-process evidence storage data on-chain latency threshold, which is adapted to the evidence storage requirements of severely delayed signing scenarios. Finally, the effectiveness of the adjusted target on-chain latency threshold is verified to ensure that it is within the preset reasonable threshold range, avoiding problems such as excessively long on-chain latency, untimely evidence storage, or excessively frequent on-chain storage and system resource consumption caused by excessively large or small gain coefficients. The entire process, through the coordinated introduction of exponential smoothing, linear interpolation, and outlier removal algorithms, not only solves the problems of static on-chain latency thresholds and low accuracy of gain coefficient queries in traditional methods, but also improves the accuracy and adaptability of latency adjustments. This ensures a high degree of matching between the timing of on-chain data storage and the instruction receipt status, guaranteeing both the timeliness and integrity of the stored data while avoiding waste of system resources. Furthermore, it strengthens the standardization and reliability of the entire process of matching and scheduling for evidence storage, meeting regulatory requirements for the timeliness and traceability of evidence storage.

[0026] As the fourth module of this system, the scheduling execution and closed-loop management module is used to push matching scheduling instructions to the target terminal in real time, collect the auditor's execution feedback data and on-site status data to form a full closed-loop management of scheduling-execution-feedback, and feed back the execution data to the big data resource pool module to realize the iterative optimization of the profile.

[0027] It should be noted that the matching and scheduling instructions include the target auditor's unique identifier, audit task details, time constraints, geographical constraints, and compliance requirements; Through a pre-set encrypted communication interface, the matching scheduling instructions are pushed in real time to the target terminal bound to the unique identifier of the target auditor. The target terminal is a dedicated work terminal for auditors and supports instruction receipt, status feedback and data reporting functions. Auditor execution feedback data includes the status of dispatch instruction receipt, audit task start status, audit task progress, audit task completion status, and audit exception information. On-site status data includes the auditor's real-time geographical location, on-site audit scene photos / videos, audit site environmental parameters, audit process records, and other information. During the collection process, each piece of data is timestamped and identified by its source to ensure data traceability. If a matching scheduling instruction is not acknowledged, a secondary push notification mechanism is triggered; if the audit task is found to be lagging behind, a progress warning is generated and pushed to the auditor's terminal and the scheduling management terminal; if an abnormality is detected at the audit site, real-time feedback is sent to the intelligent matching and dynamic scheduling engine module, triggering an abnormal rescheduling assessment; if the audit task is detected to be completed, the execution result is confirmed and synchronized to the compliance verification and process evidence storage module.

[0028] In this embodiment, firstly, a structured matching and scheduling instruction containing the target auditor's unique identifier, audit task details, time constraints, geographical constraints, and compliance requirements is generated. The instruction content is then encrypted end-to-end using the AES-256 encryption algorithm. The encrypted matching and scheduling instruction is pushed in real-time to a dedicated work terminal bound to the target auditor's unique identifier via a preset encrypted communication interface (based on HTTPS + SM4 national cryptographic protocol). This terminal has a built-in instruction decryption module, which can only be unlocked after auditor authentication. It supports instruction receipt, status feedback, and data reporting functions. Subsequently, based on a real-time data acquisition and verification algorithm, the terminal collects auditor execution feedback data (including scheduling instruction receipt status, audit task start status, audit task progress, audit task completion status, and audit anomaly descriptions) and on-site status data (including the auditor's real-time geographical location, on-site audit scene photos / videos, audit site environmental parameters, and audit process records). During the acquisition process, an immutable timestamp is generated for each data item using a blockchain timestamp algorithm, and the terminal's unique hardware identifier is appended as the source identifier. Simultaneously, unstructured data (photos / videos) is also processed. The video undergoes MD5 hash verification to ensure data integrity and traceability. Next, the system introduces a multi-dimensional anomaly monitoring algorithm to perform real-time analysis of all collected data. If the algorithm detects that a matching scheduling instruction has not been signed for within the preset signing time limit, a priority-based secondary push notification mechanism is triggered, following the sequence of "SMS + terminal pop-up +..." The system pushes voice reminders in a gradient manner, while recording the reminder trigger time and number of reminders. If the algorithm determines that the review task is lagging behind by calculating the difference between the task progress and the time constraint (the lag threshold is set based on the 3σ principle of historical task completion time), it automatically generates a progress warning message containing the lag time, the node to be completed, and suggested remedial measures, and pushes it to the reviewer terminal and the scheduling management terminal respectively. If the algorithm detects anomalies in the review site through dimensions such as geographical location deviation, abnormal keyword recognition (based on BERT semantic analysis model), and environmental parameter threshold comparison in the on-site status data, it encapsulates the abnormal data into a standardized abnormal event message and feeds it back to the intelligent matching and dynamic scheduling engine module in real time, triggering an abnormal rescheduling evaluation based on a multi-objective optimization algorithm. The algorithm calculates the optimal rescheduling scheme by comprehensively considering factors such as the reviewer's availability, geographical accessibility, and qualification matching degree.If the algorithm verifies that the audit task is completed (requiring consistency among three types of data: task completion receipt, on-site acceptance record, and compliance verification result), the validity of the execution result is confirmed through the consensus verification algorithm. The result is then synchronized to the compliance verification and process evidence storage module for on-chain evidence storage. The entire process uses encryption algorithms to ensure secure instruction transmission, leverages multi-dimensional anomaly monitoring algorithms to achieve full-scenario risk identification, and relies on consensus verification algorithms to ensure the reliability of the execution result. This not only solves the problems of insecure instruction transmission, non-standard data collection, and untimely anomaly response in traditional scheduling, but also achieves intelligent, secure, and traceable management and control of the entire scheduling-execution-feedback process.

Claims

1. An intelligent matching system for certification auditing and auditor competency profiling based on big data, characterized in that: The system includes: a big data resource pool and auditor capability profile construction module, an intelligent matching and dynamic scheduling engine module, a compliance verification and process evidence storage module, and a scheduling execution and closed-loop management module. The big data resource pool and auditor capability profile construction module are used for the collection, cleaning, fusion and standardized storage of multi-source data, and to construct a dynamically iterative digital capability profile of auditors based on the standardized multi-source data. The intelligent matching and dynamic scheduling engine module is used to structurally decompose the target certification project, generate a standardized requirement feature vector for the project, perform similarity calculation and multi-objective constraint optimization matching between the standardized requirement feature vector for the project and the auditor's capability profile, so as to generate a compliance matching list, and perform real-time re-matching and conflict resolution for dynamic sudden scheduling events. The compliance verification and process evidence preservation module is used to perform full-process legal compliance verification on real-time re-matching, perform trusted evidence preservation and traceability of the entire matching and scheduling process, and generate and execute matching and scheduling instructions. The scheduling execution and closed-loop management module is used to push matching scheduling instructions to the target terminal in real time, collect auditor execution feedback data and on-site status data to form a full closed-loop management of scheduling-execution-feedback, and feed back the execution data to the big data resource pool module to realize profile iterative optimization.

2. The intelligent matching system for certification auditing and auditor competency profiling based on big data as described in claim 1, characterized in that, The specific steps for constructing a dynamically iterative digital competency profile of auditors are as follows: The multi-source data includes audit performance data, real-time status data, certification project basic data, regulatory compliance data, and spatiotemporal travel data. The digital capability profile of the auditor includes legal qualification tags, professional capability tags, performance compliance tags, spatiotemporal resource tags, and dynamic status tags. By calling the standardized dataset preprocessed from the big data resource pool, and based on the preset tagging system, feature extraction and classification labeling are performed on multi-source data to generate legal qualification tags, professional capability tags, performance compliance tags, spatiotemporal resource tags, and dynamic status tags. The legal qualification tags, professional capability tags, performance compliance tags, spatiotemporal resource tags, and dynamic status tags are then linked and integrated to construct an initial profile of the auditor's digital capabilities.

3. The intelligent matching system for certification auditing and auditor competency profiling based on big data as described in claim 2, characterized in that, The construction of a dynamically iterative digital capability profile for auditors also includes: Establish a real-time data interaction mechanism. When the big data resource pool collects a new round of multi-source data, it will automatically trigger the data preprocessing process, generate incremental standardized data, and update, supplement and correct the corresponding tags of the auditor's initial digital capability profile, so as to realize the real-time dynamic update of the auditor's digital capability profile. The quality of the constructed digital capability profiles of auditors is verified. By combining historical data retained in the big data resource pool and execution feedback data returned from the scheduling and closed-loop management modules, the consistency between the digital capability profile labels of auditors and the actual profile labels is compared, so as to eliminate deviation labels and correct erroneous labels. Based on updates to certification industry standards, adjustments to regulatory requirements, and changes in business scenarios, we optimize the labeling system and feature extraction rules to achieve continuous iterative optimization of auditors' digital capability profiles, ensuring that the profiles are highly aligned with auditors' actual capabilities, business needs, and regulatory requirements.

4. The intelligent matching system for certification auditing and auditor competency profiling based on big data as described in claim 1, characterized in that, The specific steps for similarity calculation and multi-objective constraint optimization matching are as follows: Step 1: Obtain the similarity weight and the review type matching weight. The similarity weight is used to adjust the proportion of each dimension's matching degree in the overall similarity calculation, and the review type matching weight is used to represent the weight proportion of the review type matching degree in the comprehensive similarity calculation. Step 2: When the audit type of the target certification project is a special audit, the audit type matching weight adjustment mechanism is triggered; when the audit type of the target certification project is a regular surveillance audit, the audit type matching weight adjustment mechanism is triggered. Step 3: Determine whether the matching failure rate is within the preset failure rate threshold range. If so, maintain the similarity baseline weight unchanged. If not, determine whether the matching failure rate is greater than the preset failure rate threshold upper limit. A mapping relationship between audit and audit type matching weight adjustment factors is constructed. This mapping relationship is used to represent the pre-established one-to-one correspondence between the audit type of the target certification project and the corresponding weight adjustment factor. The mapping of special audit type outputs the matching weight increase factor, and the mapping of regular supervision audit type outputs the matching weight decrease factor, so that the intelligent matching and dynamic scheduling engine module can directly query the corresponding weight adjustment factor based on the input audit type.

5. The intelligent matching system for certification auditing and auditor competency profiling based on big data as described in claim 4, characterized in that, The specific steps for triggering the review type matching weight adjustment mechanism are as follows: If the review type matching weight adjustment mechanism is triggered, the corresponding review type is input into the review-review type matching weight adjustment factor mapping relationship, the corresponding matching weight adjustment factor is output, and the result of doubling the review type matching baseline weight and the matching weight adjustment factor and then rounding down is used as the target review type matching weight. The specific steps for triggering the review type matching weight reduction mechanism are as follows: If the review type matching weight reduction mechanism is triggered, the corresponding review type will be input into the review-review type matching weight adjustment factor mapping relationship, and the corresponding matching weight reduction factor will be output. The result of multiplying the review type matching baseline weight and the matching weight reduction factor and then rounding up will be used as the target review type matching weight.

6. The intelligent matching system for certification auditing and auditor competency profiling based on big data as described in claim 4, characterized in that, The specific steps for determining whether the matching failure rate is within the preset failure rate critical interval are as follows: A failure rate-similarity weight correction factor mapping table is pre-constructed to establish a one-to-one correspondence between the matching failure offset, the matching failure deviation and the corresponding similarity weight correction factor, providing a clear basis for factor query for the dynamic correction of the similarity benchmark weight. The target similarity weight obtained after the doubling process meets the normalization constraint requirements of the similarity weight. If the matching failure rate is within the preset failure rate critical interval, the similarity benchmark weight remains unchanged. The preset failure rate critical interval represents the closed interval formed by the preset lower limit of the failure rate and the preset upper limit of the failure rate.

7. The intelligent matching system for certification auditing and auditor competency profiling based on big data as described in claim 6, characterized in that, The determination of whether the matching failure rate is within the preset failure rate critical range also includes: If the matching failure rate is greater than the preset failure rate threshold, the matching failure offset is input into the matching failure rate-similarity weight correction factor mapping table, and the similarity weight gain factor is output. The current similarity benchmark weight and the similarity weight gain factor are multiplied to obtain the target similarity weight. The matching failure offset represents the degree of positive deviation between the matching failure rate and the preset failure rate threshold. If the matching failure rate is less than the preset failure rate threshold, the matching failure deviation is input into the matching failure rate-similarity weight correction factor mapping table, and the similarity weight reduction factor is output. The current similarity benchmark weight and the similarity weight reduction factor are multiplied to obtain the target similarity weight. The matching failure deviation represents the degree of negative deviation between the matching failure rate and the preset failure rate threshold.

8. The intelligent matching system for certification auditing and auditor competency profiling based on big data as described in claim 1, characterized in that, The specific process for performing full-process legal compliance verification of real-time rematching is as follows: A preset initial on-chain latency threshold is set, which is the baseline time interval between the generation time of the scheduling instruction and the on-chain operation of the full-process evidence storage data. Obtain the scheduling instruction generation time parameter and the instruction receipt feedback time parameter, and calculate the actual instruction receipt delay, wherein the actual instruction receipt delay is the time difference between the instruction receipt feedback time and the scheduling instruction generation time. The actual instruction receipt delay is compared with the preset receipt delay range, and the whole process evidence storage data up-to-chain delay threshold is dynamically adjusted based on the comparison result. A pre-constructed mapping relationship between instruction receipt delay and on-chain delay threshold is established. This mapping relationship is used to characterize the pre-established segmented quantitative correspondence between the actual instruction receipt delay, the receipt delay offset, and the corresponding on-chain delay threshold gain coefficient. This allows the compliance verification and process evidence preservation module to directly query the corresponding gain coefficient based on the input actual instruction receipt delay or receipt delay offset, and then dynamically adjust the initial on-chain delay threshold to obtain the target full-process evidence preservation data on-chain delay threshold that is adapted to the current instruction receipt status.

9. The intelligent matching system for certification auditing and auditor competency profiling based on big data as described in claim 8, characterized in that, The full-process legal compliance verification for real-time rematching also includes: If the actual instruction receipt delay is within the preset receipt delay interval, it is determined that the instruction receipt is delayed. The current actual instruction receipt delay is input into the instruction receipt delay-on-chain delay threshold mapping relationship, and the first-level gain coefficient of the on-chain delay threshold is output. The preset initial on-chain delay threshold and the first-level gain coefficient of the on-chain delay threshold are multiplied to obtain the target full-process evidence storage data on-chain delay threshold. The preset receipt delay interval represents the closed interval formed by the preset lower limit of receipt delay and the preset upper limit of receipt delay. If the actual instruction receipt delay is less than the preset receipt delay lower limit, it is determined that the instruction is received immediately, and the initial on-chain delay threshold remains unchanged. If the actual instruction receipt delay is greater than the preset receipt delay upper limit, it is determined that the instruction receipt is delayed. The receipt delay offset is input into the instruction receipt delay-on-chain delay threshold mapping relationship, and the second-level gain coefficient of the on-chain delay threshold is output. The preset initial on-chain delay threshold and the second-level gain coefficient of the on-chain delay threshold are multiplied to obtain the target full-process evidence storage data on-chain delay threshold. The receipt delay offset represents the degree of positive deviation between the actual instruction receipt delay and the preset receipt delay upper limit.

10. The intelligent matching system for certification auditing and auditor competency profiling based on big data as described in claim 1, characterized in that, The matching and scheduling instructions include the target auditor's unique identifier, audit task details, time constraints, geographical constraints, and compliance requirements; The matching and scheduling instructions are pushed in real time to the target terminal that is bound to the unique identifier of the target auditor through a preset encrypted communication interface. The target terminal is a dedicated work terminal for auditors and supports instruction receipt, status feedback and data reporting functions. The auditor's execution feedback data includes the status of dispatch instruction receipt, audit task start status, audit task progress, audit task completion status, and audit exception description information. The on-site status data includes the auditor's real-time geographical location, on-site audit scene photos / videos, audit site environmental parameters, audit process records, and other information. During the collection process, each piece of data is timestamped and identified by its source to ensure data traceability. If a matching scheduling instruction is not acknowledged, a secondary push notification mechanism is triggered; if the review task is found to be lagging behind, a progress warning message is generated and pushed to the reviewer's terminal and the scheduling management terminal. If an anomaly is detected at the audit site, it is fed back to the intelligent matching and dynamic scheduling engine module in real time, triggering anomaly rescheduling assessment; if the audit task is detected to be completed, the execution result is confirmed and synchronized to the compliance verification and process evidence storage module.

Images