A business object label generation and optimization method and system based on a BI system

Abstract

The application relates to a business object label generation and optimization method and system based on a BI system, which comprises the following steps: extracting a multi-source data set of all business objects from a BI system; inputting the multi-source data set into a general label model to generate a scenario label; generating an object classification label set through a rule engine; fusing the scenario label and the object classification label set to form a comprehensive label set; and updating the comprehensive label set in real time according to business feedback data. The application combines a general label model and a rule engine, solves the problem that a traditional rule engine needs to be frequently manually updated, reduces maintenance cost, improves response speed, overcomes the defect that a static label system cannot reflect business dynamics in real time, and has the effects of improving the timeliness and accuracy of label generation and optimization.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method and system for generating and optimizing business object labels based on a BI system. Background Technology

[0002] In current business intelligence (BI) systems, the generation and optimization of business object tags primarily rely on traditional rule engines or static tagging systems. Existing technologies typically generate tags by manually defining fixed rules (such as user spending thresholds or number of promotional activity participations) or by batch-generating tags based on training general models using historical data. For example, the retail industry commonly uses the RFM model (recent purchase, purchase frequency, purchase amount) to classify customers, while the financial industry uses credit scoring models to identify high-risk customers.

[0003] However, these methods have significant drawbacks in practical applications: First, traditional rule engines require frequent manual updates to adapt to business changes, resulting in high maintenance costs and delayed responses; in addition, static tag systems cannot reflect business dynamics in real time (such as sudden changes in user behavior and fluctuations in market trends), leading to delayed tag updates and affecting the timeliness of decision-making. Summary of the Invention

[0004] To address the aforementioned shortcomings and improve the timeliness and accuracy of business tag generation and optimization in BI systems, this application provides a method and system for generating and optimizing business object tags based on BI systems.

[0005] The above-mentioned objective of this application is achieved through the following technical solution:

[0006] A method for generating and optimizing business object tags based on a BI system, comprising the following steps:

[0007] Extract multi-source datasets of all business objects from the BI system;

[0008] The multi-source dataset is input into a pre-trained general label model, which then generates contextualized labels for business objects based on the multi-source dataset.

[0009] The rule engine generates a set of object classification labels corresponding to business objects based on multi-source datasets;

[0010] The received contextualized tags and object classification tags associated with the same business object are merged to form a comprehensive tag set associated with that business object;

[0011] Receive business feedback data from the BI system in real time and update the comprehensive tag set based on the business feedback data.

[0012] By adopting the above technical solution, multi-source datasets of all business objects are extracted from the BI system and input into a pre-trained general label model. The general label model generates contextualized labels corresponding to the business objects based on the multi-source datasets. Simultaneously, a rule engine generates object classification label sets corresponding to the business objects based on the multi-source datasets. The contextualized labels and object classification label sets associated with the same business object are then merged to form a comprehensive label set associated with that business object. Business feedback data from the BI system is received in real time, and the comprehensive label set is updated based on the business feedback data, realizing the automated generation and dynamic optimization of business object labels. This application solves the problem of frequent manual rule updates required by traditional rule engines by combining a general label model and a rule engine, reducing maintenance costs and improving response speed. At the same time, by receiving business feedback data in real time and dynamically updating the comprehensive label set, the shortcomings of static label systems in reflecting business dynamics in real time are overcome, which improves the timeliness and accuracy of label generation and optimization, thereby enhancing the decision support capability and business adaptability of the BI system.

[0013] In a preferred embodiment, this application can be further configured as follows: the general label model includes a recognition layer, a data reuse layer, a feature extraction layer, and a label output layer; the step of inputting the multi-source dataset into the pre-trained general label model, enabling the general label model to generate contextualized labels corresponding to business objects based on the multi-source dataset, includes the following steps:

[0014] The recognition layer identifies the input multi-source dataset, determines the industry domain of the multi-source dataset based on the recognition results, and outputs an industry confidence matrix based on the multi-source dataset.

[0015] The data reuse layer obtains corresponding industry-specific information based on the industry confidence matrix;

[0016] The feature extraction layer extracts industry-wide features, vertical domain features, and business scenario features based on multi-source datasets and general industry information;

[0017] The label output layer outputs contextualized labels based on industry-wide features, vertical domain features, and business scenario features, using multi-source datasets.

[0018] By adopting the above technical solution, multi-source datasets are input into a pre-trained general label model. The recognition layer identifies the input multi-source datasets and determines the industry domain based on the recognition results, outputting an industry confidence matrix. The data reuse layer obtains corresponding industry general information based on the industry confidence matrix. The feature extraction layer extracts industry general features, vertical domain features, and business scenario features based on the multi-source datasets and industry general information. The label output layer outputs scenario-based labels based on the multi-source datasets based on industry general features, vertical domain features, and business scenario features, realizing the automated generation of scenario-based labels for business objects. This application solves the limitations of traditional label generation methods that rely on single rules or static models through layered processing and multi-dimensional feature extraction. It can more accurately capture the industry characteristics and scenario-based needs of business objects. At the same time, by introducing industry confidence matrices and industry general information, the accuracy and adaptability of label generation are improved, thereby providing high-quality scenario-based label support for BI systems and enhancing the accuracy and timeliness of business decisions.

[0019] In a preferred embodiment, this application can be further configured as follows: the step of generating a set of object classification labels corresponding to business objects based on a multi-source dataset using a rule engine includes the following steps:

[0020] Obtain the industry confidence matrix output by the recognition layer, and match the execution rules corresponding to the multi-source datasets based on the pre-stored multi-dimensional business rule set and the industry confidence matrix;

[0021] A rule execution chain is constructed based on the matching execution rules, thereby generating a tag inference chain;

[0022] The multi-source dataset is input into the label inference chain to generate the object classification label set corresponding to the business object.

[0023] By adopting the above technical solution, the industry confidence matrix output by the recognition layer is obtained. Based on the pre-stored multi-dimensional business rule set and the industry confidence matrix, the execution rules corresponding to the multi-source datasets are matched. Based on the matched execution rules, a rule execution chain is constructed to generate a label inference chain. The multi-source datasets are input into the label inference chain to generate the object classification label set corresponding to the business objects, thereby realizing the automated generation of the business object classification label set. This application solves the limitations of traditional rule engines that rely on single rules or static logic by combining the industry confidence matrix and the multi-dimensional business rule set. It can more flexibly adapt to the needs of different industries and business scenarios. At the same time, by constructing the rule execution chain and the label inference chain, the traceability and logical consistency of the label generation process are ensured, thereby providing more accurate and reliable object classification label support for the BI system and improving the scientificity and effectiveness of the BI system's label generation and optimization business decisions.

[0024] In a preferred embodiment, this application can be further configured as follows: the step of constructing a rule execution chain based on the matching execution rules to generate a label inference chain includes the following steps:

[0025] When constructing the rule execution chain, check whether there are rule conflicts between the matched execution rules;

[0026] When a rule conflict is detected, it is handled based on a pre-set conflict resolution strategy.

[0027] The matching execution rules are hierarchically classified, and a label inference chain is generated based on the classification results.

[0028] By adopting the above technical solution, the system detects whether there are rule conflicts between matching execution rules when constructing the rule execution chain. When a rule conflict is detected, it is processed based on a pre-set conflict resolution strategy, and the matching execution rules are classified. A tag inference chain is generated based on the classification result, thereby achieving efficient and intelligent construction of the rule execution chain. This application solves the problem of inconsistent tag generation caused by rule conflicts in traditional rule engines by introducing rule conflict detection and conflict resolution strategies, ensuring the logical consistency and accuracy of the tag inference chain. At the same time, by classifying rules and generating tag inference chains, the order and efficiency of rule execution are optimized, thereby providing the BI system with a more efficient and reliable object classification tag generation capability, and improving the accuracy and timeliness of BI system tag generation and optimization business decisions.

[0029] In a preferred embodiment, this application may be further configured as follows: before the step of fusing the received contextualized tags and object classification tag sets associated with the same business object to form a comprehensive tag set associated with the business object, the application includes the following steps:

[0030] Map the contextualized tags associated with the same business object to the object category tag set;

[0031] Detect whether there are duplicate or conflicting labels between the contextualized labels and the mapped object classification label set;

[0032] When duplicate or conflicting tags exist, the corresponding pre-set deduplication or conflict resolution strategy will be used for processing.

[0033] By adopting the above technical solution, the contextualized tags associated with the same business object are mapped to the object classification tag set. The system detects whether there are duplicate or conflicting tags between the contextualized tags and the mapped object classification tag set. When duplicate or conflicting tags are found, a pre-set deduplication strategy or conflict resolution strategy is applied to handle the situation, achieving intelligent fusion of contextualized tags and object classification tag sets. This application solves the problem of inconsistency in tag sets caused by tag duplication and conflict by introducing tag mapping and conflict detection mechanisms, ensuring the accuracy and consistency of the comprehensive tag set. Simultaneously, the application of deduplication and conflict resolution strategies optimizes the logic and efficiency of tag fusion, thereby providing the BI system with higher quality and more reliable comprehensive tag set support, and improving the accuracy and comprehensiveness of BI system tag generation and optimization business decisions.

[0034] In a preferred embodiment, this application can be further configured as follows: the step of fusing the received contextualized tags and object classification tag sets associated with the same business object to form a comprehensive tag set associated with the business object includes the following steps:

[0035] The multi-source dataset is input into the pre-trained emotion recognition model, enabling the emotion recognition model to extract emotion features from the multi-source dataset.

[0036] Based on the emotion features output by the emotion recognition model, emotion tags are generated that correspond to business objects.

[0037] The emotional tags, contextual tags, and object classification tags associated with the same business object are merged to form a comprehensive tag set associated with that business object.

[0038] By adopting the above technical solution, multi-source datasets are input into a pre-trained emotion recognition model, enabling the model to extract emotion features from the multi-source datasets. Based on the emotion features output by the emotion recognition model, emotion tags corresponding to business objects are generated. Contextualized tags, object classification tag sets, and emotion tags associated with the same business object are fused to form a comprehensive tag set, achieving accurate and dynamic generation of business object tags. This application, by introducing emotion recognition technology, can more comprehensively reflect the dynamic behavior and emotional state of business objects. At the same time, by fusing emotion tags with contextualized tags and object classification tag sets, the accuracy and business value of the tags are improved, thus providing more comprehensive tag support for the BI system.

[0039] The second objective of this invention is achieved through the following technical solution:

[0040] A business object tag generation and optimization system based on a BI system includes:

[0041] The data extraction module is used to extract multi-source datasets of all business objects from the BI system;

[0042] The contextualized label generation module is used to input multi-source datasets into a pre-trained general label model, enabling the general label model to generate contextualized labels corresponding to business objects based on the multi-source datasets.

[0043] The object classification label set generation module is used to generate the corresponding object classification label set for business objects based on multi-source datasets through the rule engine;

[0044] The tag fusion module is used to merge the contextualized tags and object classification tag sets associated with the same business object to form a comprehensive tag set associated with that business object.

[0045] The dynamic update module is used to receive business feedback data from the BI system in real time and update the comprehensive tag set based on the business feedback data.

[0046] By adopting the above technical solution, the data extraction module is used to extract multi-source datasets of all business objects from the BI system; the scenario-based label generation module is used to input the multi-source datasets into a pre-trained general label model, so that the general label model generates scenario-based labels corresponding to the business objects based on the multi-source datasets; the object classification label set generation module is used to generate object classification label sets corresponding to the business objects based on the multi-source datasets through the rule engine; the label fusion module is used to fuse the received scenario-based labels and object classification label sets associated with the same business object to form a comprehensive label set associated with the business object; and the dynamic update module is used to receive business feedback data from the BI system in real time and update the comprehensive label set based on the business feedback data.

[0047] In a preferred embodiment, this application may be further configured to include:

[0048] The recognition layer module is used to identify the input multi-source dataset, determine the industry domain of the multi-source dataset based on the recognition results, and output an industry confidence matrix based on the multi-source dataset.

[0049] The data reuse layer module is used to obtain corresponding general industry information based on the industry confidence matrix.

[0050] The feature extraction layer module is used to extract industry-wide features, vertical domain features, and business scenario features based on multi-source datasets and general industry information.

[0051] The label output layer module is used to output contextualized labels based on multi-source datasets, based on industry-wide features, vertical domain features, and business scenario features.

[0052] By adopting the above technical solution, the recognition layer module is used to identify the input multi-source dataset, and to determine the industry domain of the multi-source dataset based on the recognition results, and output an industry confidence matrix based on the multi-source dataset; the data reuse layer module is used to obtain the corresponding industry general information based on the industry confidence matrix; the feature extraction layer module is used to extract industry general features, vertical domain features and business scenario features based on the multi-source dataset and industry general information; the label output layer module is used to output scenario-based labels based on the multi-source dataset based on the industry general features, vertical domain features and business scenario features.

[0053] In summary, this application includes at least one of the following beneficial technical effects:

[0054] 1. This application solves the problem of traditional rule engines requiring frequent manual rule updates by combining a general tag model and a rule engine, reducing maintenance costs while improving response speed. At the same time, by receiving business feedback data in real time and dynamically updating the comprehensive tag set, it overcomes the defect that static tag systems cannot reflect business dynamics in real time, which improves the timeliness and accuracy of tag generation and optimization, thereby enhancing the decision support capability and business adaptability of the BI system. Attached Figure Description

[0055] Figure 1 This is a flowchart of an embodiment of a business object tag generation and optimization method based on a BI system according to this application;

[0056] Figure 2 This is a flowchart of step S20 in an embodiment of a business object tag generation and optimization method based on a BI system according to this application;

[0057] Figure 3 This is a flowchart of step S30 in an embodiment of a business object tag generation and optimization method based on a BI system according to this application;

[0058] Figure 4 This is a flowchart of step S32 in an embodiment of a business object tag generation and optimization method based on a BI system according to this application;

[0059] Figure 5 This is a flowchart of an implementation prior to step S40 in an embodiment of a business object tag generation and optimization method based on a BI system according to this application;

[0060] Figure 6 This is a flowchart of step S40 in an embodiment of a business object tag generation and optimization method based on a BI system according to this application. Detailed Implementation

[0061] The following is in conjunction with the appendix Figure 1-6This application will be described in further detail.

[0062] In one embodiment, such as Figure 1 As shown, this application discloses a method for generating and optimizing business object tags based on a BI system, which specifically includes the following steps:

[0063] S10: Extract multi-source datasets of all business objects from the BI system;

[0064] In this embodiment, the BI system, or Business Intelligence system, is a technology platform for collecting, processing, and analyzing enterprise data, aiming to provide data support and insights for business decisions; the business objects are the entities analyzed and managed in the BI system, such as customers, products, and orders; the multi-source dataset is a diverse collection of data extracted from different data sources (such as transaction systems, user behavior logs, external data, etc.), wherein the multi-source dataset includes structured data (such as transaction records and inventory data) and semi-structured data (such as log files and user feedback text);

[0065] Specifically, structured and semi-structured data related to the business object are extracted from multiple data sources in the BI system to form a multi-source dataset corresponding to the business object, ensuring the comprehensiveness and diversity of the data;

[0066] S20: Input the multi-source dataset into the pre-trained general label model, so that the general label model can generate contextualized labels corresponding to business objects based on the multi-source dataset;

[0067] In this embodiment, the general label model is a pre-trained machine learning model used to generate contextualized labels for business objects based on multi-source datasets; the contextualized labels are labels generated based on specific business scenarios or needs, such as "high-value customers" and "promotion-sensitive users".

[0068] Specifically, a pre-trained general labeling model is used to analyze multi-source datasets and generate contextualized labels related to business objects.

[0069] S30: Generate a set of object classification labels corresponding to business objects based on multi-source datasets using a rule engine;

[0070] In this embodiment, the rule engine is a system that performs logical reasoning and decision-making based on predefined rules, used to generate a set of object classification tags for business objects; the set of object classification tags is a collection of business object classification tags generated by the rule engine, such as "new customer" and "old customer";

[0071] Specifically, the rule engine is used to process multi-source datasets according to predefined business rules to generate classification labels for business objects;

[0072] S40: Merge the received contextualized tags and object classification tags associated with the same business object to form a comprehensive tag set associated with the business object;

[0073] In this embodiment, the comprehensive tag set is a set of business object tags formed by merging scenario-based tags and object classification tags, such as "high-value old customers" or "new customers sensitive to promotions";

[0074] Specifically, the contextualized tags and object classification tags will be integrated to form a more comprehensive and accurate set of business object tags;

[0075] S50: Receives business feedback data from the BI system in real time and updates the comprehensive tag set based on the business feedback data;

[0076] In this embodiment, the business feedback data is actual business usage data from the BI system, including user behavior, changes in business metrics, etc., which is used to optimize the tag generation process;

[0077] Specifically, based on actual business feedback data from the BI system, the comprehensive tag set is dynamically optimized and updated to ensure the timeliness and accuracy of the tags;

[0078] Furthermore, business feedback data is analyzed to evaluate the actual effectiveness of the current general tag model, rule engine, and comprehensive tag set. The evaluation is conducted using the following metrics: Tag accuracy: the degree to which tags match actual business scenarios; Tag coverage: the scope of coverage of tags for business objects; Tag usage rate: the actual frequency of tag usage in the BI system; Business metric improvement: the effect of tags on improving business decisions and metrics (such as conversion rate, user retention rate, etc.).

[0079] Furthermore, based on feedback data, the general label model can be iteratively optimized, including: model parameter adjustment: optimizing model parameters based on feedback data to improve the model's predictive and generalization capabilities; feature engineering optimization: adjusting feature extraction and selection strategies based on feedback data to ensure the effectiveness and relevance of features; model retraining: retraining the model using the latest feedback data to ensure the model can adapt to business changes; and recording the model update process and results to form a model update log, supporting subsequent traceability and optimization.

[0080] Furthermore, based on feedback data, the rule engine can be dynamically adjusted, including: rule optimization: adjusting the logic, conditions, and weights of existing rules according to feedback data to improve the accuracy and usability of the rules; rule addition: adding new rules to support more granular tag generation based on new business needs in the feedback data; rule elimination: removing rules that are ineffective or no longer applicable from the rule engine; recording the process and results of rule updates to form a rule update log to support subsequent traceability and optimization;

[0081] Furthermore, based on feedback data, the comprehensive tag set is dynamically optimized, including: tag correction: correcting the accuracy, consistency, and semantic expression of tags according to feedback data; tag addition: adding tags to support richer business scenarios based on new business needs in the feedback data; tag elimination: removing tags with low usage or no longer applicable from the comprehensive tag set; recording the tag update process and results to form a tag update log to support subsequent traceability and optimization.

[0082] In one embodiment, the general labeling model includes a recognition layer, a data reuse layer, a feature extraction layer, and a label output layer, such as... Figure 2 As shown, step S20 includes the following steps:

[0083] S21: The recognition layer identifies the input multi-source dataset, and determines the industry domain of the multi-source dataset based on the recognition results, outputting an industry confidence matrix based on the multi-source dataset;

[0084] S22: The data reuse layer obtains corresponding general industry information based on the industry confidence matrix;

[0085] S23: The feature extraction layer extracts industry-wide features, vertical domain features, and business scenario features based on multi-source datasets and industry-general information;

[0086] S24: The label output layer outputs scenario-based labels based on industry-general features, vertical domain features, and business scenario features, based on multi-source datasets.

[0087] In this embodiment, the recognition layer is the first layer of the general labeling model, used to identify and determine the industry domain of the input multi-source dataset, and output an industry confidence matrix. The industry confidence matrix represents the confidence distribution of the multi-source dataset in different industry domains and is used to guide subsequent data processing and label generation. The data reuse layer is the second layer of the general labeling model, used to obtain industry-related general information based on the industry confidence matrix, such as industry standards and industry characteristics. Industry general information includes industry-related general knowledge and data, such as customer behavior patterns in the retail industry and risk assessment standards in the financial industry. The feature extraction layer is a general labeling layer. The third layer of the label model is used to extract industry-wide features, vertical domain features, and business scenario features based on multi-source datasets and industry-wide information. Industry-wide features are general features applicable to multiple industries, such as customer age and spending amount. Vertical domain features are unique features specific to a particular industry or domain, such as repurchase rate in the retail industry and credit score in the financial industry. Business scenario features are features related to specific business scenarios, such as participation in promotional activities and changes in user behavior. The label output layer is the fourth layer of the general label model, used to generate scenario-based labels based on the extracted features, such as "high-value customers" and "promotion-sensitive users".

[0088] Specifically, the identification layer analyzes multi-source datasets to identify their respective industry sectors, such as retail, finance, and manufacturing, and outputs an industry confidence matrix representing the confidence distribution of the dataset across different industries. The data reuse layer, based on the industry confidence matrix, retrieves general industry-related information from a pre-stored industry knowledge base, such as industry standards and characteristics. The feature extraction layer combines multi-source datasets and general industry information to extract general features applicable to multiple industries, such as customer age, spending amount, vertical domain features specific to each industry (e.g., repurchase rate, credit score), and features specific to business scenarios (e.g., promotional activity participation, changes in user behavior). The label output layer generates contextualized labels based on the extracted features, such as "high-value customers" and "promotion-sensitive users," providing support for business decisions.

[0089] For example, for the financial industry, the identification layer analyzes multi-source datasets to determine if they belong to the financial industry and outputs an industry confidence matrix; the data reuse layer obtains general information about the financial industry, such as credit scoring standards and risk assessment models; the feature extraction layer extracts general industry features (such as customer income and asset size), vertical domain features (such as credit scores and loan default rates), and business scenario features (such as investment preferences and risk tolerance); the label output layer generates scenario-based labels, such as "high-risk customers," "high-net-worth customers," and "conservative investors."

[0090] For the manufacturing industry, the identification layer analyzes multi-source datasets to determine if they belong to the manufacturing industry and outputs an industry confidence matrix; the data reuse layer obtains general information about the manufacturing industry, such as production efficiency standards and equipment maintenance cycles; the feature extraction layer extracts general industry features (such as equipment uptime and output), vertical domain features (such as equipment failure rate and production efficiency), and business scenario features (such as order delivery cycle and raw material inventory level); the label output layer generates scenario-based labels, such as "high-efficiency production line", "high failure rate equipment", and "products with tight inventory".

[0091] Furthermore, the examples given in this embodiment are all broad industry categories, such as retail, finance, manufacturing, healthcare, education, and logistics. The specific examples can be adjusted to more precise and specific industry categories based on actual needs. For example, retail can be further subdivided into e-commerce retail and offline retail, finance can be further subdivided into banking and insurance, and manufacturing can be further subdivided into automobile manufacturing and electronics manufacturing, to meet more precise usage needs.

[0092] In one embodiment, such as Figure 3 As shown, step S30 includes the following steps:

[0093] S31: Obtain the industry confidence matrix output by the recognition layer, and match the execution rules corresponding to the multi-source dataset based on the pre-stored multi-dimensional business rule set and the industry confidence matrix;

[0094] S32: Construct a rule execution chain based on the matching execution rules, thereby generating a label inference chain;

[0095] S33: Input the multi-source dataset into the label inference chain to generate the object classification label set corresponding to the business object.

[0096] In this embodiment, the industry confidence matrix is ​​a matrix representing the confidence distribution of multi-source datasets in different industry domains, used to guide rule matching and label generation; the multi-dimensional business rule set is a predefined set of business rules covering rules of different dimensions, such as field-level rules, logical rules, industry rules, etc., used to generate object classification labels; the execution rule is a rule obtained by matching the industry confidence matrix and the multi-dimensional business rule set, used to process the multi-source dataset and generate object classification labels; the rule execution chain is an execution chain constructed by the matched execution rules according to a specific order and logical relationship, used to guide the label generation process; the label inference chain is an inference logic chain generated based on the rule execution chain, used to transform the multi-source dataset into an object classification label set; the object classification label set is a set of business object classification labels generated based on the rule engine, such as "individual customer", "enterprise customer", "new customer", "old customer", etc.

[0097] Furthermore, the multidimensional business rule set includes industry normative rules (such as anti-money laundering transaction thresholds in the financial industry), enterprise business rules (such as membership level classification standards in the retail industry), data feature rules (such as the definition of discretization intervals for numerical fields), and spatiotemporal constraint rules (such as the effective time range of promotional activities).

[0098] Specifically, the industry confidence matrix output by the recognition layer is obtained, and combined with the pre-stored multi-dimensional business rule set, the execution rules corresponding to the multi-source datasets are matched to ensure that the rules are compatible with the data features and industry characteristics. According to the matched execution rules, the rule execution chain is constructed according to priority and logical relationship to form a label inference chain, clarifying the order and logical relationship of rule execution. The multi-source datasets are input into the label inference chain, and the rule engine performs inference and calculation on the data to generate the object classification label set corresponding to the business object.

[0099] In one embodiment, such as Figure 4 As shown, step S32 includes the following steps:

[0100] S321: When constructing a rule execution chain, detect whether there are rule conflicts between the matched execution rules;

[0101] S322: When a rule conflict is detected, it is handled based on a pre-set conflict resolution strategy;

[0102] S323: Hierarchize the matching execution rules and generate a label inference chain based on the hierarchical results.

[0103] In this embodiment, the rule execution chain is an execution chain constructed by matching execution rules according to a specific order and logical relationship, used to guide the label generation process; rule conflict refers to the conflict between different rules in the rule execution chain due to overlapping conditions or inconsistent conclusions, such as two rules being applicable to the same dataset but generating different labels; conflict resolution strategy is a predefined strategy used to resolve rule conflict problems, such as priority resolution, weight resolution, and logical resolution; rule classification is the classification of rules according to their importance, applicability, or business value, such as high-priority rules, medium-priority rules, and low-priority rules; and label inference chain is an inference logic chain generated based on the rule execution chain, used to transform multi-source datasets into object classification label sets.

[0104] Furthermore, the rule execution chain can be designed as a hierarchical structure to obtain a tag inference chain that supports tag inference from coarse-grained to fine-grained:

[0105] Level 1 Inference: Based on the industry confidence matrix and core business rules, generate coarse-grained classification labels (such as industry category and business type) for business objects;

[0106] Second-level inference: Based on coarse-grained tags, combine fine-grained rules (such as field-level rules and time series rules) to generate more refined tags (such as customer value level and product life cycle stage);

[0107] Level 3 reasoning: Based on fine-grained labels, personalized labels (such as customer preferences and risk levels) are generated through advanced rules (such as machine learning rules and semantic rules);

[0108] The result of each level of reasoning serves as the input for the next level of reasoning, forming a progressively refined chain of tag-based reasoning.

[0109] Specifically, when constructing the rule execution chain, it detects whether there are conflicts between matching execution rules (such as overlapping rule conditions but inconsistent conclusions). Through rule logic analysis and data feature comparison, it identifies potential rule conflict points and records the specific information of conflicting rules, including rule conditions, conclusions, and conflict types, providing a basis for subsequent resolution. When a rule conflict is detected, it is handled using a pre-set conflict resolution strategy, such as: priority resolution: assigning priority to each rule, with higher priority rules overriding lower priority rules; weight resolution: assigning weights to rules based on their historical execution performance or industry confidence matrix, with rules with higher weights being executed first; and logic resolution: merging or adjusting conflicting rules through logical reasoning (such as Boolean operations and fuzzy logic) to generate a new set of rules. Consistency Conclusions and Business Resolution: Combining business scenarios and actual needs, select the rule conclusions that best meet business objectives, while recording the conflict resolution process to form a traceable conflict handling log, supporting subsequent optimization and auditing; classify the matched execution rules and generate a tag inference chain based on the classification results. Classify the matched execution rules according to their importance, applicability, or business value. For example: high-priority rules: rules directly related to core business indicators are executed first; medium-priority rules: rules related to auxiliary business indicators are executed second; low-priority rules: rules related to peripheral business indicators are executed last. Based on the classification results, construct the rule execution chain according to the priority order, generate a tag inference chain, and clarify the order and logical relationship of rule execution.

[0110] Taking the retail industry as an example, if two conflicting rules are detected, such as "users with a spending amount greater than 1,000 yuan are marked as high-value customers" and "users with a spending frequency greater than 5 times are marked as high-value customers", a priority resolution strategy is adopted. The "spending amount rule" is assigned high priority, covering the "spending frequency rule". The "spending amount rule" is listed as a high-priority rule, and a tag inference chain is generated. For example, the "spending amount rule" is executed first, and then the "spending frequency rule" is executed.

[0111] For example, in the retail industry, the rules for execution may include various rules on customer purchasing behavior, such as purchase frequency and purchase amount. When building the rule execution chain, if it is found that a customer meets both the "high-value customer" rule and the "potential churn customer" rule, then it needs to be handled through conflict resolution strategies to ultimately generate a customer classification label that best meets the current business needs.

[0112] In one embodiment, such as Figure 5 As shown, before step S40, the following steps are included:

[0113] S37: Map the contextualized tags associated with the same business object to the object classification tag set;

[0114] S38: Detect whether there are duplicate or conflicting labels between the contextualized labels and the mapped object classification label set;

[0115] S39: When duplicate or conflicting tags exist, the corresponding pre-set deduplication strategy or conflict resolution strategy shall be used for processing.

[0116] In this embodiment, tag mapping involves associating contextualized tags corresponding to the same business object with a set of object classification tags, thereby establishing a correspondence between the two. Duplicate tags are those where the contextualized tags and the object classification tags have the same name or the same semantics. Conflicting tags are those where the contextualized tags and the object classification tags have the same name but different semantics, or the same semantics but different generation logic. The deduplication strategy is a predefined strategy used to handle duplicate tag issues, such as retaining high-confidence tags or merging similar tags. The conflict resolution strategy is a predefined strategy used to handle conflict tag issues, such as priority resolution, weight resolution, and logic resolution.

[0117] Specifically, the system aligns contextualized tags and object classification tags, identifies the relationships between them, and maps contextualized tags to object classification tags using a tag mapping table or semantic matching algorithm to ensure consistency between tags. The tag mapping process is recorded to create a traceable tag relationship log, supporting subsequent analysis and optimization. The system also detects duplicate tags (e.g., tags with the same name but different meanings, or tags with the same meaning but different names) and conflicting tags (e.g., tags with the same name but inconsistent conclusions, or tags with the same conclusions but inconsistent generation logic). The detection results, including specific information about duplicate and conflicting tags, are recorded to provide a basis for subsequent processing. When duplicate or conflicting tags are found, pre-set deduplication or conflict resolution strategies are applied: Deduplication strategy: For duplicate tags, high-confidence tags are retained or similar tags are merged to ensure the uniqueness and accuracy of the tags; Conflict resolution strategy: For conflicting tags, a multi-dimensional resolution strategy is adopted, including priority resolution, weight resolution, logical resolution, and business resolution.

[0118] Taking the retail industry as an example, the scenario-based tag "high-value customer" is mapped to the object classification tag set "VIP customer". Duplicate tags "high-value customer" and "VIP customer" are detected, as well as conflicting tags "promotion-sensitive user" and "new customer". A deduplication strategy is used to merge "high-value customer" and "VIP customer", and a priority resolution strategy is used to select "promotion-sensitive user" to cover "new customer".

[0119] In one embodiment, such as Figure 6 As shown, step S40 includes the following steps:

[0120] S41: Input the multi-source dataset into the pre-trained emotion recognition model, so that the emotion recognition model can extract emotion features from the multi-source dataset;

[0121] S42: Generate emotion tags that correspond to business objects based on the emotion features output by the emotion recognition model;

[0122] S43: Merge the emotion tags, contextual tags, and object classification tags associated with the same business object to form a comprehensive tag set associated with that business object;

[0123] In this embodiment, the emotion recognition model is a pre-trained machine learning model used to extract emotion features from multi-source datasets, such as text sentiment analysis models and speech emotion recognition models; the emotion features are emotion-related features extracted from multi-source datasets, including emotion categories (such as positive, negative, neutral), emotion intensity (such as high, medium, low), and emotion change trends (such as the frequency of emotion fluctuations); the emotion tags are tags generated based on emotion features and related to business objects, such as "highly satisfied customers", "complaint-prone users", and "emotionally fluctuating users".

[0124] Specifically, emotion-related data (such as user comments, customer service dialogue records, social media text, voice interaction data, etc.) is extracted from multi-source datasets. This emotion-related data is then input into a pre-trained emotion recognition model to extract emotion features, including emotion category, emotion intensity, and emotion change trend. Based on the extracted emotion features, emotion labels related to business objects are generated, such as: "Highly Satisfied Customers": customers with positive emotion features and high emotion intensity; "Complaint-Prone Users": users with negative emotion features and high emotion intensity; "Emotionally Fluctuating Users": users with high frequency of emotion changes. The emotion labels, contextual labels, and object classification label sets are mapped to identify the relationships between the three. The mapped labels are then fused to generate a comprehensive label set, such as: "High-Value, Highly Satisfied Customers": the contextual label of "High-Value Customers" is fused with the emotion label of "Highly Satisfied Customers"; "High-Risk Complaint-Prone Customers": the object classification label of "High-Risk Customers" is fused with the emotion label of "Complaint-Prone Users". The fused labels are then optimized, including deduplication, semantic correction, and confidence calculation.

[0125] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0126] In one embodiment, a business object tag generation and optimization system based on a BI system is provided. This BI system-based business object tag generation and optimization system corresponds one-to-one with the business object tag generation and optimization method based on a BI system described in the above embodiment. The BI system-based business object tag generation and optimization system includes:

[0127] The data extraction module is used to extract multi-source datasets of all business objects from the BI system;

[0128] The contextualized label generation module is used to input multi-source datasets into a pre-trained general label model, enabling the general label model to generate contextualized labels corresponding to business objects based on the multi-source datasets.

[0129] The object classification label set generation module is used to generate the corresponding object classification label set for business objects based on multi-source datasets through the rule engine;

[0130] The tag fusion module is used to merge the contextualized tags and object classification tag sets associated with the same business object to form a comprehensive tag set associated with that business object.

[0131] The dynamic update module is used to receive business feedback data from the BI system in real time and update the comprehensive tag set based on the business feedback data;

[0132] Optional, also includes:

[0133] The recognition layer module is used to identify the input multi-source dataset, determine the industry domain of the multi-source dataset based on the recognition results, and output an industry confidence matrix based on the multi-source dataset.

[0134] The data reuse layer module is used to obtain corresponding general industry information based on the industry confidence matrix.

[0135] The feature extraction layer module is used to extract industry-wide features, vertical domain features, and business scenario features based on multi-source datasets and general industry information.

[0136] The label output layer module is used to output contextualized labels based on multi-source datasets, based on industry-wide features, vertical domain features, and business scenario features.

[0137] For specific limitations regarding a business object tag generation and optimization system based on a BI system, please refer to the limitations of a business object tag generation and optimization method based on a BI system mentioned above, which will not be repeated here. Each module in the aforementioned business object tag generation and optimization system based on a BI system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0138] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for generating and optimizing business object tags based on a BI system, characterized in that: Including the following steps: Extract multi-source datasets of all business objects from the BI system; The multi-source dataset is input into a pre-trained general label model, which then generates contextualized labels for business objects based on the multi-source dataset. The general labeling model also includes an identification layer, a data reuse layer, a feature extraction layer, and a label output layer; The recognition layer identifies the input multi-source dataset, determines the industry domain of the multi-source dataset based on the recognition results, and outputs an industry confidence matrix based on the multi-source dataset. The data reuse layer obtains corresponding industry-specific information based on the industry confidence matrix; The feature extraction layer extracts industry-wide features, vertical domain features, and business scenario features based on multi-source datasets and general industry information; The label output layer outputs contextualized labels based on industry-wide features, vertical domain features, and business scenario features, using multi-source datasets. The rule engine generates a set of object classification labels corresponding to business objects based on multi-source datasets; The step of generating a set of object classification labels corresponding to business objects based on a multi-source dataset using a rule engine further includes the following steps: Obtain the industry confidence matrix output by the recognition layer, and match the execution rules corresponding to the multi-source datasets based on the pre-stored multi-dimensional business rule set and the industry confidence matrix; A rule execution chain is constructed based on the matching execution rules, thereby generating a tag inference chain; Input the multi-source dataset into the label inference chain to generate the object classification label set corresponding to the business object; The received contextualized tags and object classification tags associated with the same business object are merged to form a comprehensive tag set associated with that business object; Receive business feedback data from the BI system in real time and update the comprehensive tag set based on the business feedback data.

2. The method for generating and optimizing business object tags based on a BI system according to claim 1, characterized in that: The step of constructing a rule execution chain based on matching execution rules to generate a label inference chain includes the following steps: When constructing the rule execution chain, check whether there are rule conflicts between the matched execution rules; When a rule conflict is detected, it is handled based on a pre-set conflict resolution strategy. The matching execution rules are hierarchically classified, and a label inference chain is generated based on the classification results.

3. The method for generating and optimizing business object tags based on a BI system according to claim 1, characterized in that: Before the step of fusing the received contextualized tags and object classification tag sets associated with the same business object to form a comprehensive tag set associated with the business object, the following steps are included: Map the contextualized tags associated with the same business object to the object category tag set; Detect whether there are duplicate or conflicting labels between the contextualized labels and the mapped object classification label set; When duplicate or conflicting tags exist, the corresponding pre-set deduplication or conflict resolution strategy will be used for processing.

4. The method for generating and optimizing business object tags based on a BI system according to claim 1, characterized in that: The step of fusing the received contextualized tags and object classification tag sets associated with the same business object to form a comprehensive tag set associated with the business object includes the following steps: The multi-source dataset is input into the pre-trained emotion recognition model, enabling the emotion recognition model to extract emotion features from the multi-source dataset. Based on the emotion features output by the emotion recognition model, emotion tags are generated that correspond to business objects. The emotional tags, contextual tags, and object classification tags associated with the same business object are merged to form a comprehensive tag set associated with that business object.

5. A business object tag generation and optimization system based on a BI system, characterized in that: include: The data extraction module is used to extract multi-source datasets of all business objects from the BI system; The contextualized label generation module is used to input multi-source datasets into a pre-trained general label model, so that the general label model generates contextualized labels corresponding to business objects based on the multi-source datasets. The general label model also includes a recognition layer, a data reuse layer, a feature extraction layer, and a label output layer. The scenario-based label generation module also includes a recognition layer module, a data reuse layer module, a feature extraction layer module, and a label output layer module; The recognition layer module is used to recognize the input multi-source dataset, determine the industry domain of the multi-source dataset based on the recognition results, and output an industry confidence matrix based on the multi-source dataset. The data reuse layer module is used to obtain corresponding industry-wide information based on the industry confidence matrix; The feature extraction layer module is used to extract industry-wide features, vertical domain features, and business scenario features based on multi-source datasets and industry-general information. The label output layer module is used to output contextualized labels based on multi-source datasets based on industry-general features, vertical domain features, and business scenario features. The object classification label set generation module is used to generate the corresponding object classification label set for business objects based on multi-source datasets through the rule engine; The step of generating a set of object classification labels corresponding to business objects based on a multi-source dataset using a rule engine further includes the following steps: Obtain the industry confidence matrix output by the recognition layer module, and match the execution rules corresponding to the multi-source datasets based on the pre-stored multi-dimensional business rule set and the industry confidence matrix; A rule execution chain is constructed based on the matching execution rules, thereby generating a tag inference chain; Input the multi-source dataset into the label inference chain to generate the object classification label set corresponding to the business object; The tag fusion module is used to merge the contextualized tags and object classification tag sets associated with the same business object to form a comprehensive tag set associated with that business object. The dynamic update module is used to receive business feedback data from the BI system in real time and update the comprehensive tag set based on the business feedback data.

6. The business object tag generation and optimization system based on a BI system according to claim 5, characterized in that: Also includes: The conflict detection module is used to detect whether there are rule conflicts between matched execution rules when constructing the rule execution chain; The detection and processing module is used to process the detected rule conflicts based on a pre-set conflict resolution strategy. The results generation module is used to classify the matched execution rules and generate a label inference chain based on the classification results.

Images