A data-driven human target matching dynamic weight adaptive adjustment system and method

Abstract

This invention discloses a data-driven dynamic weight adaptive adjustment system and method for human-target matching, belonging to the field of artificial intelligence data-driven decision-making and intelligent human-target matching technology. It is the underlying core parameter support patent (sub-patent 09) of this patent pool. Based on historical matching data, user feedback data, and scenario feature data, this invention analyzes feature contribution through statistical analysis, machine learning, and reinforcement learning, automatically adjusting the matching weights of all sub-features in the four layers of talent and the six dimensions of the target. It supports multiple trigger conditions, scenario-based weight templates, user-defined weights, interpretability of weight adjustment, stability control, and multi-objective optimization; covering nine major scenarios, it provides dynamic, accurate, and adaptive weight parameter support for the entire patent pool. This invention shortens the weight adjustment cycle from 1-3 months to 1-7 days, and improves the matching accuracy by an average of 15%-20%. It completely solves the pain points of traditional manual weight presets, rigidity, and poor scenario adaptability. It forms a complete technical closed loop with the parent patent and various sub-patents, possessing extremely high novelty, inventiveness, and practicality.

Description

Technical Field

[0001] This invention belongs to the fields of artificial intelligence data-driven decision-making, human-item intelligent matching weight optimization, feature contribution quantification analysis, and multi-objective optimization technology. It is the underlying core parameter adaptive support invention patent of this patent pool and is positioned as sub-patent 09. This invention, together with the parent patent "A Human-Item Intelligent Adaptation Quantitative Judgment System and Method" and sub-patents 01 to 08, forms a complete seven-layer technical architecture and a full-chain patent closed loop: underlying parameter adaptive → feature extraction → matching calculation → imbalance optimization → cross-scenario migration → full-cycle risk control → long-term time-series prediction.

[0002] This invention, based on historical talent matching data, user feedback data, and scenario feature data, uses three feature contribution analysis methods—statistical analysis, machine learning, and reinforcement learning—to automatically adjust the matching weights between the four layers of talent-side features (surface traits, abilities and skills, personality traits, and growth potential) and the six dimensions of target-side features (market, technology, business, finance, compliance, and value). It supports multiple trigger conditions, scenario-based weight templates, user-defined weights, interpretability of weight adjustments, stability control, and multi-objective optimization. Covering nine major scenarios across education, further education, career development, recruitment, headhunting, venture capital, technology transfer, industry-academia-research collaboration, and team building, this invention completely solves the fundamental pain points of traditional talent matching—"manually preset weights, fixed and rigid structures, poor scenario adaptability, high parameter tuning costs, lack of interpretability, and lack of stability control"—providing dynamic, accurate, and adaptive weight parameter support for all upper-level technologies in the entire patent pool, significantly improving matching accuracy and user satisfaction across all scenarios. Background Technology

[0003] In the human-identity intelligent matching technology system, feature weight allocation is the core parameter determining matching accuracy. Traditional human-identity matching systems generally adopt a manually preset fixed weight mode, which has six fundamental technical defects, severely restricting the matching effect and scene adaptability: 1. Manually preset weights are highly subjective and have low accuracy. The weights are entirely dependent on the experience of engineers or industry experts. The judgment criteria of different people vary greatly, which cannot objectively reflect the actual contribution of each feature to the matching results. As a result, the matching accuracy has been hovering at 60%-70% for a long time and it is difficult to improve it further. 2. Fixed and rigid weights cannot adapt to dynamic changes. Fixed weights cannot be automatically adjusted as business develops, market changes, and user needs evolve. For example, the technical requirements of the internet industry change extremely rapidly, and weight configurations from three years ago are no longer suitable for current recruitment needs. However, traditional systems cannot update automatically, resulting in a continuous decline in matching effectiveness. 3. High cost and long cycle for parameter tuning Each weight adjustment requires manual data analysis and repeated testing and verification, with a parameter tuning cycle of 1-3 months and labor costs exceeding tens of thousands of yuan, making it impossible to quickly respond to changes in scenarios and user needs. 4. Lacks context-specific weight adaptation capability The weighting requirements vary greatly across different scenarios, industries, and users. For example, venture capital scenarios place greater emphasis on the entrepreneur's mindset and growth potential, while technology transfer scenarios prioritize technological maturity and commercial viability. Traditional systems employ uniform weighting, which fails to achieve scenario-based adaptation, leading to a significant decrease in cross-scenario matching accuracy. 5. Lacks support for user-defined weights, failing to meet personalized needs. Different enterprises and organizations have different business priorities and therefore different levels of emphasis on features. Traditional systems do not support user-defined weights or only support simple global weight adjustments, which cannot meet users' personalized and customized needs. 6. Weight adjustment lacks interpretability and stability control. The traditional system's weight adjustment process is opaque, making it impossible to explain the reasons and basis for the adjustment, which makes it difficult for users to trust it. At the same time, it lacks a stability control mechanism, and a single adjustment may cause drastic fluctuations in weight, leading to chaotic matching results, and it is impossible to roll back to the previous version. 7. Lacks multi-objective optimization capability Traditional weighting methods only pursue maximizing matching accuracy, ignoring multiple objectives such as matching success rate, user satisfaction, and result diversity, resulting in monotonous matching results and poor user experience.

[0010] According to industry statistics, matching systems using manual fixed weights require comprehensive parameter tuning every six months on average. During this tuning period, the matching accuracy drops by 15%-20%, and business losses due to weight issues exceed 30% annually. This invention addresses all of these shortcomings by constructing a data-driven, end-to-end dynamic weight adaptive adjustment system. This system enables automatic, accurate, stable, and interpretable weight adjustment, providing underlying parameter support for the entire patent pool and completely filling a technological gap in the industry. Summary of the Invention

[0011] Purpose of the invention Addressing the core pain points of traditional human-identity matching weight systems—namely, manual pre-setting, rigidity, poor scenario adaptability, high parameter tuning costs, lack of interpretability, and absence of stability control—this invention provides a data-driven dynamic weight adaptive adjustment system and method for human-identity matching, achieving the following core objectives: 1. Construct a data-driven automatic weight adjustment system to replace manual presets, automatically analyze feature contribution based on historical data and user feedback, and dynamically adjust matching weights; 2. Achieve full feature coverage, support independent adjustment of all sub-features in the four layers on the talent side and all sub-features in the six dimensions on the target side, with a focus on supporting independent weight control of value concepts and external image features; 3. Supports multiple triggering conditions, including periodic adjustments, scene switching, data accumulation, effect decline, and manual triggering, to achieve real-time adaptive weighting; 4. Provides three feature contribution analysis methods (statistical analysis, machine learning, and reinforcement learning) to adapt to different data volumes and business scenarios; 5. Establish a scenario-based weighted template management mechanism to preset initial templates for different scenarios and automatically optimize them, while supporting users to create custom templates; 6. Supports user-defined weights and priorities. The system only makes minor adjustments based on the user's customization to ensure that it does not deviate from the user's core needs. 7. Ensure the interpretability of weight adjustments, generate weight adjustment reports, and explain the adjustment range, reasons, basis, and effect comparison; 8. Provide weight stability control, setting maximum adjustment range, smooth sliding window, and historical version rollback to prevent drastic weight fluctuations; 9. Employ a multi-objective optimization strategy to simultaneously optimize multiple objectives, including matching accuracy, success rate, user satisfaction, result diversity, and adjustment range; 10. Supports nine major application scenarios across the entire domain, providing unified dynamic weight support for all upper-layer technologies in the entire patent pool, forming a complete technology closed loop.

[0012] Technical solution

[0013] (I) Core Methods and Steps The data-driven human-identity matching dynamic weight adaptive adjustment method disclosed in this invention includes the following standardized execution steps: S1: Obtain historical talent matching data, user feedback data, and scenario feature data; historical matching data includes talent characteristics, target characteristics, matching results, and suitability scores; user feedback data includes recruitment results, investment decisions, user evaluations, and complaints and suggestions; scenario feature data includes scenario type, industry attributes, user preferences, and business rules.

[0014] S2: Based on the above data, analyze the actual contribution of each feature to the matching results; using at least one of the following methods: - Statistical analysis methods: Calculate the Pearson correlation coefficient and Spearman correlation coefficient between each feature and the matching results to quantify the contribution of the features; - Machine learning methods: Calculate the marginal contribution of each feature to the matching result using algorithms such as SHAP value, LIME, random forest feature importance, and XGBoost feature importance; - Reinforcement learning method: Treat weight adjustment as a Markov decision process, learn the optimal weight allocation strategy through interaction with the environment, and maximize long-term rewards.

[0015] S3: Based on the feature contribution analysis results, automatically adjust the matching weights of each level of talent features and each dimension of target features; the weight adjustment covers all sub-features of the four levels of talent features, abilities and skills, personality traits, and growth potential, as well as all sub-features of the six dimensions of target features, including market, technology, business, finance, compliance, and value; among them, the weights of value features and external image features can be adjusted independently.

[0016] S4: Verify the matching effect after weight adjustment, and evaluate the changes in indicators such as matching accuracy, matching success rate, user satisfaction, and result diversity; use A / B testing to compare and verify the adjusted weights with the original weights.

[0017] S5: If the verification passes (core indicators improve by ≥5% or do not decrease significantly), update the global weight library and the weight template for the corresponding scenario; if the verification fails, roll back the weight adjustment and re-analyze the feature contribution and optimize the weight.

[0018] (ii) Methods for adding core functions 1. Scenario-based weight template management: Preset initial weight templates for 9 major scenarios, including education and training, school selection, career planning, corporate recruitment, headhunting services, venture capital, technology transfer, industry-university-research cooperation, and team building; automatically optimize exclusive weight templates for each scenario based on historical data; support users to create custom weight templates based on basic templates and set the applicable scope and permissions of the templates. 2. User-defined weight support: Receives user-inputted custom base weights; performs data-driven fine-tuning based on user-defined base weights, with the fine-tuning range not exceeding a preset threshold (default 20%); retains the highest priority of user-defined weights, ensuring that system adjustments do not overwrite or deviate from the user's core needs. 3. Generate interpretable weight adjustment reports: Generate weight adjustment reports explaining the magnitude, reasons, and data basis for the weight adjustments of each feature; display a comparative analysis of the matching effect before and after weight adjustment, including changes in accuracy, success rate, and user satisfaction; and provide suggestions and explanations for weight adjustments to help users understand and make decisions. 4. Weight stability control: Set a maximum limit on the magnitude of weight adjustment (the adjustment magnitude of a single feature shall not exceed 30%) to prevent drastic fluctuations in weight; use a sliding window averaging method to smooth the weight adjustment process, and calculate the final weight based on the contribution analysis results of the most recent 3 periods; retain all historical weight versions, support rollback to any historical version at any time, and record version change logs. 5. Multi-objective optimization strategy: Employ multi-objective genetic algorithms, particle swarm optimization algorithms, etc., to simultaneously optimize at least two of the following objectives: maximizing matching accuracy, maximizing matching success rate, maximizing user satisfaction, maximizing the diversity of matching results, and minimizing the weight adjustment range; generate a Pareto optimal solution set for users to select the optimal weight scheme.

[0023] (III) System Composition The data-driven human-identity matching dynamic weight adaptive adjustment system disclosed in this invention comprises six core modules: 1. Data Acquisition Module: Used to acquire historical user-identity matching data, user feedback data, and scenario feature data, and to connect with business data and user data from other modules in the patent pool. 2. Contribution Analysis Module: This module integrates statistical analysis, machine learning, and reinforcement learning methods to analyze the actual contribution of each feature to the matching results. 3. Dynamic weight adjustment module: It is used to automatically adjust the matching weight of each feature based on the feature contribution analysis results, and supports independent adjustment of all features and multi-objective optimization. 4. Effect Verification Module: Used to verify the matching effect after weight adjustment through A / B testing and to evaluate the changes in various core indicators. 5. Weight Library Management Module: Used to store, update and manage the global weight library, scenario-based weight templates and historical weight versions, and supports template creation, editing, deletion and rollback. 6. User Interaction Module: This module supports users in defining custom weights, viewing weight adjustment reports, manually triggering weight adjustments, and managing custom weight templates.

[0029] (iv) Computer-readable storage media The present invention also protects a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method of any one of claims 1 to 10.

[0030] Core Innovation Points 1. Full-process data-driven automatic weight adjustment: Completely replaces manual preset weights, automatically analyzes and adjusts based on historical data and user feedback, shortens the weight adjustment cycle from 1-3 months to 1-7 days, and reduces parameter adjustment costs by more than 90%. 2. Independent and fine-grained adjustment of all features: Covers all sub-features in 4 layers on the talent side and 6 dimensions on the target side, supports independent control of value concepts and external image features, and refines the adjustment granularity from "level / dimension" to "atomic feature". 3. Multi-trigger condition adaptive: Supports five weight adjustment methods: periodic, scene switching, data accumulation, effect decline, and manual triggering, to achieve real-time dynamic adaptation of weights. 4. Multi-method contribution analysis: Integrating statistical, machine learning, and reinforcement learning methods to adapt to different data volumes and business scenarios, the accuracy of contribution analysis is improved to over 90%. 5. Dual-mode: Scenario-based templates + customization: Preset initial templates for 9 scenarios and automatically optimize them, while also supporting user-defined weights and templates, taking into account both standardization and personalization needs. 6. Dual guarantee of interpretability and stability: Generates transparent weight adjustment reports, and sets maximum range, smooth sliding, and version rollback mechanisms to ensure that weight adjustments are credible, stable, and controllable. 7. Multi-objective global optimization: Simultaneously optimize five objectives: accuracy, success rate, satisfaction, diversity, and adjustment range, to achieve global optimization rather than optimization of a single indicator. 8. Universal platform for all scenarios: A weight adjustment system adapts to 9 core scenarios, providing unified dynamic parameter support for all upper-layer technologies in the entire patent pool, maximizing technology reusability. Attached Figure Description

[0039] Figure 1 This is a flowchart illustrating the data-driven human-label matching dynamic weight adaptive adjustment method of the present invention. It sequentially shows the complete closed-loop process of data acquisition, feature contribution analysis, dynamic weight adjustment, effect verification, weight library update and rollback. It highlights the core technical aspects such as multi-contribution analysis method, multiple triggering conditions, scenario-based templates, user customization, interpretable generation, and stability control, and maintains complete consistency with the style of all the accompanying drawings in this patent pool.

[0040] Figure 2This is a module architecture diagram of the data-driven human-label matching dynamic weight adaptive adjustment system of the present invention. It shows the collaborative working relationship and data flow of the data acquisition module, contribution analysis module, dynamic weight adjustment module, effect verification module, weight library management module, and user interaction module, and clearly presents the core architecture logic of the system. Detailed Implementation

[0042] Example 1: Dynamic Weight Adaptive Adjustment in Enterprise Recruitment Scenarios An internet company uses this system to adjust the weighting of technical job postings. The initial preset weighting is: skills and abilities 40%, surface traits 20%, personality traits 25%, and growth potential 15%.

[0043] - Data Acquisition: The system acquires matching data from 1,000 resumes over the past 6 months, HR hiring feedback data, user satisfaction data, and scenario characteristic data from the Internet technology industry. - Contribution Analysis: The XGBoost feature importance algorithm was used for analysis. The results showed that growth potential contributed the most to the recruitment results (32%), followed by skills (28%), personality traits (25%), and surface traits (15%). Among them, the contribution of the "learning ability" sub-feature far exceeded that of other ability sub-features. - Weighting adjustment: The system automatically adjusts the weights to: growth potential 32%, ability and skills 28%, mindset traits 25%, and surface traits 15%; at the same time, the weight of the "learning ability" sub-trade is increased to 12%. - Results Verification: A two-week A / B test was conducted. After the adjustment, the matching accuracy of the weighted group increased from 72% to 83%, and HR satisfaction increased from 65% to 82%, with significant improvements in core indicators. - Weight Update: After verification, the system updates the weight template for the Internet technology recruitment scenario and synchronizes it to the global weight library. - Explainability Report: Generates a weight adjustment report that explains the adjustment range of each feature weight, the data basis, and the effect comparison, allowing HR to clearly see the reasons for the adjustment.

[0049] Example 2: Optimization of the Weight Template for Venture Capital Scenarios An angel investment institution uses the venture capital scenario weight template of this system. The initial template weights are: personality traits 35%, growth potential 30%, ability and skills 20%, surface traits 15%; target side: technology 30%, business 25%, market 20%, finance 15%, compliance 5%, value 5%.

[0050] - Data accumulation: The system has accumulated data on 200 investment projects of the institution over the past year, including project characteristics, founder characteristics, investment decisions, and subsequent performance data. - Automatic optimization: The system automatically analyzes the contribution of features every quarter and found that the "value concept" sub-feature contributes as much as 28% to the project success rate, far higher than the 10% of the initial template; the contribution of the "financial" dimension has increased to 20%. - Template Update: The system automatically increases the weight of the "value concept" sub-feature in the mindset traits to 28%, and the weight of the financial dimension on the target side to 20%, and generates a new weight template exclusive to the venture capital scenario. - Results Verification: After using the new template, the agency's project investment success rate increased from 22% to 35%, and post-investment risk decreased by 28%.

[0054] Example 3: User-defined weights and system fine-tuning A military industrial enterprise needs to create a custom weighted template for recruiting military technical personnel, requiring that the weight of the "confidentiality qualification" sub-feature be no less than 30% and the weight of the "political affiliation" sub-feature be no less than 20%.

[0055] - User-defined: Users can input their own base weights through the interactive module: Confidentiality qualification 30%, political affiliation 20%, technical ability 25%, work experience 15%, and others 10%. - System fine-tuning: Based on historical data from the military industry, the system is fine-tuned on the basis of user-defined settings. The weight of "technical capability" is fine-tuned to 27%, and "work experience" is fine-tuned to 13%. The fine-tuning range does not exceed 20%, and the core weight priority defined by the user is retained. - Template saving: After user confirmation, the system saves the custom weight template and sets it as the default template for the enterprise. - Results Verification: After using the custom template, the matching accuracy of military technical personnel increased from 68% to 85%, meeting the personalized needs of enterprises.

[0059] Example 4: Weight Stability Control and Version Rollback A headhunting platform conducted a quarterly weight adjustment. The initial system calculation showed that the weight of the "external image" feature was increased from 10% to 45%, far exceeding the preset maximum adjustment range of 30%.

[0060] - Stability control: The system automatically limits the adjustment range to within 30%, and finally adjusts the weight of "external image" to 13%; at the same time, it uses a sliding window averaging method, combined with the contribution data of the previous two quarters, to smooth the adjustment process and avoid drastic fluctuations in weight. - Version history: The system records the version number, adjustment time, adjustment content, and effect data of this weight adjustment. - Version rollback: After the adjustment was launched, it was found that the matching results for some positions were abnormal. Users can roll back to the previous stable version with one click through the interaction module. The system will immediately restore the original weight configuration and record the reason for the rollback for subsequent model optimization.

[0063] Example 5: Multi-objective optimization weight configuration An education platform needs to simultaneously optimize the accuracy of college matching, student satisfaction, and outcome diversity.

[0064] - Multi-objective settings: Users set optimization objectives as follows: maximize matching accuracy, maximize student satisfaction, maximize result diversity, and minimize weight adjustment range. - Multi-objective optimization: The system uses a multi-objective genetic algorithm to generate a Pareto optimal solution set, providing three weighting schemes: Option A: Accuracy 88%, Satisfaction 82%, Diversity 75% 2. Option B: Accuracy 85%, Satisfaction 88%, Diversity 78% 3. Option C: Accuracy 86%, Satisfaction 85%, Diversity 82% - Solution Selection: Users select Solution C based on their business needs, and the system applies this weight configuration and updates the weight library. - Results Verification: After application, student satisfaction on the platform increased by 12%, result diversity increased by 18%, and matching accuracy remained high.

[0067] Example 6: Automatic weight adjustment triggered during scene switching When a user switches from a corporate recruitment scenario to a team building scenario, the system automatically triggers a weight adjustment: - Scene Recognition: The system recognizes the current scene as team building and automatically loads the weighted template for the team building scene. - Template Application: The team building scenario template has the following weights: personality traits 35%, complementary abilities 30%, alignment of values ​​25%, and surface traits 10%, with a focus on increasing the weights of values ​​and complementary abilities. - Immediate effect: The weight template takes effect immediately, and subsequent team formation matching will use this weight configuration. - Automatic optimization: The system will continuously collect matching data in this scenario and automatically optimize the template weight every quarter.

[0071] Beneficial effects

[0072] I. Beneficial effects from a technical perspective 1. Revolutionary upgrade of the weight adjustment system: It has been upgraded from manually preset fixed weights to data-driven dynamic adaptive weights, which completely solves the problems of subjectivity and rigidity in the traditional weight system. 2. Significantly improved adjustment accuracy and efficiency: The accuracy of feature contribution analysis has been improved to over 90%, the weight adjustment cycle has been shortened from 1-3 months to 1-7 days, and the parameter adjustment cost has been reduced by over 90%. 3. Fine-tuning of all features: The granularity of adjustment is refined to the atomic feature level, supporting independent control of value concepts and external image, and improving matching accuracy by an average of 15%-20%. 4. Dual guarantee of stability and interpretability: Multiple stability control mechanisms prevent weight fluctuations, transparent adjustment reports enhance user trust, and the acceptability of weight adjustments is increased by more than 80%. 5. Multi-objective global optimization: Simultaneously optimize multiple business objectives to achieve global optimization rather than optimization of a single metric, resulting in a significant improvement in user experience. 6. Maximize technology reusability: One system adapts to 9 major scenarios, providing unified weight support for the entire patent pool and significantly reducing R&D and maintenance costs. 7. Continuous evolution capability: Based on the continuous accumulation of data, the weights are automatically optimized, and the system capabilities continue to improve with business development, forming a positive cycle.

[0079] II. Beneficial effects on the business level 1. Improved matching performance across the board: The average accuracy of matching across all scenarios has increased by 15%-20%, the success rate of matching has increased by more than 25%, and user satisfaction has increased by 30%. 2. Significantly reduced operating costs: It replaces a large amount of manual parameter tuning work, reducing labor costs by more than 80% and greatly shortening the parameter tuning cycle. 3. Enhanced customer loyalty: Scenario-based templates and custom weighting features meet the personalized needs of different customers, resulting in a 25% increase in customer retention rate. 4. Product competitiveness differentiation: The dynamic weight adaptive capability has become the core differentiating advantage of the product, far exceeding similar products in the industry, and significantly increasing market share. 5. Accelerated commercialization: Quickly adapt to new scenarios and industries, shortening the implementation cycle of new scenarios from 3 months to 1 week, accelerating market expansion.

[0084] III. Beneficial Effects at the Industry Level This invention establishes a unified dynamic weight adjustment standard in the field of human-label matching, promoting the industry's transformation from "human experience-driven" to "data intelligence-driven", improving the intelligence level and operational efficiency of the entire human resources, venture capital, education, and technology transfer industries, reducing resource waste caused by unreasonable weights, and promoting the high-quality development of related industries.

[0085] IV. Beneficial Effects at the Patent Layout Level This invention serves as the underlying core parameter support patent for the patent pool, filling the gap in the field of adaptive weight adjustment. Together with the upper-layer feature extraction, matching calculation, imbalance optimization, cross-scenario migration, full-cycle risk control, and long-term time-series prediction patents, it forms a complete seven-layer technical architecture and a full-chain patent barrier. It achieves full-process technical protection for human-label matching from data input to result output, significantly enhancing the overall value, defense capabilities, and commercial monetization potential of the patent pool.

Claims

1. A data-driven method for adaptive adjustment of dynamic weights in person-identity matching, characterized in that, Includes the following steps: S1: Obtain historical person-identity matching data, user feedback data, and scene feature data; S2: Based on the data, analyze the actual contribution of each feature to the matching results; S3: Based on the feature contribution analysis results, automatically adjust the matching weights of each level of talent features and each dimension of target features; S4: Verify the matching effect after weight adjustment, and evaluate the changes in matching accuracy, success rate and user satisfaction; S5: If the verification passes, update the global weight library; if the verification fails, roll back the weight adjustment and re-analyze.

2. The method according to claim 1, characterized in that, The weight adjustment described in step S3 covers all sub-characteristics at four levels: talent-side surface traits, abilities and skills, personality traits, and growth potential, as well as all sub-characteristics at six dimensions: target-side market, technology, business, finance, compliance, and value; among them, the weights of value characteristics and external image characteristics can be adjusted independently.

3. The method according to claim 1, characterized in that, The triggering conditions for weight adjustment include any one or more of the following combinations: - Reaching a preset periodic adjustment cycle; - Switch to different application scenarios; - A certain amount of new matching data and feedback data has been accumulated; - The matching performance index is lower than the preset threshold; - Received a weight adjustment instruction manually triggered by the user.

4. The method according to claim 1, characterized in that, The feature contribution analysis described in step S2 employs at least one of the following methods: - Statistical analysis method: calculating the correlation coefficient between each feature and the matching result; - Machine learning methods: Calculate the contribution of each feature through feature importance evaluation algorithms (such as SHAP, LIME, random forest feature importance); - Reinforcement learning method: Learns the optimal weight allocation strategy through interaction with the environment.

5. The method according to claim 1, characterized in that, It also includes the steps for managing scenario-based weight templates: Pre-set corresponding initial weight templates for different application scenarios; Based on historical data for each scenario, automatically optimize the exclusive weight template for each scenario; Users can create custom weight templates based on the basic template.

6. The method according to claim 1, characterized in that, It also includes steps to support user-defined weights: Receive user-inputted custom base weights; Based on user-defined base weights, data-driven fine-tuning is performed; Retain the priority of user-defined weights to ensure that system adjustments do not deviate from the core needs of users.

7. The method according to claim 1, characterized in that, It also includes the step of generating interpretability for weight adjustment: Generate a weight adjustment report, explaining the magnitude, reasons, and basis for the weight adjustments of each feature; Show a comparative analysis of the matching effect before and after weight adjustment; Provide suggestions and explanations for weight adjustments.

8. The method according to claim 1, characterized in that, It also includes a weight stability control step: Set a maximum limit on the weight adjustment to prevent drastic fluctuations in weight; The weight adjustment process is smoothed by using a sliding window averaging method. It retains historical weighted versions and supports rolling back to any historical version at any time.

9. The method according to claim 1, characterized in that, The weight adjustment described in step S3 adopts a multi-objective optimization strategy, simultaneously optimizing at least two of the following objectives: - Maximizing matching accuracy; - Maximize the success rate of matching; - Maximize user satisfaction; - Maximize the diversity of matching results; - Minimize the magnitude of weight adjustments.

10. The method according to claim 1, characterized in that, The method can be applied to any one or more of the following scenarios: Education and training scenarios, school selection scenarios, career planning scenarios, corporate recruitment scenarios, headhunting service scenarios, venture capital scenarios, technology transfer scenarios, industry-university-research cooperation scenarios, and team building scenarios.

11. A data-driven human-identity matching dynamic weight adaptive adjustment system, characterized in that, include: The data acquisition module is used to acquire historical matching data, user feedback data, and scene feature data. The contribution analysis module is used to analyze the actual contribution of each feature to the matching results; The dynamic weight adjustment module is used to automatically adjust the matching weight of each feature based on its contribution. The effect verification module is used to verify the matching effect after weight adjustment; The weight library management module is used to store, update, and manage the global weight library and scenario-based weight templates; The user interaction module supports users in defining weights, viewing weight adjustment reports, and manually triggering weight adjustments.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 10.

Images