Unified social credit code data quality control method based on traceability technology

Abstract

The application discloses a unified social credit code data quality control method based on a traceability technology, relates to the technical field of quality control, and comprises the following steps: collecting original data sets indexed by a unified social credit code, calculating activity, and forming a data stream with activity annotation; constructing a tree-shaped credit primitive structure through a depth-first algorithm, and determining an authoritative data source in combination with a Gaussian integral minimization model; completing conflict resolution by using dynamic weight distribution, and outputting a consistent credit archive; sequentially performing structured verification, semantic consistency verification and business logic comparison, identifying deviation conditions of key fields, and generating a difference report; completing consistency verification through version number verification and block chain storage based on the difference report; and if the problem is not solved within a time limit, the problem is transferred to manual work order processing. The application effectively reduces risks caused by multi-source data conflicts and deviations, and provides accurate and reliable data basis for credit information management, financial supervision and decision support.

Description

Technical Field

[0001] This invention relates to the field of quality control technology, and in particular to a method for quality control of unified social credit code data based on traceability technology. Background Technology

[0002] In recent years, with the construction of the social credit system and the development of big data technology, the unified social credit code has become an important index for enterprise and organizational information management, promoting research on the collection, integration, and application of various credit data. Existing technologies, through multi-source data collection and data cleaning methods, can obtain core information such as enterprise entity, equity structure, registered capital, and business scope, and standardize the data to provide basic data support for credit evaluation, risk control, and supervision. Meanwhile, data quality control methods are continuously being optimized, for example, using activity indicators to assess data update frequency and reliability, and improving data processing efficiency through hierarchical indexing and structured storage. Nevertheless, how to achieve structured association, conflict resolution, and reliable verification of key fields in a multi-source, heterogeneous data environment remains a key focus and challenge of current research.

[0003] CN119648246A discloses a tracking and tracing system based on big data analysis. This system cleans and verifies raw data through a data quality verification module and uses data comparison algorithms to verify the consistency of multi-source data. It also employs anonymization and access control technologies to ensure data confidentiality during storage and transmission. This technology can improve data accuracy and regulatory efficiency to some extent and promote information sharing among participants. However, the technical solution mainly relies on data comparison and access control methods, lacking fine-grained processing of the structured correlations and activity differences between key fields in multi-source data. It cannot fully address the issues of heterogeneous data conflicts, dynamic authoritative source determination, and hierarchical identification of key field deviations.

[0004] CN120234211B discloses an intelligent data quality monitoring method and system. By monitoring the integrity, accuracy, and consistency of data streams, it detects and processes abnormal data and provides report generation and alarm mechanisms. While this method is effective in monitoring single data dimensions or static rules, it still has shortcomings in practical applications: it cannot perform recursive structured analysis of semantic dependencies of key fields, nor can it dynamically generate hierarchical difference groups based on data activity. This results in insufficient precision in identifying data deviations and limited risk visualization and decision support capabilities.

[0005] In summary, existing big data tracking and quality monitoring technologies still suffer from limitations in resolving multi-source data conflicts and insufficient structured and hierarchical analysis of key fields when processing data related to the Unified Social Credit Code. This invention provides a Unified Social Credit Code data quality control method based on source tracing technology. It uses the Unified Social Credit Code as the root node and recursively attaches key fields and activity levels according to semantic dependencies, forming a multi-level, logically clear tree structure to achieve structured association of key fields. Furthermore, it utilizes statistical optimization methods to determine authoritative sources in multi-source data and resolves field conflicts based on dynamic weight adjustments, achieving consistent credit profile output. Summary of the Invention

[0006] The purpose of this section is to outline some aspects of the embodiments of the present invention and to briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section, as well as in the abstract and title of the present application, to avoid obscuring the purpose of this section, the abstract and title of the invention. Such simplifications or omissions shall not be used to limit the scope of the present invention.

[0007] In view of the aforementioned existing problems, the present invention is proposed.

[0008] Therefore, the technical problem solved by this invention is: how to improve the consistency, reliability and traceability of multi-source data.

[0009] To address the aforementioned technical problems, this invention provides the following technical solution: Collect a raw data set indexed by the unified social credit code, calculate activity levels, and form a data stream labeled with activity levels; construct a tree-like credit primitive structure from the data stream using a depth-first search algorithm, and determine authoritative data sources using a Gaussian integral minimization model; resolve conflicts using dynamic weight allocation and output a consistent credit profile; based on the consistent credit profile, sequentially perform structured verification, semantic consistency verification, and business logic comparison to identify deviations in key fields and generate a difference report; based on the difference report, complete consistency verification through version number verification and blockchain notarization; if the issue remains unresolved after a timeout, it is transferred to a manual work order for processing.

[0010] In a preferred embodiment of the present invention, the calculation of the activity level includes: extracting the questioning frequency, update time difference, and new data weight factor for each record in the original dataset and calculating the activity level according to the following formula:

[0011]

[0012]

[0013] in, Activity level; For questioning frequency factor, which is a continuous value of questioning frequency normalized to [0,1]; , 2. 3 represents the weighting coefficient; The decay rate over time. Half-life; This is the time decay coefficient; This is the difference between the current time and the last time the data was updated. The maximum effective time window is defined; if the activity level is greater than the 80th percentile of the historical activity distribution, it is marked as high activity; otherwise, it is marked as low activity.

[0014] As a preferred embodiment of the present invention, the step of constructing a tree-like credit primitive structure using a depth-first search algorithm includes: receiving the data stream and using the unified social credit code field as the root node of the tree structure, while binding an activity value; taking the key fields in the data stream as input and generating a semantic dependency matrix according to the inherent logical relationship between the key fields, wherein a matrix element of 1 indicates the existence of a parent-child relationship or logical dependency, otherwise it is 0; starting from the root node, recursively expanding the key fields marked as having a dependency relationship in the semantic dependency matrix one by one, and attaching each key field and its corresponding activity value together to form a child node and its activity value, thereby obtaining a tree-like credit primitive structure; the semantic dependency matrix is ​​generated based on a predefined business rule base, which defines the parent-child dependency relationship between the key fields.

[0015] As a preferred embodiment of the present invention, the step of determining the authoritative data source by combining the Gaussian integral minimization model includes: extracting path fields and activity levels for each branch path in the tree-like credit primitive structure to form a path field set; calculating the path field deviation for each path field: if the field type is a numeric field, then using the arithmetic difference to calculate the absolute value of the difference between the current value of the field and the weighted average of all candidate values ​​of the corresponding field in the same data source set; if the field type is a date field, then calculating the difference in the number of days between the current value of the field and the weighted average of all candidate values ​​of the corresponding field in the same data source set. The absolute value; if the field type is a text field, the field deviation is calculated based on the edit distance or string similarity; for all path fields of each path, the deviation of each path field is input into a normal distribution function to obtain the field Gaussian integral value; the mean of the normal distribution function is the mean of the historical values ​​of the field, and the standard deviation is the standard deviation of the historical values; the Gaussian integral value of the field in the path and the activity are calculated by weighted summation to obtain the path matching score; the path matching score and the activity are weighted summation to obtain the path residual; the data sources corresponding to the paths whose path residuals are less than the preset residual threshold are included in the authoritative data source set.

[0016] As a preferred embodiment of the present invention, the step of resolving conflicts and outputting a consistent credit profile using dynamic weight allocation includes: obtaining the path field deviation for conflicting path fields in the tree-like credit primitive structure; adjusting the path field weights proportionally based on the path field deviation and path residual, and normalizing the path weights of the same path field after adjustment; the initial value of the path field weight is the activity level value; using the adjusted path weights to perform weighted fusion on the conflicting path fields, selecting the path field value with the highest weighted score as the authoritative field value, and recording the corresponding data source; combining all path fields after dynamic weight allocation and conflict resolution to generate a consistent credit profile, including the authoritative field value, the corresponding authoritative data source, and the activity level.

[0017] In a preferred embodiment of the present invention, the formula for calculating the proportional adjustment of the path field weight is as follows:

[0018]

[0019] in, The path deviation ratio is calculated by dividing the path field deviation by the path residual. This is the initial value for the path field weight. This refers to the adjusted path field weight.

[0020] In a preferred embodiment of the present invention, the identification of deviations in key fields includes: extracting the value of each authoritative field, the corresponding path residual, and the corresponding activity level from the consistency credit profile; constructing a deviation detection vector, wherein the elements of the deviation detection vector are triples <authoritative field name, path residual, activity level>; classifying the deviation detection vector into deviation levels, marking fields with path residuals below a first threshold as low deviation, fields with path residuals between the first and second thresholds as medium deviation, and fields with path residuals above the second threshold as high deviation; generating a key field deviation distribution table based on activity level and deviation level, marking fields with high activity and high deviation as sensitive fields, and fields with low activity and low deviation as stable fields.

[0021] As a preferred embodiment of the present invention, the step of generating a difference report based on a hierarchical decision tree model includes: constructing a two-dimensional deviation projection matrix based on a key field deviation distribution table, wherein rows represent deviation levels and columns represent activity intervals; performing hierarchical clustering operations on the two-dimensional deviation projection matrix to divide authoritative fields with the same deviation level and activity level into the same difference group; assigning difference labels to each difference group, and generating a difference report by combining the historical version records of authoritative fields within the difference group.

[0022] As a preferred embodiment of the present invention, the consistency verification through version number verification and blockchain storage includes: verifying sensitive fields in the difference report according to their version numbers in the set of authoritative data sources; if the version number sequence of the same unified social credit code field is discontinuous in different data source paths or if there are cases where the same version number corresponds to different authoritative field values, then a version number conflict is considered to have occurred; if a version number conflict occurs, priority is given to comparing the consistency of authoritative field values; if the authoritative field values ​​are consistent, then only the version number record is updated; if the authoritative field values ​​are different, then the authoritative field values ​​and the corresponding authoritative data sources are written into the blockchain to generate an immutable difference certificate.

[0023] In a second aspect, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, they implement the steps of the unified social credit code data quality control method based on traceability technology as described in the present invention.

[0024] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, they implement the steps of the unified social credit code data quality control method based on traceability technology as described in the present invention.

[0025] The beneficial effects of this invention are as follows: This invention can dynamically identify high-risk and sensitive fields based on data update frequency and historical activity, while distinguishing stable fields with low activity, thereby enhancing data consistency and reliability; by classifying and clustering deviations in key fields, this invention improves the ability to detect abnormal and conflicting data, which helps to handle potential problems in a timely manner and reduce the need for manual intervention; based on a traceable verification and evidence storage mechanism, it further ensures that the data processing process is transparent and tamper-proof.

[0026] This invention effectively reduces the risks caused by conflicts and biases in multi-source data, providing an accurate and reliable data foundation for credit information management, financial supervision, and decision support, while improving the efficiency and transparency of overall data governance. Attached Figure Description

[0027] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0028] Figure 1 This is a flowchart of the unified social credit code data quality control method based on traceability technology as shown in this invention. Detailed Implementation

[0029] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0030] Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort should fall within the scope of protection of this invention.

[0031] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0032] According to an embodiment of the present invention, in combination Figure 1 The flowchart shown illustrates a unified social credit code data quality control method based on traceability technology, which includes:

[0033] S1: Collect the original data set indexed by the unified social credit code, calculate the activity level, and form a data stream labeled with the activity level.

[0034] S1.1: Collect the raw data set indexed by the unified social credit code, covering 12 key fields.

[0035] By simultaneously scraping data from official platforms, credit reporting agencies, and local government departments, the collected data mainly consisted of 12 key fields. These 12 key fields include: entity name, unified code, legal person information, equity structure, registered capital, establishment date, registered address, business scope, enterprise status, operating period, registration authority, and key management personnel.

[0036] Ideally, to balance data coverage and timeliness, a hybrid strategy is adopted during the crawling process, with asynchronous submission as the main method and synchronous submission as a supplement. Priority queues are set for crawling tasks from different sources, and data sources with high update frequency are processed first.

[0037] Specifically, data source tasks with high update frequency are added to a priority queue, and multi-threaded parallel scheduling of the crawling process is used, while monitoring the crawling status and response time. After each crawl, the crawled record is compared with the historical record to confirm whether there are duplicates or conflicts, ensuring the integrity of fields, source identifier, and crawling time of each record.

[0038] The final output is the original data set.

[0039] S1.2: Calculate activity labels to form a data stream with activity labels.

[0040] Input the original dataset, and for each record, extract the frequency of questioning, the difference in update time, and the weighting factor of new data to calculate the activity level using the following formula:

[0041]

[0042]

[0043] in, Activity level, ranging from [0,1]; The frequency factor for questioning is a continuous value of the questioning frequency normalized to [0,1]. , 2. 3 represents the weighting coefficient, with recommended values ​​of 0.4, 0.3, and 0.3, and adjustments are allowed based on business needs. The decay rate over time. Half-life; This is the time decay coefficient; This is the difference between the current time and the last time the data was updated. This is the maximum effective time window used for linear freshness truncation, and the default value is 720 hours (i.e., 30 days).

[0044] Among them, the frequency of questioning refers to the number of times the corresponding field content is marked as suspected error or inconsistency by a third party or system, and the normalized value to [0,1] is used as the frequency of questioning factor.

[0045] If the activity level is greater than the 80th percentile of the historical activity distribution, it is marked as high activity; otherwise, it is marked as low activity.

[0046] S2: Construct a tree-like credit primitive structure from the data stream using a depth-first search algorithm, and determine the authoritative data source by combining it with a Gaussian integral minimization model; resolve conflicts using dynamic weight allocation, and output a consistent credit profile.

[0047] S2.1: Construct a tree-like credit primitive structure using a depth-first search algorithm.

[0048] S2.1.1: Receives the data stream and uses the Unified Social Credit Code field as the root node of the tree structure, while also binding an activity value. The activity value serves as the initial weight of the node, and the setting of the root node ensures that each credit record has a unified index in the tree structure.

[0049] S2.1.2: Take the key fields in the data stream as input and generate a semantic dependency matrix according to the inherent logical relationship between the key fields. The matrix element is 1 to indicate the existence of a parent-child relationship or logical dependency, otherwise it is 0.

[0050] It should be noted that the semantic dependency matrix is ​​generated based on a predefined business rule base. This base defines the parent-child dependencies, logical order, and constraints between key fields. For example, legal entity information depends on the entity name, and equity structure depends on legal entity information. When generating the matrix, the field activity level of each record needs to be recorded so that this activity level information can be synchronously attached to the corresponding child nodes during subsequent recursion, thus binding node activity level with the actual field activity level.

[0051] S2.1.3: Starting from the root node, recursively expand each key field marked as a dependency in the semantic dependency matrix, and attach each key field and its corresponding activity level to form child nodes and child node activity levels, thus obtaining a tree-like credit primitive structure.

[0052] For example, information about the legal entity and key management personnel depends on the entity name / unified code; the registered address depends on the registration authority; and the equity structure depends on the legal entity information.

[0053] During each expansion, the activity level of the current node is passed to its child nodes, and weighted in conjunction with the activity level of the child nodes' own fields to form the child node's activity level. During the recursive process, the node path, node activity level, and node origin are recorded to form a complete tree-like credit primitive structure.

[0054] S2.2: Determine the authoritative data source by combining the Gaussian integral minimization model.

[0055] S2.2.1: Extract the path field and activity level for each branch path in the tree-like credit primitive structure to form a path field set. Each path field includes a field value, field type, and activity level.

[0056] S2.2.2: Calculate the path field deviation for each path field:

[0057] If the field type is numeric, the arithmetic difference is used to calculate the absolute value of the difference between the current value of the field and the weighted mean of all candidate values ​​of the corresponding field in the same data source set; where the weighted mean is provided by activity level, and fields with higher activity levels have a larger weight in the mean calculation.

[0058] If the field type is a date field, calculate the absolute value of the difference in days between the current value of the field and the weighted average of all candidate values ​​of the corresponding field in the same data source set.

[0059] If the field type is text, the field bias is calculated based on edit distance or string similarity. Specifically, edit distance calculation includes character-by-character comparison, insertion, deletion, and replacement operations, ultimately obtaining a quantified similarity value between the current field value and the weighted average string in the candidate value set. Text bias calculation allows for authoritative evaluation of unstructured text fields such as company names and registered addresses.

[0060] For all path fields of each path, input the deviation of each path field into the normal distribution function to obtain the Gaussian integral value of the field; the mean of the normal distribution function is the mean of the historical values ​​of the field, and the standard deviation is the standard deviation of the historical values.

[0061] S2.2.3: Calculate the path matching score by weighted summation of the Gaussian integral value and activity level of the fields in the path.

[0062] Specifically, the path matching score is calculated based on the deviation of all fields in each path. The Gaussian integral value of each field represents the probability density that the deviation between the current value of the corresponding field and the weighted mean of the candidate value set conforms to a normal distribution. The field weight is provided by the field's own activity level, and its calculation is first performed by multiplying the Gaussian integral value, activity level, path hierarchy decay coefficient, and field category weight; the path matching score is the sum of the weighted values ​​of all fields on the path. Among them, the root node field contributes the most to path matching, and the contribution of child nodes gradually decreases with the increase of hierarchy. The path hierarchy decay coefficient ranges from [0,1]. The closer it is to 1, the slower the decay, and the more emphasis is placed on leaf node fields; the smaller the path hierarchy decay coefficient, the faster the decay, emphasizing the importance of higher-level node fields. The default value is recommended to be 0.9, which can be adjusted according to actual business needs. The field category weight is used to distinguish the degree of contribution of different types of fields to the authority determination in path matching, and can be predefined through the business rule base.

[0063] S2.2.4: Calculate the path matching score and the activity score weighted sum to obtain the path residual, which represents the difference between the overall path deviation and the activity prior.

[0064] Specifically, the path residual is calculated as the absolute value of the difference between the path matching score and the cumulative activity value. By taking the absolute value, the degree to which the path matching score deviates from the path activity value can be clearly quantified.

[0065] S2.2.5: Include the data sources corresponding to paths with path residuals less than a preset residual threshold into the authoritative data source set.

[0066] S2.3: Utilize dynamic weight allocation to resolve conflicts and output a consistent credit profile.

[0067] For conflicting path fields in the tree-like credit primitive structure, obtain the path field deviation; that is, multiple data sources corresponding to the same unified social credit code have inconsistent values ​​on the same field.

[0068] Based on the path field deviation and path residual, the path field weights are adjusted proportionally, and the path weights of the same path field are normalized after adjustment.

[0069] The formula for calculating the proportional adjustment of path field weights is as follows:

[0070]

[0071] in, The path deviation ratio is calculated by dividing the path field deviation by the path residual. This is the initial value for the path field weight. The adjusted path field weights are initially set to the activity level values.

[0072] The conflicting path fields are weighted and merged using adjusted path weights, and the path field value with the highest weighted score is selected as the authoritative field value, while the corresponding data source is recorded. It can be seen that this invention not only solves the problem of multi-source field conflicts but also ensures the traceability of the authoritative field value's source, while taking into account field activity and historical deviations, thus forming a highly reliable and consistent credit profile.

[0073] All path fields, after dynamic weight allocation and conflict resolution, are combined to generate a consistent credit profile, which includes authoritative field values, corresponding authoritative data sources, and activity levels.

[0074] S3: Based on the consistent credit profile, perform structured verification, semantic consistency verification and business logic comparison in sequence to identify deviations in key fields and generate a difference report.

[0075] S3.1: Identify deviations in key fields.

[0076] Extract the value of each authoritative field, the corresponding path residual, and the corresponding activity level from the consistency credit profile to construct a deviation detection vector. The elements of the deviation detection vector are triples <authoritative field name, path residual, activity level>.

[0077] It should be noted that the deviation detection vector, as a quantitative tool, can accurately describe the deviation level of each field and its activity weight in the data stream. By correlating path residuals with activity, subsequent analysis can identify which field deviations may affect the reliability of the overall credit profile.

[0078] The deviation detection vector is classified into deviation levels. Fields with path residuals below a first threshold are marked as low deviation, fields with path residuals between the first and second thresholds are marked as medium deviation, and fields with path residuals above the second threshold are marked as high deviation. The first and second thresholds are determined by business strategies or statistical quantiles of historical data and can be adjusted according to different business scenarios; this embodiment of the invention is not unique.

[0079] A key field deviation distribution table is generated based on activity level and deviation level. Fields with high activity and high deviation are marked as sensitive fields, while fields with low activity and low deviation are marked as stable fields.

[0080] Specifically, a two-dimensional mapping is generated based on activity level and deviation level. Sensitive fields indicate that these fields have high influence and large deviation in multi-source data, which may have a significant impact on business decisions and credit assessment; stable fields indicate that the data has high reliability and makes a stable contribution to the overall credit profile.

[0081] The deviation distribution table allows technicians to intuitively grasp the deviation of key fields and provides basic data for difference group classification and risk control.

[0082] Furthermore, during the deviation identification process, historical authoritative data source information for each field is recorded to trace the field's origin during deviation analysis. By combining historical version data and path residuals, deviation detection can not only reflect the current state of a field but also provide analysis of field change trends, offering a reference for enterprise credit monitoring and auditing.

[0083] S3.2: Generate a discrepancy report based on the hierarchical decision tree model.

[0084] A. Based on the key field deviation distribution table, construct a two-dimensional deviation projection matrix by combining deviation level and activity level, where rows represent deviation level and columns represent activity level intervals.

[0085] The matrix elements are corresponding sets of authoritative fields, and the matrix can intuitively present the distribution of field bias and activity. The construction of a two-dimensional matrix allows technical personnel to quickly identify high-risk fields through matrix visualization.

[0086] Hierarchical clustering is performed on the two-dimensional bias projection matrix to group authoritative fields with the same bias level and activity level into the same difference group.

[0087] In this embodiment of the invention, performing hierarchical clustering operations includes:

[0088] B. Form preliminary candidate difference groups from the set of fields located in the same cell of the two-dimensional deviation projection matrix, and record the number of fields in each group, average activity, deviation level distribution, and parent node information (root node or upper-level key field); traverse the fields in each candidate difference group and determine the parent-child relationship based on the semantic dependency matrix:

[0089] If a field in a group has a parent node in another cell of the two-dimensional deviation projection matrix, the corresponding field is moved out of the current group and attached to the child level of the parent node's group, forming a child node hierarchy; if a field in a group does not have a parent node, it is retained in the top-level difference group of the current group.

[0090] For the adjusted candidate difference groups, recursively aggregate them according to their deviation level and average activity level: fields with high deviation and high activity are prioritized to form the top-level difference group; fields with low deviation or low activity are included as child nodes in the top-level group if they have a parent-child dependency with fields in the top-level difference group; otherwise, they form medium- and low-level difference groups; the average activity and deviation distribution of the fields within each aggregation level are updated.

[0091] Optionally, a consistency check is performed on the generated hierarchical difference groups: the difference in field activity within the group is within a reasonable range (based on historical activity fluctuations); the dependency relationship between fields within the group and the parent node conforms to the definition of a semantic dependency matrix (1 indicates dependency, 0 indicates no dependency).

[0092] C. Assign a difference label to each difference group and generate a difference report by combining the historical version records of the authoritative fields within the difference group.

[0093] The difference report includes difference labels, a set of authoritative field names within the difference group, path residual intervals and activity distributions of fields within the difference group, and historical authoritative data source identifiers for each field within the difference group.

[0094] It should be noted that the generation of the discrepancy report also includes traceability and auditability features. Each discrepancy group records the field source, path residuals, and activity change trajectory, ensuring that the root cause and change process of field anomalies can be traced in subsequent credit profile optimization, manual intervention, or regulatory review.

[0095] S4: Based on the difference report, complete the consistency verification through version number verification and blockchain evidence storage; if the problem is not resolved within the time limit, it will be transferred to a manual work order for processing.

[0096] In this embodiment of the invention, consistency verification through version number verification and blockchain evidence storage includes:

[0097] For sensitive fields in the discrepancy report, the version number in the set of authoritative data sources is used for verification. If the version number sequence of the same unified social credit code field is not continuous in different data source paths, or if there are cases where the same version number corresponds to different authoritative field values, then a version number conflict is considered to have occurred.

[0098] If a version number conflict occurs, the authoritative field value consistency comparison is performed first. If the authoritative field values ​​are consistent, only the version number record is updated; if the authoritative field values ​​are different, the authoritative field values ​​and the corresponding authoritative data source are written to the blockchain to generate an immutable difference certificate.

[0099] Furthermore, updating the version number record involves creating a latest version number record entry for each field, recording the field value, latest version number, source data source path, and generation timestamp. The new entry is then inserted into the existing version number sequence to ensure the sequence is complete, continuous, and traceable. These operations do not modify the actual field values, ensuring the stability of consistent credit profile data, while simultaneously forming a complete historical version number chain for future auditing and anomaly tracing.

[0100] By setting timeout thresholds for sensitive field error correction and blockchain notarization, a manual work order is automatically generated if consistency verification cannot be completed within a predefined time or if version number conflicts that cannot be automatically corrected exist. The work order includes field information, conflict type, involved data source path, activity level, and path residuals. The work order is sent to data management or auditing personnel through a unified interface, ensuring that anomalies that cannot be automatically handled can be manually intervened and ultimately corrected. The manual work order processing flow includes work order reception, field value verification, conflict cause analysis, and authoritative field value confirmation. Technical personnel correct and confirm the abnormal fields based on the discrepancy report and historical data records, and then send the correction results back.

[0101] Other aspects disclosed in embodiments of the present invention also include one or more processors and a memory.

[0102] The memory is used to store operable instructions that, when executed by one or more processors, cause one or more processors to perform operations, including the flow of the unified social credit code data quality control method based on traceability technology in the foregoing embodiments, especially... Figure 1 The flowchart of the method is shown.

[0103] Other aspects disclosed in the embodiments of the present invention also propose a computer-readable medium for storing software including instructions executable by one or more computers, which, upon execution, cause the one or more computers to perform operations including the flow of the unified social credit code data quality control method based on traceability technology as described in the foregoing embodiments, particularly... Figure 1 The flowchart of the method is shown.

[0104] It should be recognized that embodiments of the present invention may be implemented or carried out by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable storage medium.

[0105] The method can be implemented using standard programming techniques, including a non-transitory computer-readable storage medium configured with a computer program in the computer program, wherein the storage medium is configured such that the computer operates in a specific and predefined manner.

[0106] Each program can be implemented in a high-level procedural or object-oriented programming language to communicate with the computer system; however, if required, the program can be implemented in assembly or machine language.

[0107] In any case, the language can be either compiled or interpreted.

[0108] Furthermore, for this purpose, the program can run on programmed application-specific integrated circuits.

[0109] The processes described herein (or variations and / or combinations thereof) can be executed under the control of one or more computer systems configured with executable instructions, and can be implemented by hardware or a combination thereof as code (e.g., executable instructions, one or more computer programs, or one or more applications) that commonly executes on one or more processors. The computer program includes a plurality of instructions executable by one or more processors.

[0110] Furthermore, the method can be implemented in any suitable computing platform, including but not limited to personal computers, minicomputers, mainframes, workstations, networked or distributed computing environments, standalone or integrated computer platforms, or in communication with charged particle tools or other imaging devices.

[0111] Various aspects of the present invention can be implemented in machine-readable code stored on a non-transitory storage medium or device, whether portable or integrated into a computing platform, such as a hard disk, optical read and / or write storage medium, RAM, ROM, etc., such that it can be read by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the processes described herein.

[0112] Furthermore, machine-readable code, or parts thereof, can be transmitted via wired or wireless networks.

[0113] When such media includes instructions or programs that combine with a microprocessor or other data processor to implement the steps described above, the invention described herein includes these and other different types of non-transitory computer-readable storage media.

[0114] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A data quality control method for unified social credit codes based on traceability technology, characterized in that, include: Collect raw data sets indexed by the unified social credit code, calculate activity levels, and form a data stream labeled with activity levels; The data stream is used to construct a tree-like credit primitive structure through a depth-first search algorithm, and an authoritative data source is determined by combining it with a Gaussian integral minimization model; conflict resolution is completed by dynamic weight allocation, and a consistent credit profile is output. Based on the consistent credit profile, structured verification, semantic consistency verification, and business logic comparison are performed sequentially to identify deviations in key fields and generate a difference report. Based on the aforementioned difference report, consistency verification is completed through version number verification and blockchain evidence storage; If the issue is not resolved within the time limit, it will be transferred to a manual work order for processing. The method of constructing a tree-like credit primitive structure using a depth-first search algorithm includes: receiving the data stream and using the Unified Social Credit Code field as the root node of the tree structure, while binding an activity value; taking the key fields in the data stream as input and generating a semantic dependency matrix according to the inherent logical relationships between the key fields, where a matrix element of 1 indicates the existence of a parent-child relationship or logical dependency, and otherwise 0; starting from the root node, recursively expanding each key field marked as having a dependency relationship in the semantic dependency matrix, and attaching each key field and its corresponding activity value together to form a child node and its activity value, thus obtaining a tree-like credit primitive structure; the semantic dependency matrix is ​​generated based on a predefined business rule base, which defines the parent-child dependency relationships between key fields; The method of determining authoritative data sources using the Gaussian integral minimization model includes: extracting path fields and activity levels for each branch path in the tree-like credit primitive structure to form a path field set; calculating the path field deviation for each path field: if the field type is a numeric field, the absolute value of the difference between the current value of the field and the weighted mean of all candidate values ​​of the corresponding field in the same data source set is calculated using arithmetic difference; if the field type is a date field, the absolute value of the difference in the number of days between the current value of the field and the weighted mean of all candidate values ​​of the corresponding field in the same data source set is calculated; if the field type is a text field, the field deviation is calculated based on edit distance or string similarity; for all path fields of each path, the deviation of each path field is input into a normal distribution function to obtain the field Gaussian integral value; the mean of the normal distribution function is the mean of the historical values ​​of the field, and the standard deviation is the standard deviation of the historical values; calculating the path matching score by weighted summing of the field Gaussian integral value and activity level in the path; calculating the path residual by weighted summing of the path matching score and activity level; and including the data source corresponding to the path with a path residual less than a preset residual threshold into the authoritative data source set.

2. The unified social credit code data quality control method based on traceability technology as described in claim 1, characterized in that: The calculation of the activity level includes: For each record in the original dataset, extract the frequency of questioning, the difference in update time, and the weighting factor of new data, and calculate the activity level using the following formula: in, Activity level; For questioning frequency factor, which is a continuous value of questioning frequency normalized to [0,1]; , 2. 3 represents the weighting coefficient; The decay rate over time. Half-life; This is the time decay coefficient; This is the difference between the current time and the last time the data was updated. The maximum effective time window; If the activity level is greater than the 80th percentile of the historical activity distribution, it is marked as high activity; otherwise, it is marked as low activity.

3. The unified social credit code data quality control method based on traceability technology as described in claim 2, characterized in that: The process of resolving conflicts using dynamic weight allocation and outputting a consistent credit profile includes: For conflicting path fields in the tree-like credit primitive structure, obtain the path field deviation; Based on the path field deviation and path residual, the path field weights are adjusted proportionally, and the path field weights for the same path field are normalized after adjustment. The initial value for the path field weight is the activity level value; The conflicting path fields are weighted and merged using the adjusted path field weights. The path field value with the highest weighted score is selected as the authoritative field value, and the corresponding data source is recorded. All path fields, after dynamic weight allocation and conflict resolution, are combined to generate a consistent credit profile, which includes authoritative field values, corresponding authoritative data sources, and activity levels.

4. The unified social credit code data quality control method based on traceability technology as described in claim 3, characterized in that: The formula for calculating the proportionally adjusted path field weight is as follows: in, The path deviation ratio is calculated by dividing the path field deviation by the path residual. This is the initial value for the path field weight. This refers to the adjusted path field weight.

5. The unified social credit code data quality control method based on traceability technology as described in claim 4, characterized in that: The deviations in identifying key fields include: Extract the value of each authoritative field, the corresponding path residual, and the corresponding activity level from the consistency credit profile to construct a deviation detection vector. The elements of the deviation detection vector are triples <authoritative field name, path residual, activity level>. The deviation detection vector is divided into deviation levels. Fields with path residuals below a first threshold are marked as low deviation, fields with path residuals between the first and second thresholds are marked as medium deviation, and fields with path residuals above the second threshold are marked as high deviation. A key field deviation distribution table is generated based on activity level and deviation level. Fields with high activity and high deviation are marked as sensitive fields, while fields with low activity and low deviation are marked as stable fields.

6. The unified social credit code data quality control method based on traceability technology as described in claim 5, characterized in that: The generation of the difference report includes: Based on the key field deviation distribution table, a two-dimensional deviation projection matrix is ​​constructed by combining deviation level and activity level, where rows represent deviation level and columns represent activity level intervals; Hierarchical clustering is performed on the two-dimensional deviation projection matrix to divide authoritative fields with the same deviation level and activity level into the same difference group; Assign a difference label to each difference group and generate a difference report by combining the historical version records of the authoritative fields within the difference group.

7. The unified social credit code data quality control method based on traceability technology as described in claim 6, characterized in that: The consistency verification process, which involves version number verification and blockchain notarization, includes: For sensitive fields in the discrepancy report, the version number in the set of authoritative data sources is used for verification. If the version number sequence of the same unified social credit code field is not continuous in different data source paths or there are cases where the same version number corresponds to different authoritative field values, then a version number conflict is considered to have occurred. If a version number conflict occurs, the authoritative field value consistency comparison is performed first. If the authoritative field values ​​are consistent, only the version number record is updated; if the authoritative field values ​​are different, the authoritative field values ​​and the corresponding authoritative data source are written to the blockchain to generate an immutable difference certificate.

8. A computer-readable medium for storing software, characterized in that: The software includes instructions executable by one or more computers, which cause the one or more computers to perform operations, including the process of the unified social credit code data quality control method based on traceability technology as described in any one of claims 1 to 7.

Images