Business data feature sharing method and system based on machine learning

Abstract

The invention provides a business data feature sharing method and system based on machine learning, and the method comprises the steps: firstly obtaining a business link feature set of a talent service field, including a personnel core, post association and service process features, and a business demand description of a feature calling main body; then constructing a feature anchoring rule to perform association processing to obtain a feature anchoring result, inputting the feature anchoring result into a pre-training machine learning model to calculate a shared adaptation parameter, and constructing a shared execution link including feature extraction, processing and delivery nodes according to the shared adaptation parameter; and carrying out association mapping on the features output by the sharing execution link and a talent service business system of the feature calling main body, generating a feature sharing result containing a plurality of identifiers, and feeding back the feature sharing result to the business system to support business decision operation, thereby realizing efficient and accurate sharing of talent service business data features.

Description

Technical Field

[0001] This invention relates to the field of machine learning technology, and more specifically, to a method and system for sharing business data features based on machine learning. Background Technology

[0002] In the talent services sector, with the diversification of business operations and the deepening of digital transformation, a large amount of valuable data features have accumulated across different business processes. Simultaneously, multiple stakeholders have diverse needs for these features. However, the talent services sector currently faces numerous challenges in sharing business data features.

[0003] On the one hand, existing business data management methods are often fragmented, with each business segment independently storing and managing its own data characteristics, lacking unified standards and specifications. For example, core personnel characteristics, job-related characteristics, and service process characteristics may be managed by different departments or systems, making it difficult to effectively integrate and share these characteristics. This forces the feature requesting entity to communicate and coordinate with multiple departments or systems when obtaining the required characteristics, resulting in low efficiency and a high risk of data inconsistencies.

[0004] On the other hand, there is a lack of precise matching mechanisms for the business needs of feature requesters. The business needs described by feature requesters are often broad and difficult to precisely correlate with specific business process features. For example, a feature requester might only state that it needs certain types of business data features, but without clearly specifying the specific requirements for these features, such as timeliness or application direction. This results in the inability to meet the actual needs of feature requesters when sharing business data features, reducing the value and effectiveness of feature sharing.

[0005] Furthermore, traditional data sharing methods lack flexibility and intelligent processing capabilities. When constructing sharing links and performing feature processing, fixed patterns and methods are often used, making it impossible to dynamically adjust according to the different needs of the feature-receiving entities. This makes shared business data features potentially unsuitable for complex and ever-changing business scenarios, limiting the innovation and development of talent service businesses. Summary of the Invention

[0006] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide a business data feature sharing method based on machine learning, the method comprising: The system acquires a set of business process features in the talent service field and a description of the business needs of the feature caller. The set of business process features includes core personnel features, job-related features, and service process features. The core personnel features are generated based on personnel capability representation information, qualification certification information, and service preference information in the talent service field. The job-related features are generated based on job requirement definition information, responsibility boundary information, and adaptation standard information in the talent service field. The service process features are generated based on business interaction record information, service cycle record information, and feedback evaluation information in the talent service field. The description of the business needs of the feature caller includes the type range, timeliness requirements, and application direction of the business data features required by the feature caller. Construct feature anchoring rules between business process features and business requirement descriptions, and perform association processing on business process features and business requirement descriptions in the business process feature set based on the feature anchoring rules to obtain feature anchoring results; The feature anchoring results are input into a pre-trained machine learning model to calculate shared adaptation parameters. A shared execution link for business process features is constructed based on shared adaptation parameters. The shared execution link includes a feature extraction node, a feature processing node, and a feature delivery node. The feature extraction node corresponds to the storage location of the business process feature set. The feature processing node processes the business process features based on the shared adaptation parameters. The feature delivery node corresponds to the feature receiving interface of the feature calling subject. The business process characteristics output by the shared execution link are associated and mapped with the talent service business system of the feature calling entity, generating a feature sharing result that includes business process characteristic identifier, feature calling entity identifier, shared adaptation parameters and business system interface identifier. The feature sharing result is then fed back to the talent service business system of the feature calling entity to support business decision-making operations.

[0007] Furthermore, embodiments of the present invention also provide a business data feature sharing system based on machine learning, characterized in that it includes: A processor; a machine-readable storage medium for storing machine-executable instructions of the processor; wherein the processor is configured to execute the aforementioned machine-based business data feature sharing method by executing the machine-executable instructions.

[0008] In another aspect, embodiments of the present invention also provide a computer program product, the computer program product including machine-executable instructions, the machine-executable instructions being stored in a computer-readable storage medium, a processor of a computer device reading the machine-executable instructions from the computer-readable storage medium, the processor executing the machine-executable instructions, causing the computer device to execute the above-described business data feature sharing method based on machine learning.

[0009] Based on the above, by comprehensively acquiring the feature set of business processes in the talent service field and the business requirement description of the feature caller, the feature set of business processes covers core personnel characteristics, job-related characteristics, and service process characteristics, comprehensively reflecting the key information of talent service business from multiple dimensions. The business requirement description of the feature caller clarifies the type range, timeliness requirements, and application direction of the required features, making feature sharing more targeted and purposeful. The constructed feature anchoring rules realize the effective association between business process features and business requirement descriptions. The feature anchoring results obtained through association processing are used to calculate sharing adaptation parameters using a pre-trained machine learning model, giving full play to the intelligent analysis and processing capabilities of machine learning. It can dynamically generate the optimal sharing adaptation parameters according to different business needs and feature situations, improving the flexibility and adaptability of feature sharing. The business process feature sharing execution link constructed based on the sharing adaptation parameters includes feature extraction, processing, and delivery nodes. This business process feature sharing execution link can accurately extract business process features according to the needs of the feature caller, perform targeted processing, and finally deliver the processed features to the feature caller, ensuring the efficiency and accuracy of feature sharing. By associating and mapping the business process characteristics output by the shared execution link with the talent service business system of the characteristic calling entity, detailed characteristic sharing results are generated and fed back to the business system, realizing efficient and accurate sharing of business data characteristics in the talent service field, and effectively improving the quality and efficiency of talent service business. Attached Figure Description

[0010] Figure 1 This is a schematic diagram of the execution flow of the business data feature sharing method based on machine learning provided in an embodiment of the present invention.

[0011] Figure 2 This is a schematic diagram of exemplary hardware and software components of a machine learning-based business data feature sharing system provided in an embodiment of the present invention. Detailed Implementation

[0012] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 This is a flowchart illustrating a machine learning-based business data feature sharing method according to an embodiment of the present invention. The following is a detailed description of the machine learning-based business data feature sharing method.

[0013] Step S110: Obtain the set of business process features in the talent service field and the business requirement description of the feature calling subject. The set of business process features includes core personnel features, job-related features, and service process features. The core personnel features are generated based on personnel's ability representation information, qualification certification information, and service preference information in the talent service field. The job-related features are generated based on job requirement definition information, responsibility boundary information, and adaptation standard information in the talent service field. The service process features are generated based on business interaction record information, service cycle record information, and feedback evaluation information in the talent service field. The business requirement description of the feature calling subject includes the type range, timeliness requirements, and application direction of the business data features required by the feature calling subject.

[0014] In the talent service field, taking a talent service platform as an example, this platform needs to provide business data feature sharing services to its corporate clients (i.e., the feature callers). First, the platform needs to acquire a set of features for each business process. Core personnel features are generated based on job seekers' information on the platform, such as extracting competency information from resumes, including professional skills like Java programming and project management, certifications like PMP and computer proficiency certificates, and service preferences like desired work location and work mode preferences (remote or on-site). Job-related features are generated based on job postings from companies, including requirements such as educational background and years of work experience, job boundary information defining the main tasks and scope of authority, and matching standards such as the degree of match between the job and the job seeker's skills. Service process features originate from business interaction records on the platform, such as communication records between job seekers and companies, interview arrangement records, service cycle records (e.g., the time from job posting to recruitment completion), and feedback and evaluation information including company evaluations of job seekers' interview performance and job seekers' evaluations of the company's recruitment process. Meanwhile, enterprise clients, as the main entities calling for features, will submit a description of their business requirements. For example, the types and ranges of the required features may be "core features of personnel related to software development positions and job-related features", the timeliness requirement is "feature data updated within the last three months", and the application direction is "optimizing the talent screening process for software development positions".

[0015] Step S120: Construct feature anchoring rules between business process features and business requirement descriptions, and perform association processing on the business process features and business requirement descriptions in the business process feature set based on the feature anchoring rules to obtain feature anchoring results.

[0016] In the talent service platform example above, to effectively correlate the acquired business process features with the business needs descriptions of enterprise clients, it is necessary to construct feature anchoring rules and perform correlation processing. The talent service platform first needs to clarify the key information of each business process feature and business need description, and then formulate rules to determine the degree of matching between them. This determines which business process features meet the needs of enterprise clients, forming feature anchoring results for subsequent calculation of shared adaptation parameters.

[0017] Step S121: Extract feature dimension information of each business process feature from the business process feature set. The feature dimension information includes the ability dimension, qualification dimension, and preference dimension of the core personnel feature; the demand dimension, responsibility dimension, and adaptation dimension of the job-related feature; and the interaction dimension, cycle dimension, and evaluation dimension of the service process feature.

[0018] In this talent service platform, the core characteristics of personnel can be further categorized into: ability dimension (including technical skills, communication skills, and learning ability); qualification dimension (including educational level, professional certificate type, and industry certification level); and preference dimension (including preferred work location region, preferred work time type (e.g., flexible work, fixed hours), and preferred corporate culture). Job-related characteristics include educational requirements (e.g., bachelor's degree or above), professional requirements (e.g., computer-related majors), and work experience requirements (e.g., three years or more of relevant experience); responsibility dimension (including main job duties (e.g., responsible for software module development), collaborating departments (e.g., collaboration with product and testing departments), and reporting to (e.g., technical supervisor); and fit dimension (e.g., skill matching threshold (e.g., 80% or higher match) and personality trait fit requirements (e.g., teamwork type). The interaction dimensions of service process characteristics include communication frequency (e.g., average number of daily communications), communication channels (e.g., email, instant messaging tools), and communication content type (e.g., interview invitations, salary negotiations); the cycle dimension involves the recruitment cycle duration (e.g., the number of days from posting to hiring) and the time spent in each stage (e.g., the time spent on resume screening, the time spent on the interview stage); the evaluation dimension includes the company's evaluation level of job seekers (e.g., excellent, good, average) and the job seeker's satisfaction rating of the company (e.g., a five-star rating). By extracting these characteristic dimensions, we can clearly understand the specific composition of the characteristics of each business process.

[0019] Step S122: Extract requirement dimension information, requirement timeliness information, and requirement application information from the business requirement description of the feature calling subject. The requirement dimension information corresponds to the dimension type of the feature required in the business requirement description, the requirement timeliness information corresponds to the effective time range of the feature required in the business requirement description, and the requirement application information corresponds to the usage scenario direction of the feature in the business requirement description.

[0020] Enterprise clients may submit business requirements that include: "Technical skills and qualifications from the core characteristics of personnel related to software development positions, updated within the last three months, as well as requirement and suitability data from the job-related characteristics, to optimize the talent selection process for software development positions." The required dimensions extracted from this description are the technical skills and qualifications from the core characteristics of personnel, and the requirement and suitability dimensions from the job-related characteristics; the timeliness information is "updated within the last three months"; and the application information is "optimizing the talent selection process for software development positions."

[0021] Step S123: Formulate feature dimension matching conditions based on feature dimension information and demand dimension information. When the feature dimension information of the business process features and the demand dimension information meet the overlap ratio requirement or mapping relationship requirement, it is determined that the feature dimension matching conditions are met. The feature dimension matching conditions include the overlap ratio requirement of the same dimension type and the mapping relationship requirement of the related dimension type.

[0022] In this talent service platform, there are requirements for the overlap ratio of the same dimension types. For example, it is stipulated that when the number of identical feature dimension types of business process characteristics and requirement dimension types accounts for more than 70% of the total number of requirement dimension types, the overlap ratio requirement is met. Assuming there are four requirement dimension types (technical ability, qualifications, requirements, and suitability), if three of the feature dimension types of a certain business process characteristic are identical to these four types, the overlap ratio is 75%, meeting the 70% threshold requirement. Regarding the mapping relationship requirements for related dimension types, the platform will pre-define some relationships between dimensions. For example, there is a mapping relationship between the "technical ability dimension" of a core personnel characteristic and the "requirement dimension" of a job-related characteristic. When such related dimension types exist in a business process characteristic, the mapping relationship requirement is considered met.

[0023] Step S124: Formulate feature timeliness alignment conditions based on the generation time information of business process features and the timeliness information of demand. The feature timeliness alignment conditions include that the generation time of the business process feature is within the effective time range corresponding to the timeliness information of demand, or that the update frequency of the business process feature is consistent with the timeliness requirement corresponding to the timeliness information of demand. When the business process feature meets any of the above conditions, it is determined that it meets the feature timeliness alignment conditions.

[0024] The generation time information of business process features can be obtained from the metadata of the feature data. For example, if the generation time of a person's core feature is March 15, 2024, and the timeliness information of the enterprise customer's requirement is "updated within the last three months" (assuming the current time is June 10, 2024), then March 15, 2024 falls within the valid time range of March 10, 2024 to June 10, 2024, satisfying the condition that the generation time is within the valid time range. Regarding the update frequency, if the timeliness information of the requirement requires the feature data to be updated once a month, and the update frequency of a certain business process feature is set to be updated monthly, then this feature satisfies the condition that the update frequency is consistent with the timeliness requirement, meeting the feature timeliness alignment condition.

[0025] Step S125: Formulate feature application adaptation conditions based on the application scope information and application requirement information of business process features. The feature application adaptation conditions include whether the application scope information of business process features covers the usage scenario direction corresponding to the application requirement information, or whether the application method of business process features is compatible with the usage method corresponding to the application requirement information. When the business process features meet any of the above conditions, it is determined that they meet the feature application adaptation conditions.

[0026] The application scope of business process features is labeled during feature generation. For example, if the application scope of a person's core feature is labeled as "talent screening and job matching," and the enterprise customer's application requirement is "optimizing the talent screening process for software development positions," then the application scope of the business process feature covers the usage scenario of the application requirement information, meeting the application compatibility criteria. Alternatively, if the application method of the business process feature is "providing feature data for algorithm model training to support decision-making," and the corresponding usage method of the application requirement information is also through algorithm model-based talent screening decisions, then the two application methods are compatible and also meet the application compatibility criteria.

[0027] Step S126: Integrate the feature dimension matching conditions, feature timeliness alignment conditions, and feature application adaptation conditions into feature anchoring rules, and determine the judgment logic and combination method of each condition.

[0028] The aforementioned feature dimension matching conditions, feature timeliness alignment conditions, and feature application adaptation conditions are integrated into feature anchoring rules. The judgment logic is as follows: First, determine whether the business process feature meets the feature dimension matching condition. If it does, then determine whether it meets the feature timeliness alignment condition. Finally, check whether it meets the feature application adaptation condition. Only when all three conditions are met can it be determined that the business process feature and the business requirement description are successfully anchored. The combination method uses "AND" logic, meaning that all three conditions must be met simultaneously.

[0029] Step S127: Based on the feature anchoring rules, associate each business process feature in the business process feature set with the business requirement description one by one, record the judgment results that meet each condition, bind the business process features that meet all conditions with the business requirement description, and generate feature anchoring results. The feature anchoring results include the bound business process feature identifier, the business requirement description identifier, and the judgment result details of each condition.

[0030] The platform system iterates through each feature in the feature set of business processes. Taking a person's core features as an example, the feature dimensions include technical ability, communication ability, and qualifications, while the requirement dimensions are technical ability and qualifications. First, it checks the feature dimension matching conditions. The technical ability and qualification dimensions of this person's core features overlap with the requirement dimensions, with an overlap ratio of 2 / 2 = 100%, meeting the overlap ratio requirement and thus meeting the feature dimension matching conditions. The result is recorded as "Compliant". Next, it checks the generation time, which is April 20, 2024. Within the last three months, it meets the feature timeliness alignment condition, and the result is recorded as "Compliant". Then, it checks the application scope information, which includes "talent screening," covering the requirement application information "optimizing the talent screening process for software development positions," meeting the feature application adaptation condition, and the result is recorded as "Compliant". Since all three conditions are met, the core characteristics of the personnel are bound to the business requirement description to generate feature anchoring results. The business process feature identifier is the unique code of the feature in the system, and the business requirement description identifier is the number of the enterprise customer's requirement. The details of the judgment results of each condition are recorded as "Feature dimension matching condition: Meets (overlap ratio 100%)", "Feature timeliness alignment condition: Meets (generation time within the last three months)", and "Feature application adaptation condition: Meets (application scope covers the application scenario of the requirement)".

[0031] Step S1231: Classify the feature dimension information of business process features and the demand dimension information of feature calling subjects by dimension type. Classify the ability dimension, qualification dimension, and preference dimension of personnel core features, the demand dimension, responsibility dimension, and adaptation dimension of job-related features, and the interaction dimension, cycle dimension, and evaluation dimension of service process features into first-level dimension types.

[0032] Within the platform system, the core characteristics of personnel—capability, qualification, and preference—are explicitly categorized into primary dimension types, named "Core Personnel - Capability Dimension," "Core Personnel - Qualification Dimension," and "Core Personnel - Preference Dimension," respectively. Job-related characteristics—requirement, responsibility, and suitability—are categorized into "Job-Related - Requirement Dimension," "Job-Related - Responsibility Dimension," and "Job-Related - Suitability Dimension." Service process characteristics—interaction, cycle, and evaluation—are categorized into "Service Process - Interaction Dimension," "Service Process - Cycle Dimension," and "Service Process - Evaluation Dimension." This categorization makes the dimension types clearer and facilitates subsequent matching processing.

[0033] Step S1232: Divide the primary dimension type into secondary dimension subtypes. For example, the ability dimension of the core characteristics of personnel includes the professional skills sub-dimension and the general ability sub-dimension; the requirement dimension of the job-related characteristics includes the education requirement sub-dimension and the experience requirement sub-dimension; and the interaction dimension of the service process characteristics includes the communication frequency sub-dimension and the collaboration method sub-dimension.

[0034] For the primary dimension type "Core Personnel - Competency Dimension," it is further divided into secondary sub-divisions. The professional skills sub-division includes programming language skills (such as Java and Python) and framework usage skills (such as Spring and Django); the general competency sub-division includes teamwork ability, problem-solving ability, and time management ability. In the secondary sub-divisions of "Job Relevance - Requirements Dimension," the education requirement sub-division includes different levels such as bachelor's, master's, and doctoral degrees; the experience requirement sub-division includes recent graduates, 1-3 years of experience, and 3-5 years of experience. The communication frequency sub-division of "Service Process - Interaction Dimension" can be divided into high-frequency (multiple times daily), medium-frequency (once daily), and low-frequency (several times weekly); the collaboration method sub-division includes online collaboration, offline collaboration, and hybrid collaboration.

[0035] Step S1233: Calculate the number of overlaps between the business process features and the second-level sub-types of the demand dimension information under the same first-level dimension type, and take the ratio of the number of overlaps to the total number of second-level sub-types of the demand dimension information as the overlap ratio of the same dimension type.

[0036] The second-level sub-dimensions under the "Core Personnel - Technical Capability" dimension in the enterprise customer needs dimension information include Java programming skills and project management capabilities. The second-level sub-dimensions under the "Core Personnel - Capability" dimension for a certain business process feature include Java programming skills, Python programming skills, and project management capabilities. Therefore, under the same first-level dimension type, the number of overlapping second-level sub-types is 2 (Java programming skills, project management capabilities). The total number of second-level sub-types in the needs dimension information is 2, resulting in an overlap ratio of 2 / 2 = 100%.

[0037] Step S1234: Set an overlap ratio threshold. When the overlap ratio of the same dimension type is greater than the overlap ratio threshold, it is determined that the overlap ratio requirement of the same dimension type is met.

[0038] The platform system sets the overlap ratio threshold to 70%. In the example above, the overlap ratio is 100%, which is greater than 70%, therefore it is determined that the overlap ratio requirement for the same dimension type is met. If the overlap ratio of another business process feature is 60%, then this requirement is not met.

[0039] Step S1235: Establish a mapping relationship table for related dimension types. The mapping relationship table includes the mapping relationship between the ability dimension of personnel core characteristics and the demand dimension of job-related characteristics, the mapping relationship between the preference dimension of personnel core characteristics and the interaction dimension of service process characteristics, and the mapping relationship between the adaptation dimension of job-related characteristics and the evaluation dimension of service process characteristics. Each mapping relationship includes the related fields and related weights between dimensions.

[0040] The platform system establishes a mapping table for related dimension types. For example, the "Java programming skills" field in the ability dimension of personnel core characteristics is mapped to the "Java development experience requirement" field in the demand dimension of job-related characteristics, with a correlation weight of 0.8; the "remote work preference" field in the preference dimension of personnel core characteristics is mapped to the "online collaboration method" field in the interaction dimension of service process characteristics, with a correlation weight of 0.7; and the "skill matching degree" field in the adaptation dimension of job-related characteristics is mapped to the "interview evaluation level" field in the evaluation dimension of service process characteristics, with a correlation weight of 0.9.

[0041] Step S1236: Check whether there is a related dimension type in the mapping relationship table between the feature dimension information and the requirement dimension information of the business process features. If there is a related dimension type, calculate the matching quantity and the weighted value of the related field respectively. The matching quantity and the related weight are judged according to the preset independent threshold. When the matching quantity meets the quantity threshold requirement and the related weight meets the weight threshold requirement, it is determined that the mapping relationship requirement of the related dimension type is met.

[0042] Assume the requirement dimension information includes a "Java development experience required" field from the job-related characteristic requirement dimension, and the business process characteristic feature dimension information includes a "Java programming skills" field from the personnel core characteristic capability dimension. These two fields exist in the mapping relationship table and belong to the related dimension type. The matching quantity of the related field is 1, and the preset quantity threshold is 1, which meets the quantity threshold requirement. The related weight is 0.8, and the preset weight threshold is 0.6, which meets the weight threshold requirement. Therefore, it is determined that the mapping relationship requirements for the related dimension type are met.

[0043] Step S12351: Collect historical business data in the talent service field. The historical business data includes past records of matching personnel with positions, interaction records during the service process, and service evaluation records.

[0044] The platform system collects historical business data from the past five years, including tens of thousands of matching records between personnel and positions. Each record contains the job seeker's ability characteristics, the job's requirements, and the matching results. It also includes a large number of service process interaction records, such as communication logs between job seekers and company HR, interview arrangement records, etc., as well as numerous service evaluation records, covering company evaluations of job seekers and job seekers' feedback to companies.

[0045] Step S12352: Extract the correspondence between the capability dimension fields of personnel core characteristics and the demand dimension fields of job-related characteristics from historical business data, count the frequency of the same capability dimension fields and demand dimension fields appearing together, use the frequency as the association weight, and establish the mapping relationship between the capability dimension of personnel core characteristics and the demand dimension of job-related characteristics.

[0046] From historical business data, the "Java programming skills" capability dimension field and the "Java development experience requirement" requirement dimension field both appeared in 5000 matching records, and the "Python programming skills" and "Python development experience requirement" fields both appeared in 3000 matching records. These frequencies were used as association weights to establish mapping relationships, such as (Core Personnel - Capability Dimension: Java Programming Skills, Job Related - Requirement Dimension: Java Development Experience Requirement, Association Weight: 5000), (Core Personnel - Capability Dimension: Python Programming Skills, Job Related - Requirement Dimension: Python Development Experience Requirement, Association Weight: 3000), etc.

[0047] Step S12353: Extract the correspondence between the preference dimension field of the core characteristics of personnel and the interaction dimension field of the service process characteristics from historical business data, analyze the degree of influence of the preference dimension field on the interaction dimension field, quantify the degree of influence into association weight, and establish the mapping relationship between the preference dimension of the core characteristics of personnel and the interaction dimension of the service process characteristics.

[0048] Analysis of historical data revealed that when the preference dimension field of the core personnel characteristic is "remote work preference", the interaction dimension field of the service process characteristic is "online collaboration method" in a very high proportion, and the influence is relatively large, which is quantified as a correlation weight of 0.7; while the influence of "on-site work preference" corresponding to "offline collaboration method" is quantified as 0.65, thus establishing a corresponding mapping relationship.

[0049] Step S12354: Extract the correspondence between the adaptation dimension field of the job-related features and the evaluation dimension field of the service process features from historical business data, calculate the correlation coefficient between the adaptation dimension field and the evaluation dimension field, use the correlation coefficient as the association weight, and establish the mapping relationship between the adaptation dimension of the job-related features and the evaluation dimension of the service process features.

[0050] By calculating the correlation coefficient between the "skill matching degree" field of the job-related feature and the "interview evaluation level" field of the service process feature in historical data, the correlation coefficient was found to be 0.85. This value was used as the association weight to establish a mapping relationship (job-related - matching dimension: skill matching degree, service process - evaluation dimension: interview evaluation level, association weight: 0.85).

[0051] Step S12355: Integrate the established mapping relationships. Each mapping relationship entry includes the source dimension type, source dimension field, target dimension type, target dimension field, and associated weight, forming a mapping relationship table of associated dimension types.

[0052] The various mapping relationships established above are integrated to form the following mapping relationship table entries: (Source dimension type: core personnel - competency dimension, source dimension field: Java programming skills, target dimension type: job-related - requirement dimension, target dimension field: Java development experience required, association weight: 5000) (Source dimension type: core personnel - preference dimension, source dimension field: remote work preference, target dimension type: service process - interaction dimension, target dimension field: online collaboration method, association weight: 0.7) (Source dimension type: Job association - Adaptation dimension, Source dimension field: Skill matching degree, Target dimension type: Service process - Evaluation dimension, Target dimension field: Interview evaluation level, Association weight: 0.85) This process continues until a complete mapping table is formed.

[0053] Step S12356: Update the mapping relationship table based on the latest talent service business data, regularly count the corresponding relationships and changes in the association weights of each dimension field in the new business data, and adjust the association weight values ​​in the mapping relationship table so that the mapping relationship table remains adapted to the current talent service business scenario.

[0054] The platform system updates the mapping table monthly based on newly generated business data. For example, if the frequency of "Go programming skills" and "Go development experience requirements" appearing together increases to 2000 times in a new month, exceeding some fields in historical data, the association weight of this mapping relationship needs to be updated to 2000. Simultaneously, by analyzing changes in the correspondence between fields of each dimension in the new data, if it is found that the influence of "remote work preference" on "online collaboration method" decreases, the association weight may be adjusted from 0.7 to 0.65.

[0055] Step S130: Input the feature anchoring results into the pre-trained machine learning model to calculate the shared adaptation parameters and obtain the shared adaptation parameters.

[0056] In the talent service platform, after feature anchoring is completed, the feature anchoring results, including business process feature identifiers, business requirement description identifiers, and details of each condition judgment result, are input into a pre-trained machine learning model. This machine learning model is trained based on a large amount of historical feature sharing case data and can calculate sharing adaptation parameters suitable for the current scenario based on the input feature anchoring results. These parameters will be used to guide the subsequent sharing execution of features in business processes.

[0057] Step S131: Encode the business process feature identifier, business requirement description identifier, and judgment result details of each condition in the feature anchoring result, converting the business process feature identifier into feature code, the business requirement description identifier into requirement code, and the judgment result details of each condition into judgment code.

[0058] Business process feature identifiers might be strings like "PERSON_FEATURE_001," which are converted into fixed-length numerical vectors as feature codes using an encoding algorithm. Business requirement description identifiers, such as "DEMAND_123," are also converted into requirement codes. Details of the judgment results for each condition, such as "Meets feature dimension matching conditions (overlap ratio 100%)" and "Meets feature timeliness alignment conditions," are encoded. "Meets" is converted to 1, and "Does not meet" is converted to 0. Specific numerical values, such as the overlap ratio, are also encoded to form the judgment code.

[0059] Step S132: Input the feature encoding, demand encoding and decision encoding into the feature input layer of the machine learning model, perform dimension unification processing and feature concatenation operation to obtain a comprehensive feature vector.

[0060] After receiving the feature encoding, requirement encoding, and decision encoding, the feature input layer first performs dimensionality unification to ensure that each encoded vector has the same dimension. For example, if the feature encoding is 100-dimensional, the requirement encoding is 80-dimensional, and the decision encoding is 50-dimensional, they are all unified to 100 dimensions through dimensionality expansion or compression techniques. Then, the three encoded vectors with unified dimensions are concatenated to form a 300-dimensional comprehensive feature vector.

[0061] Step S133: Input the comprehensive feature vector into the association processing layer of the machine learning model. The association processing layer contains a multi-layer perceptron structure. The first layer perceptron performs linear transformation and activation function processing on the comprehensive feature vector to extract the basic association features in the comprehensive feature vector and output the basic association feature vector.

[0062] The composite feature vector is input into the first perceptron layer of the association processing layer, which contains multiple neurons. A linear transformation is performed on the composite feature vector, i.e., multiplying the input vector by the weight matrix and adding a bias term, followed by processing with the ReLU activation function. The result is output as the basic association feature vector. For example, a 300-dimensional composite feature vector, after processing by the first perceptron layer, outputs a 200-dimensional basic association feature vector, which contains the basic association information between the components of the composite feature vector.

[0063] Step S134: Input the basic association feature vector into the second layer perceptron. The second layer perceptron performs nonlinear transformation and feature interaction processing on the basic association feature vector to explore the adaptation relationship between the basic association feature vector and the talent service business scenario. The adaptation relationship includes feature usage frequency association, feature update cycle association, and feature application effect association.

[0064] The second-layer perceptron also contains multiple neurons, performing nonlinear transformations on the 200-dimensional basic correlation feature vector. This is achieved through complex calculations using multi-layer neural networks, along with feature interaction processing, combining and associating features from different dimensions. In this process, the adaptability and correlation between the basic correlation feature vector and talent service business scenarios are uncovered. For example, which features are frequently used in talent screening for software development positions, the relationship between the update cycle of these features and the screening effect, and the impact of feature application on screening accuracy, efficiency, and other application effects are analyzed.

[0065] Step S135: Set up a feature attention mechanism in the association processing layer. Use the feature attention mechanism to assign weights to the mined adaptation associations, so that the adaptation associations with higher correlation to the shared adaptation parameters get higher weights, and generate an adaptation association vector with attention weights.

[0066] The feature attention mechanism analyzes the correlation between each adaptation relationship and the shared adaptation parameters. For example, if the feature usage frequency correlation is highly correlated with the feature open range parameter in the shared adaptation parameters, then this correlation will be assigned a high weight; similarly, the feature application effect correlation, which has a significant impact on the feature application permission parameter, will also receive a high weight. Through this weight allocation, an adaptation relationship vector with attention weights is generated, highlighting important correlations.

[0067] Step S136: Input the adaptation association vector with attention weights into the parameter output layer of the machine learning model, and generate shared adaptation parameters for the business process features to the feature calling subject through regression calculation. The shared adaptation parameters include feature open range parameters, feature transmission timeliness parameters, and feature application permission parameters. The feature open range parameters define the range of business process feature fields that can be shared, the feature transmission timeliness parameters define the transmission time requirements of business process features, and the feature application permission parameters define the permission level of the feature calling subject to use the business process features.

[0068] After receiving the adaptation association vector with attention weights, the parameter output layer generates shared adaptation parameters based on the information in the vector through regression calculation. For example, the feature open scope parameter may be defined as "the technical ability field and qualification certificate field in the core features of open personnel, and the requirement description field and adaptation standard field in the job-related features"; the feature transmission timeliness parameter is defined as "the feature transmission shall be completed within 2 hours"; and the feature application permission parameter is set as "can be viewed and used to filter model training, but the original feature data cannot be modified".

[0069] Step S1351: Extract all the adaptation relationships mined by the association processing layer. The adaptation relationships include feature usage frequency association, feature update cycle association, feature application effect association, feature sensitivity association, and feature call cost association.

[0070] In addition to the feature usage frequency association, feature update cycle association, and feature application effect association mentioned above, the adaptation associations mined from the association processing layer also include feature sensitivity association (such as the sensitivity level of certain personal information features) and feature call cost association (the resource cost required to obtain and process the feature).

[0071] Step S1352: Construct an adaptation association weight evaluation function. The adaptation association weight evaluation function is generated based on the degree of influence of the adaptation association on the shared adaptation parameters. The degree of influence includes the influence coefficient of the adaptation association on the feature open range parameter, the influence coefficient on the feature transmission timeliness parameter, and the influence coefficient on the feature application permission parameter.

[0072] The adaptation association weight evaluation function considers the degree of influence of each adaptation association on three shared adaptation parameters. For example, the feature sensitivity association has a higher influence coefficient on the feature open range parameter and a lower influence coefficient on the feature transmission timeliness parameter; the feature usage frequency association has a higher influence coefficient on the feature transmission timeliness parameter. The function form may be: Weight value = Influence coefficient 1 × Influence degree of feature open range + Influence coefficient 2 × Influence degree of feature transmission timeliness + Influence coefficient 3 × Influence degree of feature application permission, where influence coefficients 1, 2, and 3 are set according to the actual business scenario.

[0073] Step S1353: Input each adaptation association into the adaptation association weight evaluation function, calculate the weight value corresponding to each adaptation association, the magnitude of the weight value is positively correlated with the total influence coefficient of the adaptation association on the shared adaptation parameters, and the total influence coefficient is the weighted sum of the three influence coefficients.

[0074] The sensitivity of features is correlated with the evaluation function. Assuming its influence coefficient on the feature open range parameter is 0.6, its influence coefficient on the feature transmission timeliness parameter is 0.2, and its influence coefficient on the feature application permission parameter is 0.5, the weighted sum of these three influence coefficients is 1.3, and the corresponding weight values ​​are calculated. Other adaptation relationships are weighted in the same way.

[0075] Step S1354: Normalize the weight values ​​of all adaptation relationships to obtain the attention weight of each adaptation relationship.

[0076] The weights of all fitting relationships are summed to obtain a total. Then, each weight value is divided by the total to make the sum of the normalized attention weights equal to 1. For example, if the total weight of all fitting relationships is 5 and the weight of a certain fitting relationship is 1.3, then its attention weight is 1.3 / 5 = 0.26.

[0077] Step S1355: Multiply the feature vector corresponding to each adaptation relationship with the attention weight of that adaptation relationship to obtain the weighted feature vector of each adaptation relationship.

[0078] Each adaptation relationship has a corresponding feature vector. For example, the feature vector associated with the frequency of use contains information such as the frequency of use and its fluctuation. This vector is multiplied by the attention weight. For example, if the attention weight is 0.2 and the feature vector is [2, 3, 4], then the weighted feature vector is [0.4, 0.6, 0.8].

[0079] Step S1356: Concatenate all weighted feature vectors of the adaptation relationships according to the original order of the adaptation relationships to generate an adaptation relationship vector with attention weights.

[0080] The weighted feature vectors obtained above are concatenated according to their original order in the adaptation association. For example, the weighted feature vector associated with feature usage frequency is concatenated first, then the weighted feature vector associated with feature update cycle is concatenated, and so on, to form a complete adaptation association vector with attention weights.

[0081] Step S1357: Compare the adaptation association vector with attention weights with the original adaptation association vector, retain the feature dimensions in the adaptation association vector with attention weights whose values ​​are greater than the corresponding values ​​in the original adaptation association vector, and strengthen the representation ability of adaptation associations that have a significant impact on shared adaptation parameters.

[0082] By comparing the values ​​of each feature dimension of the two vectors, if the value of the vector with attention weights in a certain dimension is greater than that of the original vector, then the value of that dimension is retained; otherwise, the value of that dimension is set to 0. This method highlights the features that have a more significant impact on the shared adaptation parameters and their corresponding adaptation relationships.

[0083] Step S140: Construct a shared execution link for business process features based on shared adaptation parameters. The shared execution link includes a feature extraction node, a feature processing node, and a feature delivery node. The feature extraction node corresponds to the storage location of the business process feature set. The feature processing node processes the business process features based on the shared adaptation parameters. The feature delivery node corresponds to the feature receiving interface of the feature calling subject.

[0084] Based on the shared adaptation parameters output by the machine learning model, the talent service platform begins to build a shared execution link for business process features. This link acts like a dedicated channel, extracting the selected business process features from storage, processing them, and accurately delivering them to the feature receiving interface of enterprise clients, ensuring that the feature sharing process is efficient and accurate.

[0085] Step S141: Determine the feature fields of the business process to be shared based on the feature open range parameter in the shared adaptation parameters, extract the storage location of the feature fields of the business process to be shared in the business process feature set, and determine the storage location as the feature extraction node of the shared execution link. The feature extraction node contains the storage address, storage format and reading method of the feature fields of the business process to be shared.

[0086] The feature open scope parameter in the shared adaptation parameters determines the fields to be shared, such as "technical ability fields and qualification certificate fields in the core personnel features, and requirement description fields and adaptation standard fields in the job-related features." Based on this field information, the platform system locates their storage locations in the database of the business process feature set. Assume the technical ability field is stored in the "person_features" table on database server DB1, in JSON format, and retrieved via SQL queries; the qualification certificate field is stored in the "qualification_info" table on DB1, in XML format, and retrieved via API calls. Integrating these storage addresses, formats, and retrieval methods determines the feature extraction node.

[0087] Step S142: Construct feature processing rules based on the feature open range parameter in the shared adaptation parameters. The feature processing rules include filtering rules, format conversion rules, and content integration rules for feature fields of the business process to be shared. The filtering rules filter the fields to be shared based on the field range defined by the feature open range parameter. The format conversion rules convert the fields to be shared into a format that can be accepted by the feature calling subject. The content integration rules integrate the scattered fields to be shared into structured data.

[0088] Based on the feature open scope parameters, the filtering rules are set to retain only the technical capability field, qualification certificate field, requirement description field, and adaptation standard field, excluding other irrelevant fields. The format conversion rules require that the JSON-formatted technical capability field and the XML-formatted qualification certificate field be uniformly converted to the CSV format acceptable to enterprise clients. The content integration rules stipulate that fields from different tables should be integrated according to the structure of "Personnel ID - Technical Capability - Qualification Certificate - Job Requirement Description - Adaptation Standard" to form structured data.

[0089] Step S143: The processing module constructed based on the feature processing rules is determined as the feature processing node of the shared execution link. The feature processing node receives the features of the business link to be shared from the feature extraction node, processes them according to the feature processing rules, and outputs the processed business link features.

[0090] Based on the aforementioned feature processing rules, a processing module is constructed, comprising a filtering unit, a format conversion unit, and a content integration unit, which is designated as the feature processing node. The feature extraction node transmits the extracted business process features to be shared to the feature processing node. The filtering unit first filters out the specified fields, the format conversion unit converts the field format to CSV, and the content integration unit integrates the data according to the prescribed structure, finally outputting the processed structured business process feature data.

[0091] Step S144: Based on the feature transmission timeliness parameter in the shared adaptation parameters and the feature receiving interface information of the feature calling subject, determine the feature delivery node of the shared execution link. The feature delivery node includes the address of the feature receiving interface, the interface protocol, the data receiving format and timeliness requirements. The feature transmission timeliness parameter defines the transmission time limit from the feature processing node to the feature delivery node.

[0092] The feature transmission timeliness parameter is "transmission to be completed within 2 hours". The enterprise customer's feature receiving interface information includes the interface address "http: / / enterprise.example.com / feature / receive", the interface protocol is HTTP, the data receiving format is CSV, and the timeliness requirement is "response within 10 minutes of receipt". Based on this information, the feature delivery node is determined, and the transmission time limit is clearly defined as 2 hours.

[0093] Step S145: Establish the connection relationship between the feature extraction node, the feature processing node, and the feature delivery node. Read the business process features to be shared from the feature extraction node, transmit them to the feature processing node for processing, and then transmit the processed business process features to the feature delivery node to form a complete shared execution link.

[0094] A connection is established between the three nodes through the platform's internal communication mechanism. First, the feature extraction node reads the features to be shared from the storage location using a read-only method, and then transmits the data to the input port of the feature processing node via the internal network. After processing, the feature processing node transmits the data to the interface address corresponding to the feature delivery node through a channel conforming to the HTTP protocol, thus forming a complete shared execution chain.

[0095] Step S146: Associate the link information of the shared execution link with the features of the business process to be shared and the sharing adaptation parameters. The link information includes the feature extraction node identifier, feature processing node identifier, feature delivery node identifier and link transmission requirements, and generate a shared execution link association table.

[0096] The shared execution link association table contains link information, feature identifiers of the service segments to be shared, and sharing adaptation parameters. For example, the feature extraction node identifier in the link information is "EXTRACT_NODE_001", the feature processing node identifier is "PROCESS_NODE_002", the feature delivery node identifier is "DELIVER_NODE_003", and the link transmission requirement is "bandwidth not less than 10Mbps"; the feature identifiers of the service segments to be shared are "FEATURE_001, FEATURE_002"; and the sharing adaptation parameters are the previously calculated parameters. Linking the above information to generate the association table facilitates link management and monitoring.

[0097] Step S1421: Parse the feature open range parameter in the shared adaptation parameters, and extract the field list of the business process features to be shared as defined by the feature open range parameter. The field list includes specified fields of core personnel features, specified fields of job-related features, and specified fields of service process features.

[0098] Parse the feature open scope parameters to obtain the field list of features of the business process to be shared: the specified fields for the core personnel features are "technical ability field and qualification certificate field", the specified fields for the job-related features are "requirement description field and adaptation standard field", and the service process features have no specified fields in this example, so the field list is empty.

[0099] Step S1422: Construct filtering rules based on the field list. The filtering rules include field name matching logic and field attribute matching logic. The field name matching logic is used to filter business process feature fields whose names are consistent with those in the field list. The field attribute matching logic is used to filter business process feature fields whose attributes are consistent with those in the field list. The business process feature fields to be shared are filtered from the business process feature set through the field name matching logic and the field attribute matching logic.

[0100] The field name matching logic of the filtering rules is set to match the field names in the field list exactly. For example, the "Technical Capability Field" can only match fields with the exact same name. The field attribute matching logic requires that the data type, length, and other attributes of the field be consistent with those defined in the field list. For example, if the data type of the "Technical Capability Field" is a string and the length does not exceed 200 characters, then only fields that meet these attributes will be retained during filtering. By combining these two logics, the fields to be shared can be accurately filtered from the set of business process features.

[0101] Step S1423: Obtain the feature receiving format requirements of the feature calling subject. The format requirements include data structure format, field type definition and data encoding method. Based on the format requirements, construct format conversion rules. The format conversion rules include field type conversion logic, data structure reorganization logic and data encoding conversion logic. The field type conversion logic converts the type of the field to be shared into the type required by the feature calling subject. The data structure reorganization logic reorganizes the structure of the field to be shared into the structure required by the feature calling subject. The data encoding conversion logic converts the encoding of the field to be shared into the encoding required by the feature calling subject.

[0102] Enterprise customers require a tabular data structure with the following field types: technical capability (string), qualification certificate (array), requirement description (string), and adaptation standard (numeric). The data encoding is UTF-8. The format conversion rules convert the integer encoding of the technical capability field to a string. The data structure reorganization logic restructures the nested JSON structure into a flat table structure. The data encoding conversion logic ensures that all fields use UTF-8 encoding.

[0103] Step S1424: Analyze the distribution of characteristic fields of the business process to be shared, determine the fields to be shared and their relationships in scattered storage, and construct content integration rules based on the relationships. The content integration rules include field association logic, data merging logic and redundancy removal logic. The field association logic establishes the relationship between scattered fields based on the related fields between fields. The data merging logic merges the associated scattered fields into a unified data unit. The redundancy removal logic removes duplicate content in the merged data unit.

[0104] The characteristic fields of the business processes to be shared are distributed across different database tables, but they are all linked through "Personnel ID" and "Job ID". The field association logic of the content integration rules links the scattered fields based on "Personnel ID" and "Job ID"; the data merging logic merges the associated fields into a single data unit, with each data unit corresponding to the relevant characteristics of a person and a job; the redundancy removal logic checks the merged data units and removes duplicate information, such as retaining only one record of the same qualification certificate information for the same person.

[0105] Step S1425: Integrate the filtering rules, format conversion rules, and content integration rules into a complete feature processing rule, determine the execution order and execution conditions of each rule, execute the filtering rules first, execute the format conversion rules after the filtering rules, and execute the content integration rules after the format conversion rules.

[0106] The execution order of the feature processing rules is as follows: First, the filtering rule is executed to select the fields to be shared from the original data; then, the format conversion rule is executed to convert the filtered fields into the target format; finally, the content integration rule is executed to integrate the format-converted fields into structured data. The next rule is executed only if the previous rule is executed successfully and its output meets the input requirements of the next rule.

[0107] Step S150: Associate and map the business process features output by the shared execution link with the talent service business system of the feature caller, generate a feature sharing result containing the business process feature identifier, the feature caller identifier, the shared adaptation parameters and the business system interface identifier, and feed the feature sharing result back to the talent service business system of the feature caller to support business decision-making operations.

[0108] After sharing the business process characteristics output from the execution chain, the talent service platform needs to map these characteristics to the enterprise client's talent service business system. This is similar to labeling the characteristic data and indicating its destination, ensuring that the enterprise client's business system can accurately identify and use this characteristic data.

[0109] Step S151: Extract the identification information of the business link features output by the shared execution link. The identification information includes the unique code of the business link feature, the generation time, the feature category, and the version information.

[0110] The business process features output by the shared execution link have identification information, such as the unique code "FEATURE_001_V1", the generation time "2024-06-01 10:30:00", the feature category "personnel core features", and the version information "V1.0".

[0111] Step S152: Obtain the identification information of the feature caller, which includes the name, type, registration number in the talent service field, and business scope description of the feature caller.

[0112] The enterprise customer's identification information includes the name "ABC Technology Co., Ltd.", the type "Enterprise", the registration number "ENT_789", and the business scope description "Software development, information technology services".

[0113] Step S153: Determine the interface identifier of the talent service business system of the feature calling subject. The interface identifier includes the interface name, interface address, interface version and data types supported by the interface of the business system.

[0114] The interface name of ABC Technology Co., Ltd.'s talent service business system is "FeatureReceiveAPI", the interface address is "http: / / abc-tech.com / api / feature / receive", the interface version is "V2.0", and the interface supports CSV and JSON data types.

[0115] Step S154: Map the fields of the identification information of the business process feature, the identification information of the feature calling subject, the shared adaptation parameters and the business system interface identifier, and establish the association between the information. The identification information of the business process feature and the shared adaptation parameters are associated through the unique feature code, and the identification information of the feature calling subject and the business system interface identifier are associated through the calling subject name.

[0116] Establish field mapping relationships: The unique code "FEATURE_001_V1" for the business process feature is associated with the feature open range parameter in the shared adaptation parameters; the feature caller name "ABC Technology Co., Ltd." is associated with the interface name "FeatureReceiveAPI" in the business system interface identifier. Other information is also associated in a similar way, such as associating the generation time with the feature transmission timeliness parameter.

[0117] Step S155: Construct a data structure for the feature sharing results based on the association relationship. The data structure includes a business process feature identifier field, a feature call subject identifier field, a shared adaptation parameter field, and a business system interface identifier field. Each field contains corresponding information content and data type definition.

[0118] The constructed data structure is as follows: Business process feature identifier fields include: unique code (string type), generation time (date and time type), feature category (string type), and version information (string type). Feature caller identifier fields: Name (string type), Type (string type), Registration Number (string type), Business Scope Description (string type) Shared adaptation parameter fields: Feature open range parameter (string type), Feature transmission timeliness parameter (integer type, unit: minutes), Feature application permission parameter (string type). Business system interface identifier fields: Interface name (string type), Interface address (string type), Interface version (string type), Supported data types (string type, multiple types separated by commas). Step S156: Fill in the contents of each field in the data structure. The business process feature identifier field is filled with the unique code of the business process feature and the information of the generation time. The feature caller identifier field is filled with the name and registration number of the feature caller. The shared adaptation parameter field is filled with the information of the feature open scope parameter and the feature transmission timeliness parameter. The business system interface identifier field is filled with the interface address and interface version of the business system.

[0119] Example of the filled data structure content: Business process feature identifier field: unique code "FEATURE_001_V1", generation time "2024-06-01 10:30:00" Feature call subject identifier field: Name "ABC Technology Co., Ltd.", Registration Number "ENT_789" Shared adaptation parameter fields: Feature open scope parameters "Technical capability field, Qualification certificate field, Requirement description field, Adaptation standard field", Feature transmission timeliness parameter "120" (minutes) Business system interface identifier fields: Interface address "http: / / abc-tech.com / api / feature / receive", Interface version "V2.0" Step S157: By calling the interface of the talent service business system of the feature call subject, the fully filled feature sharing result is fed back to the business system. After receiving the feature sharing result, the business system calls the features of the business process based on the shared adaptation parameters for business decision-making operations in the talent service field. The business decision-making operations include personnel and position matching decisions, service process optimization decisions, and business resource allocation decisions.

[0120] The platform system sends the fully populated feature sharing results to ABC Technology Co., Ltd.'s talent service business system via the "http: / / abc-tech.com / api / feature / receive" interface. Upon receiving the results, the talent service business system retrieves feature data from various business processes based on the feature application permission parameters in the shared adaptation parameters. For example, it uses the technical skills field and the job requirement description field for matching personnel with positions, calculating the match degree between job seekers' technical skills and job requirements through algorithms; it utilizes data such as interaction records in the service process features to optimize service processes and reduce redundant steps in the recruitment process; and it allocates business resources based on the urgency of job requirements reflected in the feature data, prioritizing more recruitment resources for urgent positions.

[0121] For example, step S1571: Obtain the interface access requirements of the talent service business system of the feature calling subject, wherein the access requirements include authentication requirements, data transmission protocol requirements and timeliness control requirements.

[0122] ABC Technology Co., Ltd.'s interface access requirements include: authentication requirements requiring the provision of an API key and access token assigned by the platform; data transmission protocol requirements requiring the use of HTTPS; and timeliness control requirements requiring data transmission response time not to exceed 30 seconds.

[0123] Step S1572: Generate interface access credentials according to the authentication requirements. The access credentials include the identity code of the feature caller, the access permission identifier, and the validity period of the credentials. Attach the access credentials to the header information of the feature sharing result.

[0124] Based on the authentication requirements, the platform system generates an access credential containing the enterprise customer identity code "ENT_789", the access permission identifier "READ_WRITE", and the credential validity period "2024-06-01 12:30:00", and adds it to the HTTP request header of the feature sharing result.

[0125] Step S1573: Encapsulate the feature sharing results according to the data transmission protocol requirements, select a transmission protocol consistent with the business system interface protocol, and encapsulate the feature sharing results into data packets that conform to the protocol format.

[0126] Choose the HTTPS protocol to encapsulate the feature sharing results. In accordance with the format requirements of the HTTPS protocol, use the feature sharing results as the request body, add the corresponding request header fields, such as setting Content-Type to "application / json", and encapsulate it into a complete HTTPS data packet.

[0127] Step S1574: Set the transmission priority of data packets based on the timeliness control requirements. The transmission priority should be consistent with the feature transmission timeliness parameter in the shared adaptation parameters. The higher the feature transmission timeliness parameter requirement, the higher the transmission priority should be set.

[0128] The feature transmission timeliness parameter in the shared adaptation parameters requires transmission to be completed within 2 hours. Based on the interface's timeliness control requirements, the data packet transmission priority is set to medium. If the feature transmission timeliness parameter requires completion within 1 hour, it is set to high priority.

[0129] Step S1575: Send the encapsulated data packet to the interface of the talent service business system of the feature caller through the network channel, and monitor the transmission status of the data packet to ensure that the data packet arrives at the business system within the time required by the feature transmission timeliness parameter.

[0130] The platform system sends data packets through a stable network channel and monitors the transmission status in real time, such as sending progress and network latency. If network latency is detected that may prevent timely delivery, the system will automatically switch to a backup network channel to ensure that the data packets are successfully delivered to the enterprise customer's business system interface within 2 hours.

[0131] Step S1576: After the business system receives the data packet, it parses the feature sharing result in the data packet, verifies the validity of the access credential, and if the credential is valid, extracts the business process features and sharing adaptation parameters from the feature sharing result.

[0132] After receiving the data packet, the enterprise customer's business system first parses out the feature sharing result and the access credentials in the header. It then verifies whether the access credentials are valid within the current time and whether the identity code and access permission identifier are correct. If the verification passes, it extracts business process feature data and sharing adaptation parameters from the feature sharing result.

[0133] Step S1577: The business system determines the scope of use of business process features based on the feature application permission parameters in the shared adaptation parameters, extracts usable feature fields based on the feature open scope parameters, calls the business decision module to analyze and process the extracted feature fields, and generates the results of business decision operations. The results of the business decision operations include matching suggestions for personnel and positions, service process optimization schemes, and business resource allocation plans.

[0134] The business system determines the scope of application of features in business processes based on the feature application permission parameters "viewable, usable for model training". Based on the feature open scope parameters, usable fields such as technical ability fields and qualification certificate fields are extracted and input into the business decision module. The matching algorithm in the business decision module analyzes the technical capabilities of personnel and job requirements, generating matching suggestions such as "Job seeker A has a 90% match with job X, it is recommended to prioritize interviewing them"; the process optimization algorithm analyzes data such as interaction records and proposes optimization solutions for service processes, such as "changing the resume screening process from manual to algorithmic initial screening, which is expected to shorten the cycle by 2 days"; the resource allocation algorithm formulates a business resource allocation plan based on the urgency of job requirements and feature data, such as "allocate 3 recruitment specialists to job Y and 2 recruitment specialists to job Z".

[0135] To enable machine learning models to accurately calculate shared adaptation parameters, talent service platforms need to train these models. Model training is a continuous optimization process that uses extensive historical data to adjust model parameters, allowing the model to better adapt to real-world business scenarios.

[0136] Step S211: Collect historical feature sharing case data, which includes a set of historical business process features, a description of the business requirements of the historical feature calling subject, historical feature anchoring results, historical sharing adaptation parameters, and corresponding business decision effect feedback data.

[0137] We collected historical feature-sharing case data from the past three years. Each case includes a set of business process features at the time, a description of the enterprise customer's business needs, feature anchoring results, calculated shared adaptation parameters, and feedback data on the final business decision-making effect, such as matching accuracy and efficiency improvement in process optimization. The data volume reaches hundreds of thousands of records, ensuring sufficient samples for model training.

[0138] Step S212: Preprocess the historical feature-sharing case data, including data cleaning, missing value imputation, outlier handling, and data standardization.

[0139] Data cleaning removes duplicate case data and obviously erroneous data, such as cases with empty business requirement descriptions. Missing value imputation uses mean or median imputation to handle missing values ​​in feature data, and mode imputation is used for categorical data. Outlier handling identifies and removes outliers using box plots. Data standardization transforms all numerical feature data to the same dimension, such as using min-max standardization to map data to the [0, 1] interval.

[0140] Step S213: Divide the data into training set, validation set and test set, and randomly distribute the preprocessed historical data into the three sets according to the set ratio.

[0141] The hundreds of thousands of preprocessed data points were divided into three sets: 70% as the training set for learning model parameters; 20% as the validation set for tuning model hyperparameters; and 10% as the test set for evaluating the final performance of the model.

[0142] Step S214: Construct the network structure of the machine learning model and set initial hyperparameters, such as learning rate, number of iterations, number of hidden layer neurons, etc.

[0143] A network structure consisting of a feature input layer, an association processing layer, and a parameter output layer is constructed. The initial hyperparameter settings are: learning rate of 0.001, number of iterations of 1000, first perceptron neurons of the association processing layer with 200 neurons, second perceptron neurons with 100 neurons, and initial uniform weight distribution of the attention mechanism.

[0144] Step S215: Train the machine learning model using the training set, calculate the prediction shared adaptation parameters through forward propagation, adjust the model parameters through backpropagation, monitor the performance of the machine learning model using the validation set, and adjust the hyperparameters based on the performance of the validation set.

[0145] During training, the training set data is input into the machine learning model, and the predicted shared adaptation parameters are calculated through forward propagation. The predicted parameters are compared with the actual historical shared adaptation parameters, and the loss function value (e.g., mean squared error) is calculated. The model's weights and biases are adjusted based on the loss function value using the backpropagation algorithm. After a certain number of training epochs, the model performance is evaluated using a validation set. If the validation set loss no longer decreases, hyperparameters such as the learning rate or the number of iterations are adjusted.

[0146] Step S216: After training is completed, use the test set to evaluate the performance of the machine learning model. If the performance indicators of the machine learning model (such as accuracy and recall) reach the preset threshold, the model training is complete; otherwise, readjust the network structure or hyperparameters and train again.

[0147] The test set evaluates the model's performance metrics, such as accuracy and recall. The preset accuracy threshold is 85%, and the recall threshold is 80%. If the model achieves 88% accuracy and 82% recall on the test set, both exceeding the thresholds, the model training is complete. Otherwise, it may be necessary to increase the number of hidden layer neurons or adjust the learning rate, and retrain the model.

[0148] In the process of collecting and sharing characteristics of business processes, a large amount of sensitive personal information and other privacy data are involved. Talent service platforms need to take privacy protection measures to prevent data leakage.

[0149] For example, collected privacy-sensitive data can be classified and graded according to its sensitivity level, into extremely sensitive data, highly sensitive data, moderately sensitive data, and low-sensitivity data. Extremely sensitive data includes personal ID numbers, bank account numbers, etc.; highly sensitive data includes mobile phone numbers, home addresses, etc.; moderately sensitive data includes work experience, project experience, etc.; and low-sensitivity data includes career interests, industry preferences, etc.

[0150] Data anonymization techniques are used to process extremely sensitive and highly sensitive data. For example, some characters in ID card numbers are replaced (the first 6 digits and the last 4 digits are displayed, and the middle digits are replaced with *), and the middle 4 digits of mobile phone numbers are replaced with *.

[0151] Data encryption technology is employed to encrypt privacy-sensitive data during transmission and storage. SSL / TLS encryption is used during transmission, and AES encryption is used for data encryption during storage. During the transmission of characteristic data from the platform to the enterprise customer's business system, SSL / TLS encryption is enabled to ensure that the data is not intercepted during transmission. Sensitive data is encrypted using the AES-256 encryption algorithm during storage, and only authorized personnel with the decryption key can access the original data.

[0152] Set up an access control mechanism to strictly control access to privacy-sensitive data. Use role-based access control policies to assign different access permissions to different roles, and only authorized roles can access data of the corresponding sensitivity level.

[0153] The platform has roles such as administrator, regular operator, and auditor. Administrators have the highest privileges and can access data at all levels; regular operators can only access medium- and low-sensitivity data; auditors can only view data access logs and cannot access the raw data. Each access to sensitive data requires authentication and authorization checks.

[0154] Regularly audit and evaluate privacy protection measures, check the effectiveness of data anonymization, encryption strength, and access control, and promptly identify and fix privacy protection vulnerabilities.

[0155] Based on the same inventive concept, please refer to Figure 2 The diagram shows a schematic block diagram of a machine learning-based business data feature sharing system 100 for performing the above-described inspection video stream processing method, provided in an embodiment of this application. The machine learning-based business data feature sharing system 100 may include a communication unit 110, a machine-readable storage medium 120, and a processor 130.

[0156] In this embodiment, both the machine-readable storage medium 120 and the processor 130 are located within the machine learning-based business data feature sharing system 100 and are separately configured. However, it should be understood that the machine-readable storage medium 120 may also be independent of the machine learning-based business data feature sharing system 100 and may be accessed by the processor 130 via a bus interface. Alternatively, the machine-readable storage medium 120 may also be integrated into the processor 130 and may communicate and interact with external systems through the communication unit 110.

[0157] The processor 130 is the control center of the machine learning-based business data feature sharing system 100. It connects various parts of the system via various interfaces and lines, and performs overall monitoring of the system by running or executing software programs and / or modules stored in the machine-readable storage medium 120 and calling data stored in the machine-readable storage medium 120. Optionally, the processor 130 may include one or more processing cores; for example, it may integrate an application processor and a modem processor. The application processor mainly handles the operating system, user interface, and applications, while the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor. The machine-readable storage medium 120 stores machine-executable instructions for implementing the scheme of this application, and the processor 130 executes the machine-executable instructions stored in the machine-readable storage medium 120 to implement the inspection video stream processing method provided in the aforementioned method embodiments.

[0158] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A method for sharing business data features based on machine learning, characterized in that, The method includes: The system acquires a set of business process features in the talent service field and a description of the business needs of the feature caller. The set of business process features includes core personnel features, job-related features, and service process features. The core personnel features are generated based on personnel capability representation information, qualification certification information, and service preference information in the talent service field. The job-related features are generated based on job requirement definition information, responsibility boundary information, and adaptation standard information in the talent service field. The service process features are generated based on business interaction record information, service cycle record information, and feedback evaluation information in the talent service field. The description of the business needs of the feature caller includes the type range, timeliness requirements, and application direction of the business data features required by the feature caller. Construct feature anchoring rules between business process features and business requirement descriptions, and perform association processing on business process features and business requirement descriptions in the business process feature set based on the feature anchoring rules to obtain feature anchoring results; The feature anchoring results are input into a pre-trained machine learning model to calculate shared adaptation parameters. A shared execution link for business process features is constructed based on shared adaptation parameters. The shared execution link includes a feature extraction node, a feature processing node, and a feature delivery node. The feature extraction node corresponds to the storage location of the business process feature set. The feature processing node processes the business process features based on the shared adaptation parameters. The feature delivery node corresponds to the feature receiving interface of the feature calling subject. The business process characteristics output by the shared execution link are associated and mapped with the talent service business system of the feature calling entity, generating a feature sharing result that includes business process characteristic identifier, feature calling entity identifier, shared adaptation parameters and business system interface identifier. The feature sharing result is then fed back to the talent service business system of the feature calling entity to support business decision-making operations.

2. The business data feature sharing method based on machine learning according to claim 1, characterized in that, The feature anchoring rules for constructing business process features and business requirement descriptions are used to associate business process features and business requirement descriptions in the business process feature set based on the feature anchoring rules to obtain feature anchoring results, including: The feature dimension information of each business process feature is extracted from the set of business process features. The feature dimension information includes the ability dimension, qualification dimension, and preference dimension of the core personnel features; the demand dimension, responsibility dimension, and adaptation dimension of the job-related features; and the interaction dimension, cycle dimension, and evaluation dimension of the service process features. Extract requirement dimension information, requirement timeliness information, and requirement application information from the business requirement description of the feature caller. The requirement dimension information corresponds to the dimension type of the feature required in the business requirement description, the requirement timeliness information corresponds to the effective time range of the feature required in the business requirement description, and the requirement application information corresponds to the usage scenario direction of the feature in the business requirement description. Based on feature dimension information and demand dimension information, feature dimension matching conditions are formulated. When the feature dimension information of the business process features and the demand dimension information meet the overlap ratio requirement or the mapping relationship requirement, it is determined that the feature dimension matching conditions are met. The feature dimension matching conditions include the overlap ratio requirement of the same dimension type and the mapping relationship requirement of the related dimension type. Based on the generation time information of business process features and the timeliness information of demand, feature timeliness alignment conditions are formulated. The feature timeliness alignment conditions include that the generation time of the business process feature is within the effective time range corresponding to the timeliness information of demand, or that the update frequency of the business process feature is consistent with the timeliness requirement corresponding to the timeliness information of demand. When the business process feature meets any of the above conditions, it is determined that it meets the feature timeliness alignment conditions. Based on the application scope information and application demand information of business process characteristics, feature application adaptation conditions are formulated. The feature application adaptation conditions include whether the application scope information of business process characteristics covers the usage scenario direction corresponding to the application demand information, or whether the application method of business process characteristics is compatible with the usage method corresponding to the application demand information. When a business process characteristic meets any of the above conditions, it is determined that it meets the feature application adaptation conditions. The feature dimension matching conditions, feature timeliness alignment conditions, and feature application adaptation conditions are integrated into feature anchoring rules, and the judgment logic and combination methods of each condition are determined. Based on feature anchoring rules, each business process feature in the business process feature set is associated with the business requirement description one by one. The judgment results that meet each condition are recorded. The business process features that meet all conditions are bound to the business requirement description to generate feature anchoring results. The feature anchoring results include the bound business process feature identifier, the business requirement description identifier, and the judgment result details of each condition.

3. The business data feature sharing method based on machine learning according to claim 2, characterized in that, The step of formulating feature dimension matching conditions based on feature dimension information and demand dimension information, and determining whether the feature dimension information of the business process features meets the feature dimension matching conditions when the feature dimension information and demand dimension information meet the overlap ratio requirement or mapping relationship requirement, includes: The feature dimension information of business process characteristics and the demand dimension information of feature caller are classified by dimension type. The ability dimension, qualification dimension, and preference dimension of personnel core characteristics, the demand dimension, responsibility dimension, and adaptation dimension of job-related characteristics, and the interaction dimension, cycle dimension, and evaluation dimension of service process characteristics are respectively classified into first-level dimension types. Divide the primary dimension type into secondary dimension subtypes; Calculate the number of overlaps between the business process features and the second-level dimension subtypes of the demand dimension information under the same first-level dimension type, and take the ratio of the number of overlaps to the total number of second-level dimension subtypes of the demand dimension information as the overlap ratio of the same dimension type. Set an overlap ratio threshold. When the overlap ratio of the same dimension type is greater than the overlap ratio threshold, it is determined that the overlap ratio requirement of the same dimension type is met. Establish a mapping relationship table for related dimension types. The mapping relationship table includes the mapping relationship between the ability dimension of personnel core characteristics and the demand dimension of job-related characteristics, the mapping relationship between the preference dimension of personnel core characteristics and the interaction dimension of service process characteristics, and the mapping relationship between the adaptation dimension of job-related characteristics and the evaluation dimension of service process characteristics. Each mapping relationship includes the related fields and related weights between dimensions. Check whether there is a related dimension type in the mapping relationship table between the feature dimension information and the requirement dimension information of the business process. If there is a related dimension type, calculate the matching quantity and the weighted value of the related field respectively. The matching quantity and the related weight are judged according to the preset independent threshold. When the matching quantity meets the quantity threshold requirement and the related weight meets the weight threshold requirement, it is determined that the mapping relationship requirement of the related dimension type is met. When the feature dimension information of a business process and the requirement dimension information meet the requirements of the same dimension type overlap ratio or the mapping relationship of related dimension types, it is determined that the feature of the business process and the description of the business requirement meet the feature dimension matching conditions.

4. The business data feature sharing method based on machine learning according to claim 3, characterized in that, The establishment of the mapping relationship table for related dimension types includes: Collect historical business data in the talent service field, including past records of matching personnel with positions, interaction records during the service process, and service evaluation records; The correspondence between the capability dimension fields of personnel core characteristics and the demand dimension fields of job-related characteristics is extracted from historical business data. The frequency of the same capability dimension field and demand dimension field appearing together is counted, and the frequency is used as the association weight to establish the mapping relationship between the capability dimension of personnel core characteristics and the demand dimension of job-related characteristics. The correspondence between the preference dimension fields of core personnel characteristics and the interaction dimension fields of service process characteristics is extracted from historical business data. The influence of the preference dimension fields on the interaction dimension fields is analyzed, and the influence is quantified into association weights to establish a mapping relationship between the preference dimension of core personnel characteristics and the interaction dimension of service process characteristics. The correspondence between the adaptation dimension field of job-related features and the evaluation dimension field of service process features is extracted from historical business data. The correlation coefficient between the adaptation dimension field and the evaluation dimension field is calculated. The correlation coefficient is used as the association weight to establish the mapping relationship between the adaptation dimension of job-related features and the evaluation dimension of service process features. The established mapping relationships are integrated, and each mapping relationship entry contains the source dimension type, source dimension field, target dimension type, target dimension field and associated weight, forming a mapping relationship table of associated dimension types; The mapping relationship table is updated based on the latest talent service business data. The corresponding relationships and association weight changes of each dimension field in the new business data are regularly statistically analyzed, and the association weight values ​​in the mapping relationship table are adjusted to ensure that the mapping relationship table is adapted to the current talent service business scenario.

5. The business data feature sharing method based on machine learning according to claim 1, characterized in that, The machine learning model includes a feature input layer, an association processing layer, and a parameter output layer. The feature input layer receives the feature anchoring results and converts them into feature vectors that the model can process. The association processing layer mines the adaptation relationship between the feature anchoring results and the talent service business scenario. The parameter output layer outputs the shared adaptation parameters of the business process features to the feature calling subject. The step of inputting the feature anchoring results into a pre-trained machine learning model to calculate shared adaptation parameters, thereby obtaining shared adaptation parameters, includes: The business process feature identifiers, business requirement description identifiers, and judgment result details of each condition in the feature anchoring results are encoded. The business process feature identifiers are converted into feature codes, the business requirement description identifiers are converted into requirement codes, and the judgment result details of each condition are converted into judgment codes. The feature encoding, demand encoding, and decision encoding are input into the feature input layer of the machine learning model. Dimensional unification and feature concatenation operations are then performed to obtain a comprehensive feature vector. The comprehensive feature vector is input into the association processing layer of the machine learning model. The association processing layer contains a multi-layer perceptron structure. The first layer perceptron performs linear transformation and activation function processing on the comprehensive feature vector to extract the basic association features in the comprehensive feature vector and outputs the basic association feature vector. The basic association feature vector is input into the second-layer perceptron. The second-layer perceptron performs nonlinear transformation and feature interaction processing on the basic association feature vector to explore the adaptation relationship between the basic association feature vector and the talent service business scenario. The adaptation relationship includes feature usage frequency association, feature update cycle association, and feature application effect association. A feature attention mechanism is set in the association processing layer. The feature attention mechanism is used to assign weights to the mined adaptation associations, so that the adaptation associations with higher correlation with shared adaptation parameters are given higher weights, and an adaptation association vector with attention weights is generated. The adaptation relationship vector with attention weights is input into the parameter output layer of the machine learning model. Through regression calculation, shared adaptation parameters for business process features to feature calling subjects are generated. The shared adaptation parameters include feature open range parameters, feature transmission timeliness parameters, and feature application permission parameters. The feature open range parameters define the range of business process feature fields that can be shared. The feature transmission timeliness parameters define the transmission time requirements of business process features. The feature application permission parameters define the permission level of feature calling subjects to use business process features.

6. The business data feature sharing method based on machine learning according to claim 5, characterized in that, The step of setting a feature attention mechanism in the association processing layer, and assigning weights to the mined adaptation associations through the feature attention mechanism, so that adaptation associations with higher correlation to shared adaptation parameters receive higher weights, and generating an adaptation association vector with attention weights, includes: Extract all the adaptation relationships mined by the association processing layer. The adaptation relationships include feature usage frequency association, feature update cycle association, feature application effect association, feature sensitivity association, and feature call cost association. Construct an adaptation association weight evaluation function, which is generated based on the degree of influence of the adaptation association relationship on the shared adaptation parameters. The degree of influence includes the influence coefficient of the adaptation association relationship on the feature open range parameter, the influence coefficient on the feature transmission timeliness parameter, and the influence coefficient on the feature application permission parameter. Each adaptation association is input into the adaptation association weight evaluation function to calculate the weight value corresponding to each adaptation association. The magnitude of the weight value is positively correlated with the total influence coefficient of the adaptation association on the shared adaptation parameters. The total influence coefficient is the weighted sum of the three influence coefficients. The weight values ​​of all adaptation relationships are normalized to obtain the attention weight of each adaptation relationship. Multiply the feature vector corresponding to each adaptation relationship with the attention weight of that adaptation relationship to obtain the weighted feature vector of each adaptation relationship; All weighted feature vectors of the adaptation relationships are concatenated according to the original order of the adaptation relationships to generate an adaptation relationship vector with attention weights. By comparing the attention-weighted adaptation relationship vector with the original adaptation relationship vector, the feature dimensions of the attention-weighted adaptation relationship vector with values ​​greater than those of the original adaptation relationship vector are retained, thereby enhancing the ability to represent adaptation relationships that have a significant impact on shared adaptation parameters.

7. The business data feature sharing method based on machine learning according to claim 1, characterized in that, The process of constructing a shared execution link based on shared adaptation parameters to identify business process characteristics includes: Based on the feature open range parameter in the shared adaptation parameters, the feature fields of the business links to be shared are determined. The storage location of the feature fields of the business links to be shared in the feature set of the business links is extracted. This storage location is determined as the feature extraction node of the shared execution link. The feature extraction node contains the storage address, storage format and reading method of the feature fields of the business links to be shared. Feature processing rules are constructed based on the feature open range parameter in the shared adaptation parameters. The feature processing rules include filtering rules, format conversion rules, and content integration rules for feature fields of business links to be shared. The filtering rules filter fields to be shared based on the field range defined by the feature open range parameter. The format conversion rules convert the fields to be shared into a format that can be accepted by the feature calling subject. The content integration rules integrate the scattered fields to be shared into structured data. The processing module built based on feature processing rules is identified as the feature processing node of the shared execution link. The feature processing node receives the features of the business link to be shared from the feature extraction node, processes them according to the feature processing rules, and outputs the processed business link features. Based on the feature transmission timeliness parameter in the shared adaptation parameters and the feature receiving interface information of the feature calling subject, the feature delivery node of the shared execution link is determined. The feature delivery node includes the address of the feature receiving interface, the interface protocol, the data receiving format and timeliness requirements. The feature transmission timeliness parameter defines the transmission time limit from the feature processing node to the feature delivery node. Establish the connection relationship between feature extraction node, feature processing node and feature delivery node. Read the business process features to be shared from the feature extraction node, transmit them to the feature processing node for processing, and then transmit the processed business process features to the feature delivery node to form a complete shared execution link. The link information of the shared execution link is associated with the characteristics of the business process to be shared and the sharing adaptation parameters. The link information includes the feature extraction node identifier, feature processing node identifier, feature delivery node identifier and link transmission requirements, and a shared execution link association table is generated.

8. The business data feature sharing method based on machine learning according to claim 7, characterized in that, The step of constructing feature processing rules based on the feature open range parameter in the shared adaptation parameters includes: Parse the feature open range parameter in the shared adaptation parameters, and extract the field list of the business process features to be shared as defined by the feature open range parameter. The field list includes specified fields of core personnel features, specified fields of job-related features, and specified fields of service process features. Filtering rules are constructed based on a list of fields. The filtering rules include field name matching logic and field attribute matching logic. The field name matching logic is used to filter business process feature fields whose names are consistent with those in the field list. The field attribute matching logic is used to filter business process feature fields whose attributes are consistent with those in the field list. The business process feature fields to be shared are filtered from the set of business process features through the field name matching logic and the field attribute matching logic. The feature receiving format requirements of the feature calling subject are obtained. The format requirements include data structure format, field type definition and data encoding method. Based on the format requirements, a format conversion rule is constructed. The format conversion rule includes field type conversion logic, data structure reorganization logic and data encoding conversion logic. The field type conversion logic converts the type of the field to be shared into the type required by the feature calling subject. The data structure reorganization logic reorganizes the structure of the field to be shared into the structure required by the feature calling subject. The data encoding conversion logic converts the encoding of the field to be shared into the encoding required by the feature calling subject. Analyze the distribution of characteristic fields of the business process to be shared, determine the scattered fields to be shared and their relationships, and construct content integration rules based on the relationships. The content integration rules include field association logic, data merging logic and redundancy removal logic. The field association logic establishes the relationship between scattered fields based on the related fields between fields. The data merging logic merges the associated scattered fields into a unified data unit. The redundancy removal logic removes duplicate content in the merged data unit. The filtering rules, format conversion rules, and content integration rules are integrated into a complete feature processing rule. The execution order and conditions of each rule are determined. The filtering rules are executed first, the format conversion rules are executed after the filtering rules, and the content integration rules are executed after the format conversion rules.

9. The business data feature sharing method based on machine learning according to claim 1, characterized in that, The process of associating and mapping the business process features output from the shared execution link with the talent service business system of the feature calling entity to generate a feature sharing result containing business process feature identifiers, feature calling entity identifiers, shared adaptation parameters, and business system interface identifiers, and then feeding back the feature sharing result to the talent service business system of the feature calling entity to support business decision-making operations, includes: Extract the identification information of the business link features output by the shared execution link. The identification information includes the unique code of the business link feature, the generation time, the feature category, and the version information. Obtain the identification information of the feature caller, which includes the name, type, registration number in the talent service field, and business scope description of the feature caller; The interface identifier of the talent service business system that determines the calling subject is determined. The interface identifier includes the interface name, interface address, interface version and data types supported by the interface of the business system. The identification information of business process features, the identification information of feature calling entities, shared adaptation parameters, and business system interface identifiers are mapped to fields to establish the association between each piece of information. The identification information of business process features and shared adaptation parameters are associated through unique feature codes, and the identification information of feature calling entities and business system interface identifiers are associated through the name of the calling entity. A data structure for feature sharing results is constructed based on the association relationship. The data structure includes a business process feature identifier field, a feature call subject identifier field, a shared adaptation parameter field, and a business system interface identifier field. Each field contains corresponding information content and data type definition. Fill in the contents of each field in the data structure. The business process feature identifier field is filled with the unique code of the business process feature and the information of the generation time. The feature calling entity identifier field is filled with the name and registration number of the feature calling entity. The shared adaptation parameter field is filled with the information of the feature open scope parameter and the feature transmission timeliness parameter. The business system interface identifier field is filled with the interface address and interface version of the business system. The interface of the talent service business system of the feature call subject is used to feed back the fully filled feature sharing results to the business system. After receiving the feature sharing results, the business system calls the features of the business process based on the shared adaptation parameters for business decision-making operations in the talent service field. The business decision-making operations include personnel and position matching decisions, service process optimization decisions, and business resource allocation decisions.

10. A business data feature sharing system based on machine learning, characterized in that, include: processor; A machine-readable storage medium for storing machine-executable instructions of the processor; The processor is configured to execute the machine learning-based business data feature sharing method of any one of claims 1 to 9 by executing the machine executable instructions.

Images