Data quality automatic evaluation method and system for complex business scenarios

Abstract

This invention belongs to the field of data quality assessment technology, specifically relating to an automatic data quality assessment method and system for complex business scenarios. This invention, by parsing scenario description information and generating corresponding initial quality assessment rules, enables data quality assessment to automatically adapt to different business scenarios, improving the flexibility and versatility of the assessment method. By conducting initial assessments of target data across multiple dimensions such as structural integrity, consistency, and time-series stability, data quality can be characterized from multiple perspectives, improving the comprehensiveness and accuracy of the assessment results. Furthermore, by constructing a quality-related topology structure, the calling, referencing, and transmission relationships between data are expressed in a structured manner, extending data quality assessment from a single data dimension to a business-related structural dimension, enhancing contextual analysis capabilities.

Description

Technical Field

[0001] This invention belongs to the field of data quality assessment technology, specifically relating to an automatic data quality assessment method and system for complex business scenarios. Background Technology

[0002] As enterprises deepen their digital transformation, business systems are increasingly characterized by multi-source heterogeneity, cross-system collaboration, and highly coupled processes. Data flows frequently through complex business chains such as order processing, payment settlement, inventory management, and logistics scheduling, leading to data quality issues that are multi-dimensional and propagate across nodes. In such complex business scenarios, data quality is not only reflected in the correctness of a single field or table, but also involves comprehensive constraints on multiple dimensions such as structural integrity, consistency, and time sequence rationality. Furthermore, the data relationships between different business nodes also significantly impact the quality results.

[0003] In actual business operations, data typically flows and transforms continuously between different system modules along with business processes. For example, it flows from the order creation system to the payment system, and then further into the inventory deduction and logistics scheduling system. In addition, in complex business chains, data nodes often have both strong and weak relationships. For example, there is a strong constraint relationship between order data and payment data, while there is a weak relationship between order data and log data.

[0004] Existing data quality assessment methods often rely on static rules or single-dimensional verification, making it difficult to adapt to the dynamic changes in complex business scenarios. Furthermore, rule configuration is usually manually maintained, resulting in poor scalability. These methods often use single data objects or tables as analysis units, lacking the ability to comprehensively depict cross-system data relationships and business flow paths. They fail to reflect the propagation impact of data quality in the business chain. Multi-dimensional quality assessments lack dynamic coordination mechanisms for dimensional conflicts, typically employing fixed weighting methods for fusion, which can easily lead to biased assessment results. They often remain at the level of static correlation expression, failing to support the propagation analysis and precise location of quality problems, and thus cannot meet the refined assessment needs of complex business scenarios. Summary of the Invention

[0005] The purpose of this invention is to provide an automatic data quality assessment method and system for complex business scenarios, which can realize multi-dimensional collaborative assessment of data quality, cross-node propagation analysis, and precise location of quality problems in complex business scenarios, thereby significantly improving the comprehensiveness, adaptability, and interpretability of data quality assessment.

[0006] The specific technical solution adopted by this invention is as follows: Automatic data quality assessment methods for complex business scenarios include: Based on the obtained initial quality assessment rules, a multi-dimensional initial quality assessment is performed on the target data to be assessed obtained in the target business scenario to obtain multi-dimensional initial assessment results. Construct a quality correlation topology based on the contextual data of the target data to be evaluated and the target data to be evaluated. Based on the quality correlation topology, obtain the difference data of the multidimensional initial evaluation results, and obtain the conflict propagation path based on the difference data; The evaluation weights are obtained based on the differential data and the initial multidimensional evaluation results. The differential data are then fused based on the evaluation weights to obtain the fused evaluation results for the quality of each dimension. The path structure information is obtained based on the quality association topology, and the comprehensive quality score of the target data to be evaluated is obtained based on the fusion evaluation results and the path structure information. The quality level threshold determination result is obtained based on the comprehensive quality score, and the corresponding quality level and quality risk location information is output based on the quality level threshold determination result.

[0007] In a preferred embodiment, the initial quality assessment rules are obtained in the following ways: Obtain rule mapping identifiers based on the acquired scenario type and business characteristics; Retrieve the corresponding rule template data from the pre-built rule library based on the rule mapping identifier; The rule template data is filtered according to the scenario type to obtain the target rule corresponding to the scenario type; Based on business characteristics and target rules, obtain the initial quality assessment rules corresponding to the target business scenario.

[0008] In a preferred embodiment, the methods for obtaining the scenario type and business characteristics include: Obtain scenario description information based on the target business scenario, and combine it with a pre-built scenario semantic knowledge base to obtain the corresponding scenario semantic representation data; Match the scene semantic representation data with the scene semantic tags in the pre-built scene semantic knowledge base to obtain the scene type corresponding to the target business scene; Based on the scenario type and scenario semantic representation data, candidate business features corresponding to the scenario type are obtained from the pre-set business feature library; Obtain the degree of scene matching between scene semantic representation data and candidate business features, and filter the candidate business features to obtain the target business features; The scenario description information is structured and labeled according to the target business characteristics to obtain the business characteristics corresponding to the target business scenario.

[0009] In a preferred embodiment, the target data to be evaluated, obtained from the target business scenario, is subjected to a multi-dimensional initial quality assessment according to initial quality assessment rules, resulting in multi-dimensional initial assessment results, including: Obtain the target data to be evaluated based on the target business scenario; Based on the target data to be evaluated and the initial quality assessment rules, obtain the structural integrity dimension rules, consistency dimension rules, and time series stability dimension rules; The multi-dimensional rule execution structure is obtained based on the structural integrity dimension rules, the consistency dimension rules, and the time sequence stability dimension rules. Based on the multidimensional rule execution structure, the target data to be evaluated is evaluated in multiple dimensions to obtain the initial evaluation results of structural integrity, consistency and time series stability. The initial assessment results of structural integrity, consistency, and time series stability are summarized to form a multidimensional initial assessment result.

[0010] In a preferred embodiment, the multidimensional rule execution structure includes a structural integrity dimension, a consistency dimension, and a time sequence stability dimension.

[0011] In a preferred embodiment, a quality association topology is constructed based on the contextual data of the target data to be evaluated and the target data to be evaluated, including: Based on the first data obtained from the target data to be evaluated, obtain contextual candidate data corresponding to the target business scenario; Obtain node data based on the target data to be evaluated and the contextual candidate data; Based on the data call information, data reference information, and data transmission information obtained from the node data, the association identifier between nodes is obtained; A quality association topology is generated based on node data and the association identifiers between nodes. The quality association topology includes nodes and associated edges.

[0012] In a preferred embodiment, difference data of multidimensional initial evaluation results are obtained based on the quality correlation topology, and conflict propagation paths are obtained based on the difference data, including: Based on the quality correlation topology, obtain the multidimensional initial evaluation results corresponding to the node data, and perform dimension alignment processing on the multidimensional initial evaluation results to form a multidimensional evaluation result sequence. Difference data is obtained from adjacent nodes in the multidimensional evaluation result sequence and quality correlation topology; Obtain the difference propagation identifier based on the correlation edges of the difference data and the quality correlation topology; Path weights are obtained based on the differential propagation identifier and the quality-related topology. The conflict propagation intensity value is obtained based on the difference data and path weights; The path conflict intensity is obtained based on the conflict propagation intensity value and the quality-related topology. The dominant path of conflict propagation is obtained based on the path conflict intensity and preset threshold conditions, and the conflict propagation path between dimensional quality is obtained based on the dominant path.

[0013] In a preferred embodiment, evaluation weights are obtained based on the discrepancy data and the initial multidimensional evaluation results. The discrepancy data is then fused according to these evaluation weights to obtain a fused evaluation result for the quality of each dimension, including: Obtain dimension-aligned difference units based on the difference data and the initial multidimensional assessment results; The dimensional difference influence factor is obtained based on the dimensional alignment difference unit and the multidimensional initial evaluation results. Path constraint factors are obtained based on dimensional difference influence factors and conflict propagation paths in the quality-related topology. Initial evaluation weights are obtained based on dimensional difference influencing factors and path constraint factors; Dynamic evaluation weights are obtained based on the initial evaluation weights and the conflict propagation path. The path-weighted difference results are obtained based on the dynamic evaluation weights and difference data. The fusion evaluation results of the quality of each dimension are obtained based on the path-weighted difference results and the multidimensional initial evaluation results.

[0014] In a preferred embodiment, path structure information is obtained based on the quality-related topology, and a comprehensive quality score for the target data to be evaluated is obtained based on the fusion evaluation results and the path structure information. A quality level threshold determination result is obtained based on the comprehensive quality score, and the corresponding quality level and quality risk location information is output based on the quality level threshold determination result, including: Obtain path structure information based on the associated edges of the quality-related topology; Obtain path weight information based on path structure information and quality-related topology; Obtain path weighted scoring factors based on the fusion evaluation results and path weight information; The overall quality score of the target data to be evaluated is obtained based on the path-weighted scoring factors. The quality level threshold determination result is obtained based on the comprehensive quality score and the preset quality threshold. Risk path nodes are obtained based on the quality level threshold determination results and path structure information, and quality level and quality risk location information are obtained based on the risk path nodes.

[0015] This invention also provides an automatic data quality assessment system for complex business scenarios, used in the aforementioned automatic data quality assessment method for complex business scenarios, comprising: The multidimensional evaluation module performs a multidimensional initial quality evaluation on the target data to be evaluated obtained from the target business scenario based on the acquired initial quality evaluation rules, and obtains the multidimensional initial evaluation results. The data relationship module constructs a quality relationship topology based on the contextual data of the target data to be evaluated and the target data to be evaluated. The conflict path module obtains the difference data of the multidimensional initial evaluation results based on the quality correlation topology, and obtains the conflict propagation path based on the difference data; The fusion evaluation module obtains evaluation weights based on the differential data and the initial multidimensional evaluation results, and then performs fusion processing on the differential data according to the evaluation weights to obtain the fusion evaluation results for the quality of each dimension. The evaluation results module obtains path structure information based on the quality association topology, and obtains a comprehensive quality score for the target data to be evaluated based on the fusion evaluation results and path structure information. It then obtains the quality level threshold determination result based on the comprehensive quality score and outputs the corresponding quality level and quality risk location information based on the quality level threshold determination result.

[0016] The technical effects achieved by this invention are as follows: This invention, by parsing scenario description information and generating corresponding initial quality assessment rules, enables data quality assessment to automatically adapt to different business scenarios, improving the flexibility and versatility of the assessment method. By conducting initial assessments of target data in multiple dimensions such as structural integrity, consistency, and time-series stability, data quality can be characterized from multiple perspectives, improving the comprehensiveness and accuracy of the assessment results. By constructing a quality association topology structure, the calling, referencing, and transmission relationships between data are expressed in a structured manner, extending data quality assessment from a single data dimension to a business association structure dimension, improving the ability to analyze contextual associations. This invention acquires discrepancy data based on a graph structure and determines the discrepancy propagation path, enabling data quality deviations to be analyzed and propagated within the business process. This identifies the impact path and spread range of quality issues. By generating evaluation weights based on the discrepancy data and multidimensional initial evaluation results, and by fusing the discrepancy data, the impact of different quality dimensions can be dynamically coordinated, improving the rationality of the fusion evaluation. By combining path structure information and fusion evaluation results, a comprehensive quality score is generated, and the quality level and risk path nodes are output, enabling the locational analysis and hierarchical management of data quality issues in the business chain. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method provided by the present invention; Figure 2 This is a system module diagram provided by the present invention. Detailed Implementation

[0018] 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.

[0019] 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.

[0020] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in a preferred embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that mutually excludes other embodiments.

[0021] Furthermore, the present invention will be described in detail with reference to the schematic diagrams. When describing the embodiments of the present invention in detail, the schematic diagrams are merely examples for ease of explanation and should not limit the scope of protection of the present invention.

[0022] Please see the appendix Figure 1 As shown, an automatic data quality assessment method for complex business scenarios is provided, including: S1. Based on the obtained initial quality assessment rules, perform a multi-dimensional initial quality assessment on the target data to be assessed obtained in the target business scenario to obtain the multi-dimensional initial assessment results. S2. Construct a quality correlation topology based on the contextual data of the target data to be evaluated and the target data to be evaluated. S3. Obtain the difference data of the multidimensional initial evaluation results based on the quality correlation topology, and obtain the conflict propagation path based on the difference data; S4. Obtain the evaluation weights based on the difference data and the initial multidimensional evaluation results, and perform fusion processing on the difference data based on the evaluation weights to obtain the fusion evaluation results of the quality of each dimension. S5. Obtain path structure information based on the quality association topology, and obtain the comprehensive quality score of the target data to be evaluated based on the fusion evaluation results and path structure information. Obtain the quality level threshold determination result based on the comprehensive quality score, and output the corresponding quality level and quality risk positioning information based on the quality level threshold determination result.

[0023] As described in steps S1 to S5 above, the scenario description information of the target business scenario is obtained and semantically parsed to determine the corresponding scenario type and business characteristics, transforming the unstructured description of the business scenario information into a structured expression. Based on the scenario type and business characteristics, initial quality assessment rules are generated from a pre-defined rule system, enabling different data quality assessment strategies for different business scenarios, thus achieving scenario adaptation of the assessment rules. Initial assessments are then performed on the target data to be assessed across multiple quality dimensions according to the initial quality assessment rules, yielding multi-dimensional initial assessment results. These results characterize the basic quality status of the target data in different dimensions such as structural integrity, consistency, and time-series stability. Finally, the contextual relationship of the target data to be assessed is obtained. The system connects data and constructs a quality association topology based on target data and contextual data. This structured representation of data calls, references, and transmission relationships forms a data association topology within the business process. Based on this topology, it obtains the difference data between multi-dimensional initial assessment results and further determines the propagation path of quality differences within the graph structure. This characterizes the propagation direction and impact scope of quality issues in the business chain. Assessment weights are generated based on the difference data and multi-dimensional initial assessment results, and the difference data is then fused based on these weights. This achieves the coordination and weighted integration of the impacts between different quality dimensions, resulting in a fused assessment result for each quality dimension. The system also obtains path structure information based on the quality association topology. By combining the fusion assessment results and performing weighted analysis on the path, a comprehensive quality score is obtained. Based on preset thresholds, a quality level is determined, and corresponding quality level and quality risk location information are output. This enables the location and hierarchical output of data quality issues within the business path. By parsing scenario description information and generating corresponding initial quality assessment rules, data quality assessment can automatically adapt to different business scenarios, improving the flexibility and universality of the assessment method. Initial assessments of target data across multiple dimensions, such as structural integrity, consistency, and time-series stability, allow for a multi-faceted characterization of data quality, improving the comprehensiveness and accuracy of the assessment results. Furthermore, by constructing a quality association topology, the calls, references, and transmissions between data are linked. By structurally expressing the relationships, data quality assessment is expanded from a single data dimension to a business-related structural dimension, improving the ability to analyze contextual relationships. By acquiring differential data based on the graph structure and determining the differential propagation path, data quality deviations can be propagated and analyzed in the business process, thereby identifying the impact path and diffusion range of quality problems. By generating assessment weights based on differential data and multi-dimensional initial assessment results, and by fusing differential data, the impact of different quality dimensions can be dynamically coordinated, improving the rationality of the fusion assessment. By combining path structure information and fusion assessment results, a comprehensive quality score is generated, and the quality level and risk path nodes are output, enabling the locational analysis and hierarchical management of data quality problems in the business chain.

[0024] It should be noted that the methods for obtaining the initial quality assessment rules include: Obtain rule mapping identifiers based on the acquired scenario type and business characteristics; Retrieve the corresponding rule template data from the pre-built rule library based on the rule mapping identifier; The rule template data is filtered according to the scenario type to obtain the target rule corresponding to the scenario type; Based on business characteristics and target rules, obtain the initial quality assessment rules corresponding to the target business scenario.

[0025] The above-mentioned scenario types are standardized and encoded. For example, "e-commerce transaction scenario" is encoded as the first-class scenario identifier "C01", and "financial risk control scenario" is encoded as "C02". Simultaneously, business features are represented using multi-dimensional feature vectorization. For example, business features such as "order payment success rate" and "inventory deduction delay" are converted into feature weight vectors or feature labels. The scenario type encoding and business feature encoding are combined according to a preset mapping rule. For example, a rule mapping identifier is generated using the method of "scenario type encoding + feature clustering identifier". "C01-F03" represents the third-class feature combination in the e-commerce scenario, thus forming a unique retrieval rule index key. For example, in the e-commerce scenario, "order payment, inventory synchronization, and logistics delay" generate... The mapping identifier "C01-F12" standardizes the generation of rule index identifiers by uniformly encoding and combining scenario types and business characteristics. This allows for the unique identification of different complex business scenarios, thereby improving rule retrieval efficiency and consistency. Based on the rule mapping identifier, corresponding rule template data is retrieved from a pre-built rule library. This pre-built rule library is a structured rule database, indexed and organized according to the rule mapping identifier. It contains quality assessment rule templates for different business scenarios, such as structural integrity rule templates, consistency rule templates, and time-series stability rule templates. Each rule template includes a rule condition expression, a computational logic structure, and applicable business characteristic constraints. The rule mapping identifier serves as the lookup identifier. The query key performs matching and retrieval within the rule base index structure. For example, it uses hash indexes or key-value mapping to locate the corresponding rule template, thereby extracting rule template data relevant to the target scenario. For instance, when the mapping identifier is "C01-F12", it retrieves order consistency verification rules, inventory synchronization verification rules, and payment status consistency rules for the corresponding e-commerce scenario. The rule template retrieval is driven by the rule mapping identifier, enabling the rule retrieval process to have strong positioning capabilities and low computational overhead, thus improving rule invocation efficiency and scenario adaptability. It also parses the scenario applicability tag fields in the rule template data. For example, the rule template may contain identifiers such as "applicable scenario, e-commerce transaction" or "applicable scenario, financial risk control," and matches the current scenario type with the tags in the rule template. The system uses applicable scenario tags for matching and judgment, retaining only rule templates consistent with the current scenario type. For example, in an e-commerce transaction scenario, only rule templates related to order processing, payment settlement, and inventory synchronization are retained, while financial rule templates such as "credit risk scoring" are filtered out. Further secondary filtering is performed based on business characteristics. For instance, if the business characteristic includes "high payment failure rate," rules related to payment anomaly detection are prioritized. Filtering rule templates through scenario type constraints makes the rules more accurately fit the current business environment, thereby avoiding rule redundancy and mismatches, improving evaluation accuracy and execution efficiency. Business characteristics are mapped to parameterized variables in the rule templates; for example, "inventory synchronization delay" is mapped to a time threshold parameter in the rule.By mapping "payment success rate" to probability threshold parameters, rule parameters are dynamically configured based on the value range of business characteristics. For example, the inventory delay threshold can be set to "≤2 seconds," and the payment success rate threshold can be set to "≥99%." The rule logic is then instantiated, transforming abstract rule templates into directly executable initial quality assessment rules. For instance, the "order consistency rule template" in an e-commerce scenario is instantiated as "the order status must be consistent with the payment status, and the update time difference must not exceed T seconds." By driving rule parameterization and instantiation through business characteristics, the rules gain scenario adaptability, significantly improving their adaptability and evaluation accuracy for complex business scenarios.

[0026] It is worth mentioning that the methods for obtaining scenario types and business characteristics include: Obtain scenario description information based on the target business scenario, and combine it with a pre-built scenario semantic knowledge base to obtain the corresponding scenario semantic representation data; Match the scene semantic representation data with the scene semantic tags in the pre-built scene semantic knowledge base to obtain the scene type corresponding to the target business scene; Based on the scenario type and scenario semantic representation data, candidate business features corresponding to the scenario type are obtained from the pre-set business feature library; Obtain the degree of scene matching between scene semantic representation data and candidate business features, and filter the candidate business features to obtain the target business features; The scenario description information is structured and labeled according to the target business characteristics to obtain the business characteristics corresponding to the target business scenario.

[0027] The above describes how scenario description information is obtained by collecting multi-source heterogeneous data from the business system. This multi-source heterogeneous data includes business process configuration files, interface call logs, database table structure information, and user operation behavior records. For example, in an e-commerce business scenario, scenario description information includes fields such as order creation process, payment process, inventory deduction process, and logistics delivery process. After unified collection of multi-source data, it is fused through text parsing, field extraction, and structured parsing to form scenario description information in a unified format. A pre-built scenario semantic knowledge base is used to store semantic features, scenario tags, and mapping information between business features corresponding to different business scenarios. This knowledge base is used to... Historical business process data, business system configuration documents, data dictionaries, business rule files, and manually annotated typical business cases are collected and organized. The collected data undergoes entity extraction, semantic classification, and tag association processing to construct a data structure. Business description information from different business domains is collected, such as order processing flow descriptions, payment business descriptions, inventory management rules, and logistics scheduling processes. Key semantic elements that characterize the business scenario are extracted from the business description information, including business objects, business actions, business process nodes, and business constraints, such as "order creation," "payment confirmation," "inventory deduction," and "logistics delivery." The extracted semantic elements are then classified and labeled to build a data structure. Establish business scenario tags and associate them with corresponding business feature information to form a scenario semantic mapping relationship, ultimately generating a pre-built scenario semantic knowledge base. For example, in an e-commerce transaction scenario, the collected scenario description information is "After the user submits an order, the system completes payment verification, updates the inventory status based on the payment result, and then triggers the logistics and delivery process." Each scenario is standardized using semantic vector encoding. The scenario description information undergoes word segmentation, entity recognition, and semantic vector encoding to generate scenario semantic representation data. This data is then matched with scenario semantic tags in the pre-built scenario semantic knowledge base. The scenario semantic tags are used to classify different business scenarios. Other standardized semantic identifiers are used, and each tag corresponds to a preset semantic vector model. For example, the target business scenario description information "After the user submits an order and completes payment, the inventory status is updated according to the payment result" is obtained. The scenario description is converted into a semantic vector through semantic encoding. The semantic vectors corresponding to tags such as "e-commerce transaction scenario", "warehouse management scenario", and "logistics delivery scenario" are stored in the preset scenario semantic knowledge base. The matching degree between the target scenario and each scenario tag is calculated by using the cosine similarity formula. For example, the similarity with "e-commerce transaction scenario" is calculated to be 0.94, the similarity with "warehouse management scenario" is 0.78, and the similarity with "logistics delivery scenario" is 0.65. All candidate tags are sorted by similarity, and the tag with the highest similarity is selected as the target scenario type output. This determines the target business scenario type as an e-commerce transaction scenario. Unstructured business descriptions are automatically mapped to standardized scenario categories, thus achieving automatic identification and unified classification of scenario types. Based on the scenario type and scenario semantic representation data, corresponding candidate business features are retrieved from a pre-built business feature library. This pre-built feature library is a pre-constructed feature database, which is categorized and stored according to different scenario types. For example, in e-commerce scenarios, it includes order status flow features, inventory synchronization features, and payment success rate features; in financial scenarios, it includes transaction anomaly features and risk identification features. Features, etc., are indexed by scenario type. Corresponding feature templates are extracted from the business feature library and combined with scenario semantic representation data for semantic expansion and matching. This allows for the selection of candidate business features most relevant to the current business semantics. For example, when the semantic representation includes "payment failure retry," "payment anomaly features" and "retry mechanism features" are prioritized for matching. This enables scenario-adaptive generation of business features, eliminating the need for manual configuration and allowing for automatic semantic-driven generation. This improves the coverage and accuracy of feature representation and measures the degree of matching between scenario semantic representation data and candidate business features, thus measuring the strong correlation between candidate features and the current business scenario. Weak relationships, including methods such as semantic vector similarity calculation, keyword overlap calculation, or feature weight matching calculation, are used. For example, in e-commerce scenarios, "inventory deduction anomaly" and "inventory synchronization failure" have high semantic consistency and therefore a high degree of matching, while "device temperature anomaly" has a low degree of matching with e-commerce scenarios. Matching is filtered based on a preset threshold, for example, setting a similarity threshold of 0.75, retaining only candidate business features higher than the threshold, and then sorting and truncating them to obtain target business features. Semantic matching mechanisms are used to refine the selection of business features, removing irrelevant or weakly related features, thereby improving the targeting and accuracy of the features. By parsing, extracting semantics, and mapping attributes from scenario description information, business behavior elements are mapped into standardized structured fields. For example, "retry 3 times after order payment failure" is transformed into structured feature fields such as "number of payment failures = 3" and "retry mechanism = enabled." Standardized processing of these structured fields using unified coding rules ensures consistency in feature expression across different business scenarios. The final output is business features corresponding to the target business scenario, stored in structured form, and directly used for data quality assessment rule generation and graph construction. This process transforms business semantic information into structured feature data, improving automation, semantic understanding capabilities, and cross-business scenario adaptability.

[0028] In a preferred embodiment, a multi-dimensional initial quality assessment is performed on the target data to be assessed obtained from the target business scenario according to the acquired initial quality assessment rules, resulting in a multi-dimensional initial assessment result, including: S101. Obtain the target data to be evaluated based on the target business scenario; S102. Based on the target data to be evaluated and the initial quality assessment rules, obtain the structural integrity dimension rules, consistency dimension rules, and time series stability dimension rules; S103. Obtain the multi-dimensional rule execution structure based on the structural integrity dimension rules, consistency dimension rules, and time sequence stability dimension rules; S104. Based on the multidimensional rule execution structure, perform multidimensional evaluation of the target data to be evaluated to obtain the initial evaluation results of structural integrity, consistency and time series stability. S105. Summarize the initial assessment results of structural integrity, consistency, and time sequence stability to form a multidimensional initial assessment result.

[0029] It should be noted that the multidimensional rule execution structure includes structural integrity, consistency, and time sequence stability dimensions.

[0030] As described in steps S101 to S105 above, the target data to be evaluated is obtained by uniformly collecting data from multiple sources within the target business system. These multiple data sources include database query interfaces, message queue subscription interfaces, log collection interfaces, and business service API interfaces. For example, in an e-commerce business scenario, the target data to be evaluated includes order table data, payment record data, inventory change data, and logistics status data; in a financial scenario, it includes transaction flow data, account balance change data, and risk control log data. By uniformly extracting data from multiple sources and standardizing fields and aligning timestamps according to a preset data model, data for evaluation is formed, achieving unified acquisition of cross-system, multi-source heterogeneous business data. Standardization and integration provide a unified data structure foundation for the data to be evaluated. Based on the target data and initial quality assessment rules, which are typically composite rules instantiated from rule templates and contain multiple quality dimension sub-rules, the initial rules are decomposed according to their evaluation objectives. These rules are categorized into structural integrity rules (for detecting missing fields, abnormal field types, etc.), consistency rules (for detecting logical consistency across data sources or fields), and time-series rules (for detecting the reasonableness of data time order and delays). For example, in order processing, "order amount cannot be empty" is a structural integrity rule, "payment amount should equal order amount" is a consistency rule, and "payment amount should equal order amount" is a consistency rule. The rule "Payment time must not be earlier than order placement time" is a time-series rule. This approach breaks down complex evaluation rules into independently executable multi-dimensional rule units, thereby improving the modularity and scalability of rule execution. Rules of different dimensions are arranged according to a preset execution strategy. For example, structural integrity rules are executed first to filter invalid data, then consistency rules are executed for logical verification, and finally time-series stability rules are executed for time series analysis. Simultaneously, execution chains are constructed based on the input-output relationships between rules. For instance, the structural integrity verification result serves as an input constraint for consistency verification, thus forming a multi-dimensional rule execution structure. For example, in e-commerce scenarios, the payment consistency verification stage only begins when the order fields are complete. By constructing this multi-dimensional rule execution structure, the system achieves... The current rule-based execution process, with its streamlined and hierarchical management, improves the efficiency and logical consistency of the evaluation process. During structural integrity evaluation, target data fields undergo missing data detection, type detection, and format detection; for example, determining if the order number is empty or if the amount field is a numeric type. During consistency evaluation, cross-field or cross-data source data are logically compared; for example, determining if the order amount matches the payment amount. During temporal stability evaluation, time fields are sorted and validated, and time differences are calculated; for example, determining if the payment time is earlier than the order creation time. The evaluation results for each dimension are output as a score and pass / fail or anomaly markers, thus forming structural integrity results, consistency results, and temporal stability results.This system enables multi-dimensional independent quality assessment of target data, allowing for a comprehensive characterization of data quality from structural, logical, and temporal perspectives. This improves the completeness and accuracy of the assessment results. The system standardizes the assessment results across all dimensions, for example, normalizing scores from different dimensions to a unified range (e.g., 0-1) and encapsulating them in a structured manner according to preset weights or equal-weighting strategies. This results in a multi-dimensional assessment vector or a multi-dimensional assessment result structure. For instance, an order's data might ultimately be represented as a multi-dimensional assessment result of "structural integrity score 0.95, consistency score 0.88, and temporal stability score 0.90." By uniformly summarizing these multi-dimensional assessment results, different dimensions can be expressed within a unified framework.

[0031] In a preferred embodiment, a quality association topology is constructed based on the contextual association data of the target data to be evaluated and the target data to be evaluated, including: S201. Obtain context candidate data corresponding to the target business scenario based on the first data obtained from the target data to be evaluated; S202. Obtain node data based on the target data to be evaluated and the context candidate data; S203. Obtain the inter-node association identifier based on the data call information, data reference information, and data transmission information obtained from the node data; S204. Generate a quality association topology based on node data and node association identifiers, wherein the quality association topology includes nodes and associated edges.

[0032] As described in steps S201 to S204 above, the unique identifier of the target data to be evaluated is used as an index, such as order ID, transaction serial number, or device number. The corresponding core data records are retrieved from the business database, data warehouse, or real-time data stream to form the first data. For example, in an e-commerce scenario, if the target data to be evaluated is an order record, the first data includes the order master table information, payment record summary, and status transition record for that order. In a financial scenario, it includes the transaction master record and account change information. Based on the first data, the scope of related queries is further expanded, for example, through foreign key associations, API call records, or data lineage queries, to obtain contextual candidate data related to the target business scenario, such as the payment logs corresponding to the order. Inventory change records and logistics information, etc., form candidate context data, realizing a multi-level data expansion and acquisition mechanism starting from core business data. This extends data acquisition beyond single-point data to include upstream and downstream business-related data. The primary data and context candidate data undergo unified encoding processing. For example, order data, payment data, inventory data, and logistics data are encoded as node identifiers, and their unique attribute fields are extracted as node feature vectors, thus forming standardized node data. For instance, in e-commerce, "order node," "payment node," "inventory node," and "logistics node" are all abstracted as independent node data, with corresponding attribute information such as timestamps, status values, and data source identifiers. This allows for the integration of heterogeneous business data... The data is uniformly abstracted into standardized node data, enabling different types of data to be expressed within the same structural system. Data call information is obtained by analyzing interface call logs or service call chains, such as records of order services calling payment services. Data reference information is obtained by analyzing database foreign key relationships or field reference relationships, such as user IDs in the order table referencing user tables. Data transmission information is obtained by analyzing message queues, data synchronization tasks, or ETL processes, such as inventory changes being transmitted to the inventory service via message queues. Based on this information, inter-node association identifiers are constructed. For example, "order node, payment node" is marked as a call association, "order node, user node" as a reference association, and "inventory node, logistics node" as a reference association. This is used to represent the transitive relationships, forming structured association identifiers. By integrating three types of information—calling, referencing, and transitive—a multi-dimensional characterization of data relationships is achieved, making the expression of relationships between nodes more complete and accurate. Node data is used as nodes in the quality association topology, and the association identifiers between nodes are used as the association edges. Calling, referencing, and transitive relationships are mapped to different types of directed or weighted edges, thus constructing a complete quality association topology. For example, in an e-commerce scenario, "order node, payment node" is the calling edge, "order node, user node" is the referencing edge, and "inventory node, logistics node" is the transitive edge. Different weight values ​​are assigned according to the association strength, thus forming a weighted directed graph structure.The resulting quality association topology is used to express the association topology and flow paths between business data. By structuring business data relationships into a graph model, the associations between data become computable and analyzable, significantly improving the ability to express data associations in complex business scenarios.

[0033] Furthermore, in this application, the quality association topology is used to describe the data flow relationships, quality assessment result distribution, and quality anomaly propagation paths between different business data objects in complex business scenarios. It is not merely used to represent static associations between ordinary data, but rather to uniformly organize business nodes, association edges, multi-dimensional quality assessment results, and path impact parameters. Taking an e-commerce transaction business scenario as an example, the target business scenario includes multiple business processes such as order processing, payment settlement, inventory deduction, and logistics delivery. Business data objects are obtained based on the target data to be evaluated, and order data, payment data, inventory data, and logistics data are respectively used as quality association nodes. For example, order data nodes include order number, user ID, etc. The data structure includes fields such as order number, product number, order amount, order status, and order time. Payment data nodes include fields such as payment number, order number, payment amount, payment status, and payment time. Inventory data nodes include fields such as product number, inventory quantity, inventory deduction status, and update time. Logistics data nodes include fields such as logistics number, delivery status, and receipt time. Connections between nodes are established based on data calls, references, and transmission between business systems. For example, if an order node transmits order amount and order status information to a payment node, a positive connection is established between the order node and the payment node; if a payment node transmits payment success status to an inventory node, a connection is established between the payment node and the inventory node. Associative edges are established between inventory nodes and logistics nodes to transmit inventory confirmation status, forming a quality association topology. Nodes store business data objects and their corresponding multi-dimensional initial quality assessment results, while the association edges describe the transmission direction and impact paths between business data. For example, if the initial quality assessment results for the order node are 95 for structural integrity, 92 for consistency, and 90 for temporal stability; for the payment node, 90 for structural integrity, 85 for consistency, and 82 for temporal stability; and for the inventory node, 88 for structural integrity, 83 for consistency, and 80 for temporal stability, then all quality assessment results are considered consistent. Storing node attributes within the quality association topology allows for direct retrieval of quality changes across different business processes based on the association paths between nodes. In the specific storage process, the quality association topology uses node information and association edge information for storage. Node information includes node identifier, node name, business type, data field information, multi-dimensional quality assessment results, data update frequency, and historical quality status. For example, an order node can be stored as "Node ID N001, Node Name: Order Data, Business Type: Transaction Business, Structural Integrity: 95, Consistency: 92, Temporal Stability: 90, Daily Data Volume: 50,000 records". Association edge information describes the data transfer between two nodes.This includes information such as the associated edge number, starting node, target node, transmitted fields, association type, transmission frequency, business importance, and path weight. For example, the associated edge from the order node to the payment node can be recorded as "starting node N001, target node N002, transmitted fields: order amount and order status, transmission frequency: 50,000 times / day, business importance level: 5, path weight: 0.90". During the traversal of the quality association topology, the starting node is mainly based on the quality anomaly node, and the path search is performed on the business nodes according to the direction of the associated edge. For example, when it is found that the consistency evaluation result of the order node is 92 points, while the consistency evaluation result of the payment node drops to 85 points, it is determined that there is a 7-point gap between the order node and the payment node. The system identifies consistency differences and traverses along associated edges starting from the order node. It discovers that the anomaly might further affect the inventory node, thus forming a quality anomaly propagation path of "order node, payment node, inventory node." During the traversal, not only is the node access order recorded, but the path weight corresponding to each associated edge is also obtained to determine the importance of different propagation paths. In assigning path weights, factors such as data transmission frequency, business importance, node correlation, and historical anomaly propagation are comprehensively considered. For example, the order node to payment node transmits 50,000 data entries daily, involving core transaction fields such as order amount and payment status; therefore, a higher initial path weight is assigned, such as 0.90. The inventory node transmits approximately 30,000 data entries daily, involving inventory deduction status; therefore, it is assigned a path weight of 0.75. The inventory node to logistics node transmits approximately 20,000 data entries daily, with a relatively low business impact; therefore, it is assigned a path weight of 0.60. Furthermore, the path weights are adjusted based on historical anomaly propagation data. For example, if an order data anomaly is detected historically, there is an 80% probability that it will lead to payment data anomalies; therefore, the path weight from the order node to the payment node is increased from 0.90 to 0.98. Conversely, an inventory node anomaly only has a 30% probability of affecting the logistics node; therefore, the inventory to logistics path weight is adjusted from 0.60 to 0.50. This ensures that the path weights reflect the quality issues in the actual business scenario. The degree of impact is determined when the quality association topology is used for conflict propagation intensity analysis. The impact of quality conflict propagation is determined based on the difference data and path weights. For example, if the consistency difference between the order node and the payment node is 7 points, corresponding to a path weight of 0.98, the conflict propagation intensity is obtained by combining the difference value and the path weight: 7 multiplied by 0.98, resulting in a propagation intensity value of 6.86. Similarly, if the consistency difference between the payment node and the inventory node is 5 points, corresponding to a path weight of 0.80, the propagation intensity value is 5 multiplied by 0.80, resulting in a propagation intensity value of 4. Further summarizing the propagation intensity of each stage in the complete propagation path of order, payment, and inventory yields a path conflict intensity value of 10.86. When the preset conflict judgment threshold is 8...Because the conflict intensity of this path (10.86) exceeds the threshold, orders, payments, and inventory are identified as the dominant propagation path for quality conflicts. This path serves as a crucial basis for dynamic weight adjustment and quality risk localization. By constructing a quality-related topology, traditional quality inspection methods targeting single data tables or fields can be extended to business-link-level quality analysis. This not only allows for the acquisition of the data's inherent quality status but also enables the analysis of the location, propagation direction, and impact scope of quality issues by considering the data flow relationships between business nodes. This achieves accurate identification, dynamic assessment, and risk localization of data quality issues in complex business scenarios, improving the accuracy, interpretability, and business adaptability of automated data quality assessment.

[0034] In a preferred embodiment, difference data of multidimensional initial evaluation results are obtained based on the quality correlation topology, and conflict propagation paths are obtained based on the difference data, including: S301. Obtain the multidimensional initial evaluation results corresponding to the node data based on the quality association topology, and perform dimension alignment processing on the multidimensional initial evaluation results to form a multidimensional evaluation result sequence. S302. Obtain difference data from adjacent nodes in the multidimensional evaluation result sequence and quality correlation topology. S303. Obtain the difference propagation identifier based on the correlation edges of the difference data and the quality correlation topology; S304. Obtain path weights based on the differential propagation identifier and the quality-related topology; S305. Obtain the conflict propagation intensity value based on the difference data and path weights; S306. Obtain the path conflict intensity based on the conflict propagation intensity value and the quality association topology; S307. Obtain the dominant path of conflict propagation based on the path conflict intensity and preset threshold conditions, and obtain the conflict propagation path between dimensional quality based on the dominant path.

[0035] As described in steps S301 to S307 above, corresponding data nodes are obtained based on the node identifiers in the quality association topology, such as order nodes, payment nodes, inventory nodes, and logistics nodes. The evaluation results for structural integrity, consistency, and temporal stability obtained by each node in the aforementioned multi-dimensional initial quality assessment stage are then read. Since different business systems may use different quality evaluation methods—for example, the order system may use a percentage-based evaluation, while the payment system may use a grade-based evaluation—it is necessary to perform dimensional alignment processing on the evaluation results of different nodes, converting the quality results in different formats into a unified data expression form. For example, the initial evaluation results for the order node are: structural integrity 95 points, consistency 92 points, and temporal stability 90 points. The initial assessment results for payment nodes are structural integrity 0.90, consistency 0.85, and temporal stability 0.82. These are then converted to percentages, resulting in "95%, 92%, 90%" for order nodes and "90%, 85%, 82%" for payment nodes. A multi-dimensional assessment result sequence is then formed according to the format of "node number, quality dimension, assessment value." For example, the order node sequence is (structural integrity 95, consistency 92, temporal stability 90), and the payment node sequence is (structural integrity 90, consistency 85, temporal stability 82). This provides a unified basis for comparing the quality assessment results of different business nodes, avoiding biased quality comparisons due to different evaluation standards across business systems. This allows the difference identification process to be based on a unified quality scale, improving the accuracy of cross-system quality analysis in complex business scenarios. Adjacent business nodes are determined based on the association edges in the quality association topology; for example, order nodes and payment nodes are directly associated, as are payment nodes and inventory nodes. The multi-dimensional evaluation result sequences corresponding to adjacent nodes are read, and differences are compared for the same quality dimensions. Difference data is obtained by comparing the evaluation results of two nodes in terms of structural integrity, consistency, and temporal stability. For example, if the consistency evaluation result of the order node is 92% and the consistency evaluation result of the payment node is 85%, the consistency difference is obtained by subtracting the evaluation values ​​of the two nodes in the same dimension. For example, if the time-series stability of an order node is 90% and that of a payment node is 82%, then the time-series stability difference data is 8%. When differences exist across multiple dimensions, these differences are combined to form a set of inter-node difference data. For instance, between order nodes and payment nodes, a difference dataset with "5% difference in structural integrity, 7% difference in consistency, and 8% difference in time-series stability" can be created. By combining this difference data with node association information in the quality association topology, we can focus on node differences that may affect data quality transmission in the actual business chain, rather than judging whether a single data object is abnormal. This improves the targeting of quality anomaly detection. Based on the difference data, combined with the association edge information in the quality association topology...The source node, target node, and transmission fields corresponding to the discrepancies are identified. For example, if a 7% consistency difference is found between the order node and the payment node during the transmission of order amount information from the order node to the payment node, a corresponding difference propagation identifier is generated based on the association edges between the order node and the payment node. This identifier can include the propagation starting node, the propagation target node, the associated data field, and the quality dimension to which the difference belongs. For example, a propagation identifier of "order node, payment node, order amount field, consistency difference" is generated. If a 5% difference is also found between the payment node and the inventory node, a propagation identifier of "payment node, inventory node, payment status field, consistency difference" is generated. This identifier clearly identifies the source of the quality difference. This approach, along with identifying potentially affected business nodes, ensures that quality anomalies are no longer isolated differences but are linked to the business data flow path. The system locates the corresponding data relationship path based on the difference propagation identifier and determines the path weight by considering factors such as the importance of the business within the path, data transmission frequency, number of nodes, and scope of impact. For example, the propagation path from the order node to the payment node is considered highly important due to its involvement of transaction amount and payment status, and the daily data transmission volume reaches 50,000 records, thus assigning a path weight of 0.90. Similarly, the path from the payment node to the inventory node, primarily involving inventory deduction status and transmitting approximately 30,000 records daily, is assigned a path weight of 0.75. Finally, the path from the inventory node to the logistics node... Since the node path only affects delivery arrangements, and the daily data transmission volume is approximately 20,000 records, a path weight of 0.60 is assigned. When multiple factors jointly influence the path weight, such as business importance having a major impact, the weight value of business-important paths is prioritized to increase. This ensures that quality anomalies in the core business chain can be identified first. By introducing path weights, the impact of different data propagation paths on overall quality can be differentiated, allowing for focused attention on anomaly propagation paths with a large impact and high business value, thus improving the accuracy of risk identification. Combining discrepancies with path weights, a conflict propagation strength value is obtained by comprehensively processing the degree of discrepancy and path importance. For example, if the consistency difference between the order node and the payment node is 7%, the corresponding path weight... If the weight is 0.90, then by multiplying the difference value by the weight, we get a conflict propagation strength value of approximately 6.3%. If the consistency difference between the payment node and the inventory node is 5%, corresponding to a path weight of 0.75, then we get a conflict propagation strength value of approximately 3.75%. This shows that even if two paths have the same degree of data difference, the final propagation strength will differ due to the different path weights. By combining the degree of difference and the degree of path impact to obtain the conflict propagation strength, we can distinguish between ordinary data fluctuations and data quality conflicts with business impact, improving the reliability of anomaly detection. The complete propagation path is obtained based on the quality association topology. For example, the order, payment, and inventory path includes three data nodes: order node, payment node, and inventory node.The conflict propagation intensity between nodes in the path is summarized and comprehensively adjusted based on the path length, the number of associated nodes, and the number of abnormal nodes in the path. For example, if the order / payment path generates a conflict propagation intensity of 6.3% and the payment / inventory path generates a conflict propagation intensity of 3.75%, then the conflict intensity of the complete propagation path can be accumulated from the intensity of the two path segments to approximately 10.05%. If the path contains multiple abnormal nodes, the path conflict intensity is further increased to reflect the continuous spread of quality issues along the business chain. This can identify the spread trend of quality anomalies and avoid only discovering local anomalies while ignoring their impact on business processes. A conflict intensity judgment threshold is pre-set, for example, 8%. When the path conflict intensity of a propagation path exceeds 8%, the path is considered to have significant quality conflict. The propagation characteristics are analyzed. For example, the conflict intensity of the order and payment paths is 10.05%, exceeding a preset threshold. Therefore, this path is identified as the dominant path for conflict propagation. The conflict intensity of the inventory and logistics paths is 4%, below the threshold, and is therefore not considered a key propagation path. After identifying the dominant path, the quality dimensions involved in that path are further analyzed. For example, the main difference between order and payment nodes is consistency, and the main difference between payment and inventory nodes is time-series stability. This forms a quality conflict propagation path of "order consistency anomaly, payment consistency anomaly, and inventory time-series anomaly." By setting a path conflict intensity judgment mechanism, key quality anomaly propagation links can be filtered from complex business data relationships, realizing the transformation from discovering data anomalies to locating the causes of anomaly propagation, and improving the depth and accuracy of data quality assessment in complex business scenarios.

[0036] For example, the order node corresponds to (0.95, 0.90, 0.92), and the payment node corresponds to (0.88, 0.86, 0.91). Dimensional alignment is performed on the evaluation results of different nodes, unifying the evaluation dimensions of each node to the same standard dimensions, such as "structural dimension, consistency dimension, and temporal dimension." Missing dimensions are filled in or interpolated to make the evaluation results of different nodes comparable within the same dimensional space. For example, if a node lacks a temporal dimension, it is filled in with the default value or historical average. The nodes are arranged according to the graph topology or business flow order to form a multi-dimensional evaluation result sequence. For example, the corresponding evaluation vectors are arranged according to the path order of "order, payment, inventory, logistics," achieving cross-dimensional alignment. The unified alignment and serialization of nodes and multi-dimensional evaluation results, within the quality association topology, defines adjacent nodes as those directly related in business processes or data flows, such as order nodes and payment nodes, or payment nodes and inventory nodes. The multi-dimensional evaluation result sequences corresponding to adjacent nodes are calculated dimension-by-dimensionally, including structural integrity differences, consistency differences, and time-series stability differences, thus yielding difference data. For example, the difference vector (0.07, 0.04, 0.01) between order nodes (0.95, 0.90, 0.92) and payment nodes (0.88, 0.86, 0.91) is used to characterize the degree of data quality deviation between different business nodes. This is achieved through graph-based adjacency... The structure extracts differences, giving the difference analysis semantic support for business flow. Related edges include call edges, reference edges, and propagation edges; different types of related edges correspond to different propagation semantics. For example, call edges represent strong propagation paths, reference edges represent medium propagation paths, and propagation edges represent weak propagation paths. By combining and encoding difference data with related edges (e.g., "difference source node, target node, association type, difference dimension"), a difference propagation identifier is generated. This identifier identifies the direction and scope of difference propagation in the business chain, transforming abstract differences into traceable propagation identifiers, giving differences path-based expression capabilities. Path weights are composed of association type weights and business importance weights; for example, call edges have higher weights than reference edges, and reference edges have higher weights than propagation edges. The weighting of paths is dynamically adjusted based on business priority or historical impact. For example, the weight of order and payment paths is set to 0.9, while that of payment and inventory is 0.8. By weighting paths in the graph, different propagation paths have different impact intensities. The path weighting mechanism characterizes the differences in importance of different business relationships, thereby improving the granularity of difference propagation analysis. The difference values ​​of each dimension in the difference data are weighted. For example, structural differences, consistency differences, and time series differences are multiplied by their corresponding path weights, and then summed or normalized to obtain the conflict propagation intensity value. For example, in the order and payment paths, the difference vector (0.07, 0.04, 0.01) combined with a weight of 0.9 yields a propagation intensity of 0.108. Static differences are transformed into propagation quantification values ​​that consider the strength of business relevance, enabling conflict expression to have propagation impact rather than merely representing local differences. For multiple propagation intensity values ​​on the same path, they are accumulated sequentially according to the path, while introducing attenuation factors, such as gradual attenuation with increasing path depth, to avoid excessive amplification of the impact of remote nodes. For example, in the path "order, payment, inventory, logistics," the propagation intensity of each segment is calculated separately, accumulated, and multiplied by an attenuation coefficient to form the path conflict intensity value. Through a path-level aggregation mechanism, local differences are expanded into full-link conflict intensity, thereby achieving the ability to characterize the overall quality risk of the business link. Conflict propagation intensity is obtained based on path conflict intensity and preset threshold conditions. The system identifies the dominant path for conflict propagation, with preset threshold conditions based on historical data statistics or empirical rules. For example, high-risk, medium-risk, and low-risk thresholds are derived through quantile statistics. The path conflict intensity is compared to these threshold conditions; if the path conflict intensity exceeds the high threshold, the path is determined to be the dominant path for conflict propagation. For instance, the "orders, payments, inventory" path is identified as the dominant path in cases of frequent payment anomalies. The system selects the paths with the highest conflict intensity or those exceeding the threshold from all paths and outputs them as the conflict propagation paths between dimensions of quality. By introducing a threshold constraint mechanism to filter propagation paths, the system can automatically identify key risk propagation links, thereby improving the ability to locate data quality risks.

[0037] In a preferred embodiment, evaluation weights are obtained based on the discrepancy data and the initial multidimensional evaluation results. The discrepancy data is then fused according to these evaluation weights to obtain a fused evaluation result for the quality of each dimension, including: S401. Obtain dimension-aligned difference units based on the difference data and the multidimensional initial assessment results; S402. Obtain the dimensional difference influence factor based on the dimensional alignment difference unit and the multidimensional initial evaluation results; S403. Obtain path constraint factors based on dimensional difference influencing factors and conflict propagation paths in the quality-related topology. S404. Obtain the initial evaluation weights based on the dimensional difference influencing factors and the path constraint factors; S405. Obtain dynamic evaluation weights based on initial evaluation weights and conflict propagation paths; S406. Obtain the path-weighted difference results based on the dynamic evaluation weights and difference data; S407. Obtain the fusion evaluation results of the quality of each dimension based on the path weighted difference results and the multidimensional initial evaluation results.

[0038] As described in steps S401 to S407 above, the initial multi-dimensional evaluation results corresponding to the interconnected nodes in the quality association topology are read. For example, the initial evaluation results of the order node, payment node, and inventory node correspond to the three quality dimensions of structural integrity, consistency, and temporal stability, respectively. Assuming the initial evaluation results of the order node are 95 points for structural integrity, 92 points for consistency, and 90 points for temporal stability, and the initial evaluation results of the payment node are 90 points for structural integrity, 85 points for consistency, and 82 points for temporal stability, the same quality dimensions are matched according to the difference data between the two nodes, forming dimension-aligned difference units. For example, the difference value corresponding to the structural integrity dimension is 5 points, the difference value corresponding to the consistency dimension is 7 points, and the difference value corresponding to the temporal stability dimension is 7 points. If the difference value for a dimension is 8, then a dimension-aligned difference unit is formed: "Structural integrity 95, 90, difference 5; Consistency 92, 85, difference 7; Temporal stability 90, 82, difference 8". This allows quality changes between different nodes to be described according to a unified dimension, achieving a precise correspondence between the initial quality assessment results and the abnormal difference data. It enables the analysis of the degree of impact of anomalies on specific quality dimensions, improving the accuracy of the quality fusion assessment process. Based on the difference magnitude in the dimension-aligned difference unit and combined with the initial assessment level of the corresponding dimension, a dimension difference impact factor is obtained. This impact factor represents the degree of impact of a change in a certain quality dimension on the overall quality result. For example, the consistency difference between order nodes and payment nodes. The difference score is 7, the initial assessment score for order node consistency is 92, and the initial assessment score for payment node consistency is 85, indicating a significant downward trend in this dimension. Therefore, the impact factor for this dimension is determined to be high based on the magnitude of the difference, for example, set to 0.75. The structural integrity difference is only 5, and both nodes maintain a high level of integrity, so the structural integrity impact factor is determined to be 0.45. The temporal stability difference is 8, and since this dimension directly affects the business processing order, the impact factor is determined to be 0.85. Scoring is based on the degree of difference; for example, the larger the difference, the closer the impact factor is to 1, and the smaller the difference, the lower the impact factor. Adjustments are also made based on the importance of quality in the business scenario, so that different quality dimensions are no longer fixed. The weighting of the data allows for dynamic adjustment of the impact based on actual quality changes, improving the adaptability of quality assessment results in complex business environments. It reads the conflict propagation paths identified in the quality-related topology, such as order nodes, payment nodes, and inventory nodes. In these paths, order data affects payment status, which in turn affects inventory deduction; therefore, these paths have strong business relevance. Path constraint factors are determined based on information such as the number of nodes in the path, the degree of node relevance, and historical anomaly propagation. For example, the order and payment paths involve the transmission of core transaction data, resulting in a large business impact; therefore, the path constraint factor is set to 0.90. The inventory and logistics paths, which only affect delivery arrangements, have a smaller impact; therefore, the path constraint factor is set to 0.60. When a difference in a quality dimension is on a high-impact propagation path, the corresponding constraint factor is increased. For example, if the consistency dimension impact factor is 0.75, its corresponding path constraint factor is 0.90, indicating that the consistency problem may spread along the core business chain and requires increased attention. If a difference is on a low-impact path, the constraint factor is decreased. This ensures that quality assessment considers not only the degree of anomaly of a single data node but also the propagation impact of anomalies in the business process, improving the comprehensiveness of quality risk identification. The dimension difference impact factor and the path constraint factor are combined and multiplied to obtain the basic weight value. The weights of multiple dimensions are then normalized to ensure that the weights of different dimensions are consistent. Within a unified evaluation scope, for example, the consistency dimension has a difference impact factor of 0.75 and a path constraint factor of 0.90, so multiplying them gives an initial weight of 0.675. The structural integrity dimension has an impact factor of 0.45 and a path constraint factor of 0.70, resulting in an initial weight of 0.315. The temporal stability dimension has an impact factor of 0.85 and a path constraint factor of 0.80, resulting in an initial weight of 0.68. These weights are then proportionally adjusted; for example, if the total weights for the three dimensions are 1.67, they are converted to approximately 0.40, 0.19, and 0.41 respectively, avoiding the inability to adapt to complex business changes caused by the fixed weights used in traditional data quality assessments. The problem is that the initial weights are adjusted based on the propagation intensity, path length, and number of anomalous nodes in the conflict propagation path. For example, the initial weight for the consistency dimension is 0.40, and its conflict propagation path is order, payment, and inventory. This path passes through three key business nodes, and the aforementioned conflict propagation intensity is 6.3%, indicating a high propagation impact. Therefore, the weight of this dimension is increased, for example, to 0.48. The initial weight for the structural integrity dimension is 0.19. Since there is only a single node anomalous and no obvious propagation path, it is maintained or reduced to 0.18. The initial weight for the temporal stability dimension is 0.41. Due to the general impact range, it is adjusted to 0.34. This is achieved through dynamic... The weights are adjusted to reflect the actual risk status in the current business environment, rather than remaining fixed. This allows data quality assessment to detect abnormal propagation trends, improves the sensitivity of evaluation for potentially spreading quality issues, and enhances risk warning capabilities. The difference data for each quality dimension and its corresponding dynamic weights are read separately, and the weighted impact value is calculated by multiplication. For example, if the consistency dimension difference is 7 points and the dynamic assessment weight is 0.48, then multiplying the difference value by the weight yields a path-weighted difference result of approximately 3.36 points. If the structural integrity difference is 5 points and the dynamic weight is 0.18, then the result is 0.9 points. If the temporal stability difference is 8 points and the dynamic weight is 0.34, then the result is 2.The score is 72. After weighted processing, it can be found that although the original difference in temporal stability is the largest, the consistency dimension has a higher final impact because it is located in the core business propagation path. This allows the quality anomaly analysis results to comprehensively reflect both the magnitude of the difference and the scope of business impact, improving the authenticity of the quality assessment results. Initial assessment results for each quality dimension are obtained and then adjusted based on the path-weighted difference results. For example, if the initial assessment result for the consistency dimension is 92 points and the path-weighted difference result is 3.36 points, the initial result is corrected according to the direction of the difference's impact, resulting in a fused assessment result of approximately 88.64 points. The initial assessment score for the integrity dimension is 95 points, and the weighted difference score is 0.9 points, resulting in approximately 94.1 points. The initial assessment score for the temporal stability dimension is 90 points, and the weighted difference score is 2.72 points, resulting in approximately 87.28 points. This forms a fusion assessment result for each dimension, such as structural integrity (94.1 points), consistency (88.64 points), and temporal stability (87.28 points). This ensures that the final quality evaluation result not only reflects the current data status but also demonstrates the potential impact of the data in complex business processes, achieving a shift from static quality inspection to dynamic quality assessment and improving the accuracy and reliability of data quality assessment in complex business scenarios.

[0039] In a preferred embodiment, path structure information is obtained based on the quality association topology, and a comprehensive quality score for the target data to be evaluated is obtained based on the fusion evaluation results and the path structure information. A quality level threshold determination result is obtained based on the comprehensive quality score, and the corresponding quality level and quality risk location information are output based on the quality level threshold determination result, including: S501. Obtain path structure information based on the associated edges of the quality-related topology; S502. Obtain path weight information based on path structure information and quality-related topology; S503. Obtain the path weighted scoring factor based on the fusion evaluation results and path weight information; S504. Obtain the comprehensive quality score of the target data to be evaluated based on the path-weighted scoring factors; S505. Obtain the quality level threshold determination result based on the comprehensive quality score and the preset quality threshold. S506. Obtain risk path nodes based on the quality level threshold determination results and path structure information, and obtain quality level and quality risk location information based on the risk path nodes.

[0040] As described in steps S501 to S506 above, the quality association topology consists of nodes and associated edges. By traversing the associated edges, for example using depth-first search or breadth-first search, starting from the initial node (e.g., the order node), its reachable nodes (e.g., payment node, inventory node, and logistics node) are identified sequentially, thereby generating a complete business path, such as "order, payment, inventory, logistics". The path is then expressed in a structured way, including information such as path length, node order, and path type, thus forming path structure information. This transforms the scattered associated edge information into a complete business path structure, elevating data relationships from local connections to a global path structure expression. The path type is identified based on the path structure information, for example, by calling... The system identifies paths, referencing paths, and transit paths, assigning basic weight values ​​to different path types. For example, the call path has the highest weight, followed by the referencing path, and the transit path has the lowest. Weights are adjusted based on path length; for instance, the longer the path, the lower its weight, reflecting the decreasing influence of remote nodes. Path weights are also adjusted based on node importance; for example, order and payment nodes have higher weights than log and auxiliary nodes. The path "order, payment, inventory" receives a weight of 0.85. This multi-factor fusion approach constructs a path weight system that ensures path importance accurately reflects the strength of business connections, thereby improving the rationality of scoring calculations and business consistency. The system also includes structural integrity scoring, consistency scoring, and time series for each node. Stability scores are uniformly aggregated and multiplied by the corresponding path weights to form a weighted score. The influence weight of each data propagation path is determined based on the path structure information in the quality association topology. For example, in an e-commerce order business scenario, three main data propagation paths are obtained: "order, payment," "payment, inventory," and "inventory, logistics." The corresponding path weights are determined to be 0.5, 0.3, and 0.2 respectively, based on business importance, data transmission frequency, and node influence range. The data quality status of each path is determined based on the fusion evaluation results. For example, the fusion evaluation result for the "order, payment" path is 0.90, the "payment, inventory" path is 0.80, and the "inventory, logistics" path is 0.2. The estimated result is 0.70. The path weights are correlated with the corresponding fusion evaluation results (i.e., the path weights are multiplied by the corresponding fusion evaluation results) to obtain the path weighted scoring factors for each path. The weighted scoring factor for the "order, payment" path is 0.45, the weighted scoring factor for the "payment, inventory" path is 0.24, and the weighted scoring factor for the "inventory, logistics" path is 0.14. In the process of summarizing the multi-path scores, the weighted scoring factors of each path are summarized to obtain the comprehensive quality score. Specifically, the weighted scoring factors of each path are accumulated to obtain a comprehensive quality score of 0.83. The preset quality level threshold is set as follows: a score greater than 0.90 is excellent, 0.80 to 0.90 is good, and less than 0.90 is good.When 80 represents risk, the preset quality threshold is determined based on historical data statistics or empirical rules. This determines the overall quality level of the target data to be evaluated as "Good." For example, if an anomaly exists in the payment amount synchronization data in the "Payment, Inventory" path, causing the path's integrated evaluation result to decrease from 0.80 to 0.50, the path weighted scoring factor decreases from 0.24 to 0.15 due to the path's high business impact weight, thus lowering the overall comprehensive quality score to 0.74. In this case, the system determines the data quality level as "Risk" and, based on the path structure information, identifies the payment node in the "Payment, Inventory" path as the primary risk-affecting node. This allows the comprehensive quality score to not only reflect the evaluation results of each quality dimension but also to be dynamically adjusted based on the degree of data propagation and impact in the business process. This improves the accuracy of data quality assessment and risk identification capabilities in complex business scenarios. The quality score is matched against a threshold range to determine the current data quality level. For example, a score of 0.90 corresponds to an excellent level. By introducing a standardized threshold grading mechanism, interpretable grading of data quality results is achieved, giving the evaluation results clear business semantics and management significance. When a quality level falls below a preset threshold, reverse tracing is performed on each path node based on path structure information to identify the key nodes with the greatest impact on the overall score. For example, risk nodes are identified through contribution analysis or anomaly propagation strength analysis, such as payment node anomalies or inventory node delays. Based on the location of the risk node in the path, risk location information is output, such as "inconsistency anomalies exist in order and payment paths." Combining the distribution of risk nodes, a quality level and risk location result are generated. The path-level reverse tracing mechanism enables the location of quality problems, further penetrating from the overall score to specific nodes and paths, achieving interpretable and localizable data quality analysis capabilities.

[0041] Please see the appendix Figure 2 As shown, the present invention also provides an automatic data quality assessment system for complex business scenarios, used in the aforementioned automatic data quality assessment method for complex business scenarios, comprising: The multidimensional evaluation module performs a multidimensional initial quality evaluation on the target data to be evaluated obtained from the target business scenario based on the acquired initial quality evaluation rules, and obtains the multidimensional initial evaluation results. The data relationship module constructs a quality relationship topology based on the contextual data of the target data to be evaluated and the target data to be evaluated. The conflict path module obtains the difference data of the multidimensional initial evaluation results based on the quality correlation topology, and obtains the conflict propagation path based on the difference data; The fusion evaluation module obtains evaluation weights based on the differential data and the initial multidimensional evaluation results, and then performs fusion processing on the differential data according to the evaluation weights to obtain the fusion evaluation results for the quality of each dimension. The evaluation results module obtains path structure information based on the quality association topology, and obtains a comprehensive quality score for the target data to be evaluated based on the fusion evaluation results and path structure information. It then obtains the quality level threshold determination result based on the comprehensive quality score and outputs the corresponding quality level and quality risk location information based on the quality level threshold determination result.

[0042] As described above, the multi-dimensional evaluation module obtains the target data to be evaluated from the business system, including order data, payment data, inventory data, and log data, and performs standardization and field alignment on these data. Based on the initial quality evaluation rules, it performs structural integrity checks, consistency logic checks, and time sequence stability checks on the target data. Structural integrity checks detect missing fields and format anomalies; consistency checks detect cross-field or cross-system data consistency; and time sequence stability checks detect time order and delay anomalies. The evaluation results from each dimension are summarized to form a multi-dimensional initial evaluation result, achieving a comprehensive assessment of data quality from a single dimension to multiple dimensions. This allows for a complete characterization of data quality status, improving the comprehensiveness and accuracy of the evaluation. (Data Relationship Module) Using the target data to be evaluated as the core node, contextual candidate data is obtained by expanding outward through foreign key associations, interface call records, and data flow logs. This includes data such as order payment records, inventory change records, and logistics status data. All data objects are abstracted into node data, and the call relationships, reference relationships, and transmission relationship information between nodes are extracted to form node association identifiers. Node data is used as graph nodes, and association identifiers are used as graph edges to construct a quality association topology. The graph is represented as a weighted directed graph structure, transforming business data from an isolated structure into a graph structure, making the relationships between data computable and traceable. The conflict path module performs dimensional alignment and dimension-by-dimensional difference calculation on the multidimensional evaluation results of adjacent nodes in the graph, thereby generating a difference... Heterogeneous data, such as structural differences, consistency differences, and time-series differences, is combined with graph association edges to generate difference propagation identifiers. Further, path weights and conflict propagation strengths are calculated based on path traversal to ultimately identify the dominant conflict propagation path, thereby determining key conflict propagation links. This extends local data quality differences to path-level propagation analysis, making data quality issues propagable and traceable, thus improving the ability to locate quality problems. The fusion evaluation module performs dimensional alignment processing on the heterogeneous data to form dimensional difference units, and calculates dimensional difference impact factors based on multidimensional initial evaluation results. Path constraint factors are introduced according to the conflict propagation paths in the quality association topology, and initial evaluation weights are generated based on these factors. The initial weights are dynamically adjusted to form dynamic evaluation weights, and path-weighted processing is applied to differential data. Finally, these weights are merged with the multi-dimensional initial evaluation results to generate a multi-dimensional fusion evaluation result. This achieves multi-factor fusion of differences, weights, and path constraints, enabling the evaluation result to simultaneously reflect both the quality of the data itself and its impact on business propagation, thus improving the robustness and accuracy of the evaluation. The evaluation result module extracts path structure information from the quality association topology, including path node order, path length, and association type, and further calculates path weight information. The fusion evaluation result is combined with the path weight information to calculate the path-weighted scoring factor, and a comprehensive quality score is generated through global aggregation. This comprehensive quality score is then matched with a preset quality threshold to determine the quality level.For example, it categorizes data into excellent, good, average, and poor levels. Based on path structure information and scoring contribution analysis, it identifies risky path nodes and outputs quality risk location information, such as indicating specific anomalies occurring in order or payment paths. This achieves a complete closed-loop analysis from global scoring to local node location, ensuring the data quality assessment results are comprehensive.

[0043] The above description is merely a preferred embodiment of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention. Structures, devices, and operating methods not specifically described or explained in this invention are implemented according to conventional methods in the art unless otherwise specified or limited.

Claims

1. An automatic data quality assessment method for complex business scenarios, characterized in that, include: Based on the obtained initial quality assessment rules, a multi-dimensional initial quality assessment is performed on the target data to be assessed obtained from the target business scenario to obtain multi-dimensional initial assessment results. Construct a quality correlation topology based on the contextual data of the target data to be evaluated and the target data to be evaluated. Based on the quality correlation topology, obtain the difference data of the multidimensional initial evaluation results, and obtain the conflict propagation path based on the difference data; The evaluation weights are obtained based on the differential data and the initial multidimensional evaluation results. The differential data are then fused based on the evaluation weights to obtain the fused evaluation results for the quality of each dimension. The path structure information is obtained based on the quality association topology, and the comprehensive quality score of the target data to be evaluated is obtained based on the fusion evaluation results and the path structure information. The quality level threshold determination result is obtained based on the comprehensive quality score, and the corresponding quality level and quality risk location information is output based on the quality level threshold determination result.

2. The automatic data quality assessment method for complex business scenarios according to claim 1, characterized in that, The methods for obtaining initial quality assessment rules include: Obtain rule mapping identifiers based on the acquired scenario type and business characteristics; Retrieve the corresponding rule template data from the pre-built rule library based on the rule mapping identifier; The rule template data is filtered according to the scenario type to obtain the target rule corresponding to the scenario type; Based on business characteristics and target rules, obtain the initial quality assessment rules corresponding to the target business scenario.

3. The automatic data quality assessment method for complex business scenarios according to claim 2, characterized in that, The methods for obtaining scenario types and business characteristics include: Obtain scenario description information based on the target business scenario, and combine it with a pre-built scenario semantic knowledge base to obtain the corresponding scenario semantic representation data; Match the scene semantic representation data with the scene semantic tags in the pre-built scene semantic knowledge base to obtain the scene type corresponding to the target business scene; Based on the scenario type and scenario semantic representation data, candidate business features corresponding to the scenario type are obtained from the pre-set business feature library; Obtain the degree of scene matching between scene semantic representation data and candidate business features, and filter the candidate business features to obtain the target business features; The scenario description information is structured and labeled according to the target business characteristics to obtain the business characteristics corresponding to the target business scenario.

4. The automatic data quality assessment method for complex business scenarios according to claim 1, characterized in that, Based on the obtained initial quality assessment rules, a multi-dimensional initial quality assessment is performed on the target data to be assessed obtained from the target business scenario, resulting in multi-dimensional initial assessment results, including: Obtain the target data to be evaluated based on the target business scenario; Based on the target data to be evaluated and the initial quality assessment rules, obtain the structural integrity dimension rules, consistency dimension rules, and time series stability dimension rules; The multi-dimensional rule execution structure is obtained based on the structural integrity dimension rules, the consistency dimension rules, and the time sequence stability dimension rules. Based on the multidimensional rule execution structure, the target data to be evaluated is evaluated in multiple dimensions to obtain the initial evaluation results of structural integrity, consistency and time series stability. The initial assessment results of structural integrity, consistency, and time series stability are summarized to form a multidimensional initial assessment result.

5. The automatic data quality assessment method for complex business scenarios according to claim 4, characterized in that, The multidimensional rule execution structure includes structural integrity, consistency, and time sequence stability dimensions.

6. Construct a quality association topology based on the contextual data of the target data to be evaluated and the target data to be evaluated, including: Based on the first data obtained from the target data to be evaluated, obtain contextual candidate data corresponding to the target business scenario; Obtain node data based on the target data to be evaluated and the contextual candidate data; Based on the data call information, data reference information, and data transmission information obtained from the node data, the association identifier between nodes is obtained; A quality association topology is generated based on node data and the association identifiers between nodes. The quality association topology includes nodes and associated edges.

7. The automatic data quality assessment method for complex business scenarios according to claim 1, characterized in that, Based on the quality correlation topology, difference data of the multidimensional initial assessment results are obtained, and the conflict propagation path is obtained based on the difference data, including: Based on the quality correlation topology, obtain the multidimensional initial evaluation results corresponding to the node data, and perform dimension alignment processing on the multidimensional initial evaluation results to form a multidimensional evaluation result sequence. Difference data is obtained from adjacent nodes in the multidimensional evaluation result sequence and quality correlation topology; Obtain the difference propagation identifier based on the correlation edges of the difference data and the quality correlation topology; Path weights are obtained based on the differential propagation identifier and the quality-related topology. The conflict propagation intensity value is obtained based on the difference data and path weights; The path conflict intensity is obtained based on the conflict propagation intensity value and the quality-related topology. The dominant path of conflict propagation is obtained based on the path conflict intensity and preset threshold conditions, and the conflict propagation path between dimensional quality is obtained based on the dominant path.

8. The automatic data quality assessment method for complex business scenarios according to claim 1, characterized in that, Evaluation weights are obtained based on the discrepancy data and the initial multidimensional evaluation results. The discrepancy data is then fused according to these weights to obtain the fused evaluation results for the quality of each dimension, including: Obtain dimension-aligned difference units based on the difference data and the initial multidimensional assessment results; The dimensional difference influence factor is obtained based on the dimensional alignment difference unit and the multidimensional initial evaluation results. Path constraint factors are obtained based on dimensional difference influence factors and conflict propagation paths in the quality-related topology. Initial evaluation weights are obtained based on dimensional difference influencing factors and path constraint factors; Dynamic evaluation weights are obtained based on the initial evaluation weights and the conflict propagation path. The path-weighted difference results are obtained based on the dynamic evaluation weights and difference data. The fusion evaluation results of the quality of each dimension are obtained based on the path-weighted difference results and the multidimensional initial evaluation results.

9. The automatic data quality assessment method for complex business scenarios according to claim 1, characterized in that, Path structure information is obtained based on the quality association topology, and a comprehensive quality score for the target data to be evaluated is obtained based on the fusion evaluation results and path structure information. A quality level threshold determination result is obtained based on the comprehensive quality score, and the corresponding quality level and quality risk location information is output based on the quality level threshold determination result, including: Obtain path structure information based on the associated edges of the quality-related topology; Obtain path weight information based on path structure information and quality-related topology; Obtain path weighted scoring factors based on the fusion evaluation results and path weight information; The overall quality score of the target data to be evaluated is obtained based on the path-weighted scoring factors. The quality level threshold determination result is obtained based on the comprehensive quality score and the preset quality threshold. Risk path nodes are obtained based on the quality level threshold determination results and path structure information, and quality level and quality risk location information are obtained based on the risk path nodes.

10. An automatic data quality assessment system for complex business scenarios, applied to the automatic data quality assessment method for complex business scenarios as described in any one of claims 1 to 9, characterized in that, include: The multidimensional evaluation module performs a multidimensional initial quality evaluation on the target data to be evaluated obtained from the target business scenario based on the acquired initial quality evaluation rules, and obtains the multidimensional initial evaluation results. The data relationship module constructs a quality relationship topology based on the contextual data of the target data to be evaluated and the target data to be evaluated. The conflict path module obtains the difference data of the multidimensional initial evaluation results based on the quality correlation topology, and obtains the conflict propagation path based on the difference data; The fusion evaluation module obtains evaluation weights based on the differential data and the initial multidimensional evaluation results, and then performs fusion processing on the differential data according to the evaluation weights to obtain the fusion evaluation results for the quality of each dimension. The evaluation results module obtains path structure information based on the quality association topology, and obtains a comprehensive quality score for the target data to be evaluated based on the fusion evaluation results and path structure information. It then obtains the quality level threshold determination result based on the comprehensive quality score and outputs the corresponding quality level and quality risk location information based on the quality level threshold determination result.

Images