Target data source model intelligent matching system and method

Abstract

The invention discloses a target data source model intelligent matching system and method, and relates to the technical field of target data source management and intelligent matching, the system mainly comprises a model library, a tag library, an interface module, a matching module and a recommendation module; the system predefines tags of five dimensions including regions, industries, application scenes, keywords and model types in a tag library, and corresponding tags are added to the models in an automatic and manual mode when the models are created or updated; when the evaluation end creates an evaluation task, target data source information is transmitted to the system in a standard JSON (JavaScript Object Notation) format; after receiving the data, the interface module transmits the data to a matching module, and the matching module analyzes evaluation target information, extracts key information and performs multi-dimensional label comparison on models in a model library according to a preset matching rule and a weight mechanism, so that the most matched model is screened out; and finally, the recommendation module feeds back a matching result to the evaluation end, and the user can directly apply the recommendation model to the evaluation task.

Description

Technical Field

[0001] The present invention relates to the field of target data source management and intelligent matching technology, and in particular to a target data source model intelligent matching system and method. Background Art

[0002] Currently, in data evaluation tasks, the evaluation targets often include multiple models, and the traditional model matching method relies on manual addition one by one.

[0003] As a result, each time a new evaluation task is created, it is necessary to manually add corresponding models for each evaluation target one by one; and because the model is the smallest granularity evaluation rule, each evaluation target involves multiple models, and manual operation is time-consuming and labor-intensive; in addition, the existing technology lacks intelligent matching means and cannot automatically match the most suitable model according to the target characteristics, resulting in inaccurate or incomplete matching results.

[0004] The above existing problems limit the efficiency and accuracy of data evaluation tasks, and an automated and intelligent model matching solution is urgently needed to solve the above defects. Summary of the invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] The present invention provides a target data source model intelligent matching system and method solution. In the traditional matching scheme, each time a new evaluation task is created, it is necessary to manually add the corresponding model for each evaluation target one by one; and because the model is the evaluation rule with the smallest granularity, each evaluation target involves multiple models, and manual operation is time-consuming and labor-intensive; in addition, the existing technology lacks intelligent matching means and cannot automatically match the most suitable model according to the target characteristics, resulting in the problem that the matching results may be inaccurate or incomplete.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, an embodiment of the present invention provides a target data source model intelligent matching system, which includes: The tag library defines tags according to five dimensions: region tags, industry tags, application scenario tags, keyword tags, and model type tags, and is used to uniformly manage tag classification; Model library, when creating a model, automatically or manually adding label information corresponding to the label library to the model, thereby realizing the association between the model and the label library; An interface module, used for receiving evaluation target information from the evaluation end; A matching module, used to parse and analyze the evaluation target information, and match it based on the label attached to the model and the label definition in the label library; The recommendation module is used to return the matched model information to the evaluation end.

[0008] As a preferred solution of the target data source model intelligent matching system described in the present invention, the tag library only defines tags in five dimensions: region, industry, application scenario, keyword, and model type, and is not directly associated with a specific model; When a model is created or updated in the model library, the corresponding tag information is attached to the model by calling the tag library.

[0009] As a preferred solution of the target data source model intelligent matching system described in the present invention, the matching module is used to extract features from the evaluation target information, including region, industry, application scenario, keywords and model type features, and compare them one by one with the label information of each model in the model library to calculate the model set that matches the evaluation target.

[0010] As a preferred solution of the target data source model intelligent matching system described in the present invention, the matching module also scores the results based on pre-set label matching weights and priority comparisons, and preferentially outputs model information with the highest score or exceeding the set threshold to the recommendation module.

[0011] As a preferred solution of the target data source model intelligent matching system described in the present invention, the evaluation target information received by the interface module is transmitted in JSON format, and the information at least includes: Whether to match the basic model's identifier basic, Task description task, The asset name assetName and asset description assetDesc, Industry, country and region, Tables and fields information, where field information includes field name, field description, whether it is the primary key, and field type.

[0012] In a second aspect, the present invention provides a target data source model intelligent matching method, comprising: Step S1, establish a tag library and define tags according to five dimensions: region, industry, application scenario, keyword and model type; Step S2, when creating or updating a model in the model library, attaching corresponding tag information in the tag library to the model automatically or manually; Step S3, the evaluation end transmits the evaluation target information to the interface module according to a pre-agreed standard format; Step S4, the matching module parses the evaluation target information to obtain region, industry, application scenario, keyword and model type characteristics; Step S5, the matching module compares the features obtained by parsing with the label information of each model in the model library, and calculates the matching degree based on the matching rules and weights; Step S6: the recommendation module returns the matched model information to the evaluation end.

[0013] As a preferred solution of the target data source model intelligent matching method described in the present invention, wherein: in step S2, when automatically adding tags to the model, the region, industry, application scenario, keyword and model type that are consistent with the tag definition in the tag library are matched by analyzing the text information or key fields of the model, and are automatically attached to the model.

[0014] As a preferred solution of the target data source model intelligent matching method described in the present invention, the matching rules in step S5 include: Label matching is performed on five dimensions: region, industry, application scenario, keyword, and model type. If multiple models meet the label matching conditions, they are scored according to the preset priority or weight. When the score reaches or exceeds the predetermined threshold, the model is judged to be highly matched with the evaluation target.

[0015] As a preferred solution of the target data source model intelligent matching method described in the present invention, the method further includes: after the recommendation module returns the matching result, the evaluation end manually screens or rechecks the recommended model to ensure that the selected model is consistent with the actual business needs.

[0016] As a preferred solution of the target data source model intelligent matching method described in the present invention, the method further includes the steps of recording and counting the matching process and usage, specifically including: Record matching logs, including evaluation target information, label comparison results, final recommendation model and its matching degree; Based on the matching logs, the definition of the tag library and the matching rules are iteratively optimized to continuously improve the accuracy and efficiency of model matching.

[0017] The beneficial effects of the present invention are: The present invention analyzes the evaluation target information, automatically extracts features such as region, industry, application scenario, keywords and model type, realizes automatic matching of models, and significantly reduces the amount of manual operations.

[0018] The present invention adopts preset matching rules and weight mechanism, and multi-dimensional cross-comparison model labels to improve the accuracy and reliability of matching.

[0019] Each tag in the tag library of the present invention can be flexibly adjusted or expanded according to actual business needs, which is convenient for integrating different data sources and model management platforms.

[0020] The present invention shortens the creation cycle of evaluation tasks through an automated intelligent matching process, improves overall work efficiency, and saves users a lot of time and labor costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0022] Figure 1 It is a schematic diagram of the five-dimensional label structure of the model library and label library of the present invention.

[0023] Figure 2 It is a schematic diagram of the rule model matching process of the present invention.

[0024] Figure 3 This is an example diagram of the JSON structure of the evaluation target information of the present invention.

[0025] Figure 4 This is a schematic diagram of the relationship between region and industry label matching in the present invention. DETAILED DESCRIPTION

[0026] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0027] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0028] Secondly, the term "one 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 term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0029] Example 1, reference Figure 1 and Figure 2, which is the first embodiment of the present invention, and provides a target data source model intelligent matching system, wherein the overall system architecture and label library are designed as follows: The target data source model intelligent matching system of the present invention includes: Tag library: defines tags according to five dimensions: region tags, industry tags, application scenario tags, keyword tags, and model type tags; Model library: used to store and maintain the information of each model, including the description of the model itself, the model identifier, and the label corresponding to the label library; Interface module: The evaluation end calls this module to transfer the evaluation target information to the model library in a standard format (such as JSON); Matching module: parses the evaluation target information and performs model matching based on the label dimension; Recommendation module: returns the matching results (model recommendation information) to the evaluation end; The five dimensions of the tag library include: See also Figure 1 , the present invention divides the model tags into five dimensions in the tag library: 1. Regional labels: such as country, province, city and other regional information; 2. Industry labels: such as finance, medical care, education, manufacturing, etc.; 3. Application scenario labels: such as risk control, user portrait, credit rating, quality inspection, etc.; 4. Keyword tags: Keywords or topics derived from summarizing and refining the model content; 5. Model type label: such as basic model, prediction model, analysis model, etc.; The tag library itself only defines the type and hierarchical structure of tags and is not directly associated with the model. Tags are only attached to specific models automatically or manually when the model is created or updated. In the present invention, in order to achieve an accurate description of the model, it is necessary to associate the tag information in the tag library with the model during the model creation or maintenance phase. For the specific process, see Figure 2 , the main steps are as follows: 1. Model creation: The user enters the basic information of the new model (name, description, applicable scenarios, etc.) in the model library; 2. Automatic label extraction: The system automatically identifies potentially applicable regions, industries, application scenarios, keywords, and model type labels based on the model description text or key fields, and suggests the user to confirm; 3. Manual label addition: Users can manually check or add other labels in the interface to ensure the accuracy and comprehensiveness of model labels; 4. Tag association: The confirmed tags will be associated with the model and stored in the model library; When the evaluation end needs to create an evaluation task and select a suitable model, the relevant information of the target data source will be transferred to the model library in a standard format through the interface module. Figure 3 , which gives an example of the JSON structure. Common fields include: 1.basic: Indicates whether the basic model is matched (true / false); 2.task: task description; 3.assetName: data asset name; 4.assetDesc: data asset description; 5.Industry: industry (such as finance, medical care, etc.); 6.country, region: country and region information; 7.tables: table information, which can contain multiple table objects; name: table name; desc: table description; fields: field array, each field contains: name: field name; desc: field description; primary: whether it is the primary key; type: field type (such as INT, VARCHAR, etc.); After the interface module passes the above JSON data to the matching module, the matching module first extracts and identifies the features such as region, industry, application scenario, keyword, model type, etc. According to the technical solution of the present invention, model matching is mainly performed based on the label dimension, including the following core points: 1. Multi-dimensional label comparison: compare the analyzed evaluation target label requirements (region, industry, application scenario, keyword, model type) with the labels of each model in the model library; 2. Weight and priority: Different weights or priorities can be set for different tag dimensions. For example, regions and industries usually have higher priorities. 3. Matching score: Calculate the comprehensive matching score for each candidate model. If it exceeds the predetermined threshold, it is considered to meet the recommendation conditions; 4. Conflict and parallel processing: When multiple models are ranked in parallel, secondary screening can be performed based on business needs or user preferences; Example 2, reference Figure 4 , which is the second embodiment of the present invention, and provides a target data source model intelligent matching method, see Figure 4, which illustrates how the regional and industry labels are matched with the model library when creating tasks on the evaluation side. The overall process is as follows: 1. Parse the evaluation target information: The matching module obtains key fields such as region and industry in the JSON data; 2. Retrieve candidate models: Filter out models that have been associated with corresponding regional and industry labels in the model library; 3. Multi-dimensional cross-matching: further compare application scenarios, keywords, model types and other labels to calculate the comprehensive matching degree; 4. Recommended model: The model that meets the conditions and its matching score are returned to the evaluation end through the recommendation module; 5. Confirmation on the evaluation side: After receiving the recommendation results, the evaluation side can automatically associate the qualified models with the new task and perform manual verification or adjustment when necessary; In summary, by establishing a multi-dimensional label library and improving label information when creating a model, the present invention can automatically complete model matching when the evaluation end submits a task, without the need to manually search or add models one by one, thus greatly reducing the amount of operations; The five dimensions of region, industry, application scenario, keyword, and model type in the label library of the present invention can be expanded or refined according to business needs to meet the application scenarios of different industries and regions; The system of the present invention can access more data sources or third-party models. It only needs to supplement the labels when creating the model to integrate it into the overall matching system. If the label library needs to be updated, the label definition can also be added, deleted or adjusted during the maintenance phase, thereby continuously improving the accuracy of model recommendations. The present invention is applicable to various scenarios such as data evaluation, quality inspection, risk control, analysis and prediction. In particular, when the evaluation target has multiple attribute information (such as industry, region, data type), the label matching method can significantly improve the efficiency and accuracy of model selection. In addition, the present invention can be modified or expanded in many ways based on the above embodiments, including: Natural language processing, machine learning and other methods can be used to automatically identify the model's topic keywords, further improving the accuracy of automatic tagging; The weights of labels in each dimension can be flexibly adjusted based on business scenarios; if finer-grained judgment is required, fuzzy matching or similarity calculation mechanisms can be introduced; The model library can be connected with other systems (such as data warehouses and metadata management platforms) to achieve comprehensive management of data sources and models; In summary, the present invention establishes a multi-dimensional label library and automatically or manually manages the labels of the model, thereby realizing the function of automatically recommending matching models when submitting tasks on the evaluation end, greatly reducing the workload of manual configuration and improving the accuracy of model matching.

[0030] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A target data source model intelligent matching system, characterized by: include, The tag library defines tags according to five dimensions: region tags, industry tags, application scenario tags, keyword tags, and model type tags, and is used to uniformly manage tag classification; A model library, which automatically or manually adds label information corresponding to the label library to the model when the model is created; An interface module, used for receiving evaluation target information from the evaluation end; A matching module, used to parse and analyze the evaluation target information, and match it based on the label attached to the model and the label definition in the label library; The recommendation module is used to return the matched model information to the evaluation end.

2. A target data source model intelligent matching system as claimed in claim 1, characterized in that: The tag library only defines tags in five dimensions: region, industry, application scenario, keyword, and model type, and is not directly associated with a specific model; When a model is created or updated in the model library, the corresponding tag information is attached to the model by calling the tag library.

3. A target data source model intelligent matching system as claimed in claim 2, characterized in that: The matching module is used to extract features from the evaluation target information, including region, industry, application scenario, keyword and model type features, and compare them one by one with the label information of each model in the model library to calculate the model set that matches the evaluation target.

4. A target data source model intelligent matching system as claimed in claim 3, characterized in that: The matching module also scores the results based on the preset tag matching weights and priority comparisons, and preferentially outputs the model information with the highest score or exceeding the set threshold to the recommendation module.

5. A target data source model intelligent matching system as claimed in claim 4, characterized in that: The evaluation target information received by the interface module is transmitted in JSON format, and the information at least includes: Whether to match the basic model's identifier basic, Task description task, The asset name assetName and asset description assetDesc, Industry, country and region, Tables and fields information, where field information includes field name, field description, whether it is the primary key, and field type.

6. A target data source model intelligent matching method, based on a target data source model intelligent matching system according to any one of claims 1 to 5, characterized in that: include: Step S1, establish a tag library and define tags according to five dimensions: region, industry, application scenario, keyword and model type; Step S2, when creating or updating a model in the model library, attaching corresponding tag information in the tag library to the model automatically or manually; Step S3, the evaluation end transmits the evaluation target information to the interface module according to a pre-agreed standard format; Step S4, the matching module parses the evaluation target information to obtain region, industry, application scenario, keyword and model type characteristics; Step S5, the matching module compares the features obtained by parsing with the label information of each model in the model library, and calculates the matching degree based on the matching rules and weights; Step S6: the recommendation module returns the matched model information to the evaluation end.

7. A target data source model intelligent matching method as claimed in claim 6, characterized in that: In step S2, when automatically adding tags to the model, the text information or key fields of the model are analyzed to match the region, industry, application scenario, keyword and model type that are consistent with the tag definition in the tag library, and automatically attach them to the model.

8. A target data source model intelligent matching method as claimed in claim 7, characterized in that: The matching rules in step S5 include: Label matching is performed on five dimensions: region, industry, application scenario, keyword, and model type. If multiple models meet the label matching conditions, they are scored according to the preset priority or weight. When the score reaches or exceeds the predetermined threshold, the model is judged to be highly matched with the evaluation target.

9. A target data source model intelligent matching method as claimed in claim 8, characterized in that: The method further includes: after the recommendation module returns the matching result, the evaluation end manually screens or rechecks the recommendation model to ensure that the selected model is consistent with the actual business needs.

10. A target data source model intelligent matching method as claimed in claim 9, characterized in that: The method further includes the steps of recording and counting the matching process and usage, specifically including: Record matching logs, including evaluation target information, label comparison results, final recommendation model and its matching degree; Based on the matching logs, the definition of the tag library and the matching rules are iteratively optimized to continuously improve the accuracy and efficiency of model matching.

Images