An automobile industry production, teaching and evaluation integrated cloud platform system

Abstract

This invention discloses a cloud platform system integrating industry, education, and assessment in the automotive industry, specifically in the field of education management. It includes: an industry resource integration and intelligent semantic association module that collects multi-source data, preprocesses it, extracts skill entities and generates vectors, and calculates skill association strength and skill popularity values; an adaptive course assembly and virtual training module that constructs a skill dependency graph, generates personalized learning paths, builds a virtual training environment, and collects operational data; a comprehensive teaching and assessment credit model module that generates a comprehensive credit value based on training data, using a three-dimensional evaluation function combined with an attention mechanism, and stores it on a blockchain; and a talent profiling and intelligent matching module that integrates data from multiple modules to construct a profile, calculates three types of matching degrees, and outputs a comprehensive score and matching level through a dynamic weight model. A closed loop is formed between modules, with matching feedback feeding back into the model optimization. This invention achieves data interoperability between industry, education, and assessment, improving the adaptability of talent training to job requirements.

Description

Technical Field

[0001] This invention relates to the field of education management technology, and more specifically, to a cloud platform system integrating production, education, and assessment in the automotive industry. Background Technology

[0002] The automotive industry is currently undergoing a rapid transformation towards electrification, intelligence, and connectivity. New jobs such as new energy battery research and development, intelligent driving algorithm optimization, and vehicle networking system operation and maintenance are emerging continuously, and the skill requirements are becoming more diversified and precise. In the field of vocational education, schools generally connect with industry needs through school-enterprise cooperation models. More than half of the schools offering automotive majors have introduced virtual training systems and combined them with online teaching platforms to carry out targeted skills training. In the talent evaluation process, the participation of industry enterprises has increased significantly, forming a comprehensive evaluation system covering theoretical knowledge and practical skills. Some leading enterprises have taken the lead in formulating evaluation standards for specific fields. Meanwhile, automotive companies, educational institutions, and evaluation organizations have accumulated a wealth of data on job requirements, teaching and training, and assessment results. Some leading companies in the industry chain have built dedicated training data platforms to try to integrate internal resources and teaching data, and promote the initial linkage between industry needs and teaching content.

[0003] However, it still has some drawbacks in practical use, such as: 1. Industry data and teaching data in the automotive industry are stored separately in the independent systems of enterprises, colleges and universities, and evaluation agencies. The lack of a unified semantic association and standardized integration mechanism leads to the reliance on manual matching of job skill requirements and teaching content, which is prone to delays or deviations in the transmission of requirements. 2. The matching of skills and positions is mostly based on static labels or human experience, without dynamic adjustment based on the real-time popularity and relevance of skills in the industry. This makes it difficult to adapt to the new skill demands that are rapidly emerging in the transformation of automobiles towards electrification and intelligence, resulting in a structural mismatch between talent supply and job demand. 3. The talent evaluation system focuses more on theoretical exam scores or the results of a single practical task, and lacks quantitative analysis of multiple dimensions such as operational proficiency, practical innovation, and breadth of knowledge application. The evaluation results are difficult to fully reflect the trainees' industry adaptability and have limited reference value. 4. There is a lack of data flow loop between the updating of teaching resources, the design of practical training content, talent evaluation standards and job matching results. Skill gaps found in the evaluation and matching process cannot be fed back into the teaching process in a timely manner, resulting in the iteration of teaching content lagging behind changes in industry demand and low efficiency of industry-education collaboration. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art, the present invention provides a cloud platform system integrating production, education, and evaluation in the automotive industry, which solves the problems mentioned in the background art through the following solutions.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a cloud platform system integrating production, education, and assessment in the automotive industry, comprising: The industrial resource integration and intelligent semantic association module collects multi-source data from the automotive industry. After algorithm-based desensitization and secure transmission, it extracts skill entities through a semantic model and transforms them into 768-dimensional vectors. It also constructs a cosine similarity skill entity association strength function and builds a skill demand heat assessment model based on multi-dimensional data to output skill heat values. The adaptive course assembly and virtual training module constructs a dependency directed graph based on the strength of skill entity associations, combines skill popularity values ​​with students' initial levels to generate learning paths, and matches teaching resources; it also builds a virtual training environment and collects students' training data in real time. The teaching evaluation comprehensive credit model module receives and preprocesses students' practical training data, constructs a three-dimensional basic evaluation function, evaluates students' practical training, and generates a comprehensive credit value through attention mechanism fusion, which is then stored on the blockchain to form an electronic certificate. Talent profiling and intelligent matching module: integrates skill popularity value, trainee training data, and comprehensive credit value to build a profile; calculates the matching degree of core skills, industry practice, and scenario adaptation, generates a comprehensive matching score through a dynamic weight model, and outputs the matching level.

[0006] The technical effects and advantages of this invention are as follows: By using industry-specific BERT semantic models and cosine similarity algorithms, we can intelligently associate and dynamically integrate scattered multi-source industry and teaching data, construct a skills semantic network, realize automated and accurate matching of job requirements and teaching resources, break down data silos, and improve the efficiency and accuracy of resource integration. Based on the AHP-entropy weight coupling model, skill popularity values ​​are generated. Combined with the skill correlation strength, learning paths are dynamically planned, and teaching and training content of high-popularity skills are prioritized. This can respond to the skill iteration rhythm of the automotive industry in real time and ensure that the talent training direction is synchronized with industry needs. We construct a three-dimensional evaluation model that assesses operational proficiency, practical innovation, and knowledge mastery. By combining attention mechanisms and blockchain notarization technology, we achieve a quantitative, comprehensive, and credible evaluation process, objectively reflecting trainees' overall abilities and providing enterprises with accurate talent references. Establish a closed-loop mechanism for matching results feedback and resource and curriculum optimization, feed back enterprise employment feedback data to the industry resource integration and curriculum assembly module, continuously iterate the skills popularity model and learning path, promote dynamic collaboration among industry, education and assessment, and improve the industry adaptability of talent training. Attached Figure Description

[0007] Figure 1 This is a schematic diagram of the overall structure of the present invention.

[0008] Figure 2 This is a schematic diagram of the adaptive course assembly and virtual training module structure of the present invention.

[0009] Figure 3 This is a schematic diagram of the module structure of the teaching evaluation integrated credit model of the present invention.

[0010] Figure 4 This is a schematic diagram of the talent profiling and intelligent matching module of the present invention. Detailed Implementation

[0011] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0012] refer to Figures 1-4 The cloud platform system for integrating production, education, and assessment in the automotive industry, as shown, includes: Industry resource integration and intelligent semantic association module: This module provides the core data foundation for the entire platform. Through secure aggregation of multi-source industry data, intelligent semantic parsing and processing, dynamic semantic network construction, and intelligent demand perception and prediction, it outputs two core quantitative data points: a cosine similarity function to output skill association strength, and a skill demand popularity assessment model to output skill popularity values, supporting subsequent course assembly and talent matching functions. A detailed analysis follows: Multi-source industry data security aggregation unit: Construct a distributed data acquisition network, deploy acquisition nodes according to the six major sub-sectors of the automotive industry chain, and integrate load balancing algorithms to allocate acquisition tasks to avoid single-point overload; It should be further explained that the six major sub-sectors of the automotive industry chain include: vehicle manufacturing, parts production, after-sales maintenance, intelligent driving research and development, new energy technology, and automotive electronics. It features a built-in enterprise data security gateway, an integrated digital certificate verification mechanism based on the national cryptographic SM2 algorithm, an API key management unit, and supports automatic key expiration and reset with customizable expiration periods to ensure trusted identity during data transmission. Deploy a data anonymization rule engine, preset a list of anonymized fields such as core enterprise technical parameters and sensitive employee information, adopt a dual processing scheme of replacement encryption and hash salting, with a salt value length of 16 bits, hash algorithm of SHA-256, and explicit encryption algorithm parameter configuration; A data traffic monitoring subsystem is built to collect data transmission bandwidth, data packet volume, connection status and other indicators in real time. It has a built-in abnormal transmission threshold (single IP daily transmission volume ≥10GB) and alarm triggering module to intercept risky transmission behaviors in real time.

[0013] Configure a dual-mode data collection strategy of incremental crawling and full synchronization, and build a data quality inspection mechanism to filter invalid data through rules such as field integrity verification, logical consistency verification, and deduplication of duplicate data, so as to ensure data integrity ≥95% and accuracy ≥98%.

[0014] Intelligent semantic parsing and processing unit: It has a built-in semantic understanding model specifically for the automotive industry, and is based on the BERT pre-trained model with industry-adapted fine-tuning. The training data covers more than 800,000 texts of various types, including job descriptions, patent documents, training materials, and technical manuals in the automotive industry. It integrates industry-specific named entity recognition models, presets core entity categories such as skill names, equipment models, process steps, and technical standards, and optimizes entity recognition algorithms to ensure an accuracy rate of ≥89% for industry terminology recognition.

[0015] A dependency parser is built to extract semantic relationships such as subject-verb, verb-object, and modifier-head between entities, and a context semantic matching mechanism is constructed to solve the problem of ambiguity in industry terms. The BERT model is used to vectorize skill entities, generating 768-dimensional high-dimensional vectors. After L2 normalization, the vectors are stored in the Milvus vector database, supporting efficient retrieval. The built-in vector similarity calculation system uses the cosine similarity algorithm as its core to optimize calculation efficiency and ensure batch processing speed of ≥1000 items / second.

[0016] Dynamic Semantic Network Building Blocks: The specific process of constructing the skill entity association strength function is as follows: Construct a cosine similarity function to calculate the association strength between any two skill entities: ,in: and are the 768-dimensional vector representations of skill 1 and skill 2 after BERT model vectorization and L2 normalization, respectively; It is the dot product of vectors; , These are the L2 norms of the vectors.

[0017] A skills demand heat assessment model is constructed based on the coupling of AHP with entropy weight and nonlinearity: in, For objective combination weights: , As subjective weights, a three-level judgment matrix is ​​constructed using the Analytic Hierarchy Process (AHP), resulting in: , , ; For objective weighting, entropy values ​​are calculated based on three years of industry data. ,pass Dynamic adjustment; The Sigmoid normalization function, For time decay term, The time decay coefficient, determined through fitting automotive industry data, controls the decay rate. The interval between the current time and the original data collection time, in months, where e is an irrational number; This is a non-linear indicator function that performs a non-linear transformation on the original indicator data to enhance data discriminability, suppress extreme value interference, and reflect the synergistic effect within the indicator. Specifically, it is as follows: Demand Dimension Function: P represents the frequency of a certain skill appearing in corporate job postings over the past year. The average frequency of job positions across all skills in the entire industry. Compress the numerical range of high-frequency skills. Amplify the skill advantages that are above the industry average demand, and ultimately highlight the skills that are in high demand and significantly better than the industry average. Innovation dimension function: C represents the total number of citations for patents related to this skill. The average number of citations for skill-related patents across the entire industry; I is the innovation factor of the skill, ranging from 0 to 1, quantified based on expert evaluation of the skill. It reflects the synergy between recognition and innovation. Suppress the excessive influence of patents with extremely high citation counts on the results; Value dimension function: D represents the intensity of demand for this skill within the enterprise, based on the frequency normalization of the enterprise's internal training plans and technology upgrade projects, with a value range of 0-1. B represents the commercial value coefficient, based on the market size and profit margin of the skill-related products, with a value range of 0-1. The product of the two reflects the binding between enterprise demand and market value, and the mean term enhances the complementary effect between the two. This skill demand heat assessment model, through a three-layer design of nonlinear index transformation, combined weight fusion, and time decay, avoids the one-sidedness of single index assessment and solves problems such as differences in the scale of raw data, interference from extreme values, and insufficient data timeliness. The final output Heat (skill) value can truly and dynamically reflect the market demand heat and development potential of automotive industry skills, providing data support for subsequent course assembly, talent training, and job matching in the system. A dynamic semantic network is constructed using the Neo4j graph database, with skill entities as nodes, association strength as edge weights, and popularity as node attributes, forming a visual semantic association graph. The built-in semantic network automatic update mechanism automatically triggers the recalculation of association strength and popularity values ​​when new skill entities are added or existing data is updated, achieving dynamic iteration and data synchronization.

[0018] Intelligent demand perception and prediction unit: Integrating an LSTM time series forecasting model, based on historical industry job demand data from the past three years, the system analyzes the cyclical and trend characteristics of skill demand, outputting quarterly and annual demand forecasts. An abnormal demand fluctuation detection mechanism is established, using the 3σ principle to set an abnormal threshold; when data exceeds the threshold, it automatically marks the abnormality and triggers a cause analysis process. A skill evolution path prediction algorithm is designed, combining multi-dimensional data such as technological development trends, patent innovation directions, and changes in enterprise demand to mine skill derivative relationships and generate a visualized skill evolution path map.

[0019] Adaptive course assembly and virtual training module: Based on the skill association strength and skill popularity values ​​output by the industry resource integration and intelligent semantic association module, and combined with personalized student data, an adaptive learning path is generated to construct a high-fidelity virtual training environment. This achieves a closed-loop operation of learning, practice, and evaluation, ensuring that course content accurately matches industry needs. The specific analysis is as follows: Learning Path Planning Unit: Built-in core automotive industry skills library, covering all automotive industry skills, clearly defining the pre-requisite, parallel, and subsequent dependencies between skills. The basic dependency strength relationship is optimized by combining the skill association strength output by the industry resource integration and intelligent semantic association module: association strength ≥ 0.7 is judged as strong dependency and is given priority in path planning; association strength < 0.3 is judged as weak dependency and the learning order can be freely adjusted.

[0020] A directed graph model of skill dependency is constructed using the Neo4j graph database. Nodes represent skill items, and edges represent dependency relationships. Edge weights are a fusion of basic dependency strength and skill association strength output by the industry resource integration and intelligent semantic association module. Dependency strength and learning priority are labeled.

[0021] Integrating the Dijkstra path optimization algorithm, based on a skill-dependent directed graph, and combining students' entrance skill level test data with learning objectives and skill popularity values, the algorithm generates the optimal learning path, prioritizing the learning order of skills with the highest popularity and highest correlation strength.

[0022] We designed a personalized learning path adaptation mechanism, which distinguishes between visual, auditory, and hands-on learning styles through a learning style questionnaire, and matched adaptation paths for different learners. The difficulty gradient of each path is ≤20% per unit.

[0023] Intelligent matching unit for teaching resources: Establish a multi-dimensional annotation system for teaching resources. Annotation dimensions include knowledge type (theoretical or practical), difficulty level (levels 1-5), presentation format (video, document, animation or virtual simulation), corresponding skill points, and applicable learning styles.

[0024] Establish a resource quality assessment system, setting accuracy assessment indicators such as an expert review pass rate of ≥98%, timeliness assessment indicators such as the proportion of resources in the past 3 years of ≥80%, and interactivity assessment indicators such as ≥3 interactive sessions per practical resource. A quality score is generated through expert review and trainee feedback.

[0025] Construct a resource-skill mapping database, identify the set of high-quality teaching resources corresponding to each skill point, and support quick retrieval by skill point; Design a dynamic resource recommendation engine that adjusts recommended content in real time based on learners' learning progress, skill mastery, and operational feedback data, providing 3-5 different alternative resources for the same skill point, with a recommendation response time of ≤1 second.

[0026] Virtual training environment construction unit: We have built a 3D equipment model library for the automotive industry. Based on real parameters of mainstream car brands, we have used Unity3D for high-fidelity modeling, covering core components such as engines, transmissions, chassis, and circuit systems. The model accuracy is ≥95%.

[0027] It integrates the PhysX physics engine to simulate the mechanical feedback, motion trajectory, and equipment linkage effects in real-world operations, and restores the physical characteristics of practical operation scenarios such as disassembly, maintenance, and debugging.

[0028] It has a built-in standardized operation procedure library, which clarifies the operation steps, key points, operation specifications, and safety requirements for each training project, and embeds a virtual environment as an operation guide.

[0029] Deploy an operation behavior collection unit to record data such as student operation steps, sequence, time consumption, key node hit rate, number of erroneous operations and types at a frequency of 10ms / time.

[0030] An operational behavior analysis model is constructed to identify problems such as non-standard operations, process omissions, and operational errors based on collected data. The model automatically generates an operational behavior analysis report, and the report results are used to adjust the learning path and supplement the trainee's talent profile, providing a matching basis for subsequent modules.

[0031] Comprehensive Credit Model Module for Teaching Evaluation: A multi-dimensional, traceable, and objective teaching evaluation system is constructed. This system collects practical training data from adaptive course assembly and virtual training modules, combines it with theoretical test and project practice data, and generates a comprehensive student credit score through non-linear coupling evaluation calculation. Blockchain storage ensures credibility and verifiability, and the final comprehensive credit score is output. The specific analysis is as follows: Multi-source data acquisition unit: Virtual operation data acquisition: Through the built-in acquisition plugin of the virtual training environment, data such as mouse or gamepad action coordinates, operation trigger events, operation sequence, error trigger points, and operation duration are recorded.

[0032] Project practice data collection: Integrate Git commit statistics tools to count code commits, lines of code, and pass rate; use document-based detection tools to check originality and completeness; and collect data on code commit volume, document quality, task completion progress, and team collaboration contributions.

[0033] Theoretical knowledge test data collection: Build a hierarchical knowledge test system: basic level accounts for 40%, advanced level 30%, and application level 30%. Set the question types and score proportions for each level, and collect data such as answering time, accuracy rate, knowledge point error distribution, and repeated error rate.

[0034] Data preprocessing module: Pandas is used for data cleaning to remove abnormal data with operation time exceeding 3 times the normal range. Numerical missing values ​​are filled with the mean, and categorical missing values ​​are filled with the mode. All data are standardized to the range [0, 1].

[0035] Basic evaluation function calculation unit: Through nonlinear synergistic coupling, an operational proficiency function is constructed based on virtual operational data. Its specific mathematical function is as follows: ,in: To ensure the correct number of operation steps, Total operation time; The importance weight coefficient for the i-th advanced operation in automotive industry virtual training is dynamically calculated based on the variance of the collected operation error rate using the entropy weight method. This is the reward coefficient for advanced operations; it becomes 1 upon completion. A new collaboration item has been added. This demonstrates the synergistic gain of multiple advanced operations; the results are... Truncation yields a new function value; The practical innovation degree function is constructed based on Copula functional coupling, and its specific mathematical function is as follows: ;in: The ternary Archimedean Copula function captures three-dimensional correlations; , , The result was obtained through maximum likelihood estimation, fitted to 1000 sets of historical data; C is the technology complexity score, and U is the uniqueness coefficient. The average score is the peer review score.

[0036] The knowledge mastery degree function is constructed based on complementary coupling, and its specific mathematical function is as follows: Among them: newly added complementary items Rewards are given to students who have a balanced foundation in basic knowledge and practical skills. , Determined through Delphi-entropy weight coupling; Scoring based on basic knowledge To score points for applying knowledge, The speed reward factor has a value of 0.1-0.3.

[0037] Function verification module: Tested with a sample of 1000 students to ensure that the calculation results are ≥90% consistent with human evaluation.

[0038] Advanced fused computing unit: The comprehensive credit score is calculated through attention mechanism coupling, and its comprehensive credit function is as follows: Where: attention weight , Determined through data fitting; product term This reflects the "barrel effect" and suppresses imbalances caused by a single prominent dimension; the results are mapped to the [0, 1] interval.

[0039] Blockchain-based evidence storage unit: Deploy a Hyperledger Fabric consortium blockchain system, with consortium nodes including relevant parties such as universities, enterprises, and industry associations, to achieve multi-party supervision and consensus verification of evidence-based data.

[0040] Setting the threshold for electronic certificate generation: Student's overall credit score When the score is 1.0, an electronic certificate will be automatically generated, which includes the student's basic information, skills list, evaluation results in various dimensions, comprehensive credit score, timestamp and other core information.

[0041] Electronic certificate processing subunit: Calculates the digital fingerprint (hash value) of the certificate, writes it into the consortium blockchain to generate a unique verification identifier, and ensures that the certificate cannot be tampered with.

[0042] Certificate verification mechanism: Provides two methods: QR code verification and interface verification. It supports third parties to query the authenticity and details of the certificate through the verification mark, and realizes cross-platform mutual recognition.

[0043] Talent profiling and intelligent matching module: Based on the skill association strength and skill popularity values ​​output by the industry resource integration and intelligent semantic association module, the practical training data output by the adaptive course assembly and virtual training module, and the comprehensive credit value output by the teaching evaluation comprehensive credit model module, a precise talent profile is constructed. By constructing three matching functions to calculate the comprehensive matching score, and combining it with a dynamic weight generation model, the matching of trainees with automotive industry positions is achieved, and standardized matching results are output. The specific analysis is as follows: Talent Profile Core Dimension Extraction Unit: Hard skill indicators: Automatically quantify the core skill level index of the automotive industry, ranging from 0 to 10 points, skill mastery proficiency, and skill update timeliness (number of new skills added in the past 6 months). Skill mastery proficiency is calculated based on operational accuracy and cumulative error rate, skill update timeliness is the number of new skills added in the past 6 months, and skill priority is marked based on the skill popularity value output by the industry resource integration and intelligent semantic association module. High-popularity skills are listed separately as the scarce skills dimension.

[0044] Industry practice data dimensions: collect data on virtual training project completion rate, operation accuracy rate, key step pass rate, and project time; and statistically analyze the output of real or simulated projects.

[0045] Job scenario adaptability dimension: Records data on the adaptation of the operating environment in the virtual training module of the adaptive course assembly, the candidate's preset regional adaptation preference, and the fit between the comprehensive skill level and the job level threshold.

[0046] Comprehensive credit dimension: The comprehensive credit value output by the teaching evaluation comprehensive credit model module and the status of the electronic certificate stored on the blockchain are directly incorporated as core indicators of talent credibility.

[0047] Basic matching function calculation unit: The core skill matching function is based on the skill association strength of the industry resource integration and intelligent semantic association module, coupling skill relevance. Its specific mathematical function is as follows: in: The skill association strength is calculated by directly calling the output of the industry resource integration and intelligent semantic association module; Based on industry demand frequency, the results were weighted and optimized using skill popularity values: Update the optimized value to the new ; This is a quantitative value representing the candidate's overall level in skill x required for position y. , where is the candidate's theoretical test score on skill i (in the range of 0-1), and is the application layer score from the knowledge mastery function of the teaching evaluation integrated credit model module. Only the scores of test questions directly related to skill i are extracted, and after normalization, the scores are mapped to [0, 10]. The cumulative error rate for skills is calculated based on the number of erroneous operations collected by the operation behavior collection unit. The average error rate threshold; Industry practice matching function: Based on adaptive course assembly and virtual training module training data, coupled through synergistic effects, its specific mathematical function is as follows: in: The completion rate of job-related virtual training projects output by the adaptive course assembly and virtual training module; The average operational accuracy rate of the relevant training projects; The weighting coefficients are determined by grey relational analysis based on the correlation strength between adaptive course assembly and virtual training module training data and post-employment performance data in enterprises, in order to ensure the output quantity consistent with the technical direction of the job. The job scenario adaptation function, based on multi-dimensional correlation and coupling, has the following specific mathematical function: in: The adaptive course assembly and virtual training module records the compatibility rate between the job work environment and the candidate training environment. The degree of matching between the candidate's overall skill level (integrated credit score) and the job level threshold; For regional adaptability; Comprehensive matching execution unit: Dynamic weight generation: Weight adjustment is driven by job characteristics and integrates skill popularity values. in, The ideal threshold for the q-th dimension of a job position is preset based on the characteristics of job types in the automotive industry. The basic weights for job types are set based on the core requirements of different job types in the automotive industry, combined with industry characteristics. Overall Match Score: The score range is 0-1 points, and is rounded to two decimal places. Matching level classification: Highly compatible: 0.8-1.0, no pre-job training required; Moderately compatible: 0.6-0.79, 1-2 weeks of specialized training; Basic compatible: 0.4-0.59, 1-3 months of systematic training; Not compatible: 0-0.39, core dimensions not met.

[0048] Matching Execution Process: Starting with job requirement analysis, based on the skill popularity value output by the Industry Resource Integration and Intelligent Semantic Association module, the system automatically extracts the core high-popularity skills contained in the job description, clarifying the priority requirements for industry-scarce skills for the position. On this basis, parallel computation is initiated, inputting objective data across the entire chain, including the skill association strength and popularity value from the Industry Resource Integration and Intelligent Semantic Association module, virtual training operation data from the Adaptive Course Assembly and Virtual Training module, and the comprehensive credit value from the Teaching Evaluation and Credit Model module. The entire calculation process is free from human intervention, ensuring the objectivity of the results. After calculation, the system enters the score ranking and recommendation stage. It prioritizes candidates with high suitability for high-popularity skills and excellent comprehensive credit values ​​from the Teaching Evaluation and Credit Model module, along with detailed data sources for each dimension's scores for companies to review and verify. Finally, a feedback iteration mechanism is formed. Based on the actual performance data of candidates after they take up their posts, as provided by companies, the system reverse-optimizes the time decay coefficient and combined weight parameters of the Heat(skill) function in the Industry Resource Integration and Intelligent Semantic Association module, while simultaneously adjusting the priority of course content and the design of training projects in the Adaptive Course Assembly and Virtual Training module.

[0049] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other. In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An automobile industry production, teaching and evaluation integrated cloud platform system, characterized in that, include: The industrial resource integration and intelligent semantic association module collects multi-source data from the automotive industry. After algorithm-based desensitization and secure transmission, it extracts skill entities through a semantic model and transforms them into 768-dimensional vectors. It also constructs a cosine similarity skill entity association strength function and builds a skill demand heat assessment model based on multi-dimensional data to output skill heat values. Adaptive course assembly and virtual training module: Construct a dependency directed graph based on the strength of skill entity associations, combine skill popularity values ​​and students' initial levels to generate learning paths, and match teaching resources; Build a virtual training environment and collect students' training data in real time; The teaching evaluation comprehensive credit model module receives students' practical training data, constructs a three-dimensional basic evaluation function after preprocessing, evaluates students' practical training, and generates a comprehensive credit value through attention mechanism fusion. After being stored on the blockchain, it forms an electronic certificate. Talent profiling and intelligent matching module: integrates skill popularity value, trainee training data, and comprehensive credit value to construct a profile; The system calculates the matching degree of core skills, industry practices, and scenario adaptation, generates a comprehensive matching score through a dynamic weight model, and outputs the matching level.

2. The cloud platform system for integration of production, teaching and evaluation in the automobile industry according to claim 1, characterized in that: The constructed cosine similarity skill entity association strength function includes: The extracted automotive industry skill entities within the module are vectorized to generate feature vectors corresponding to each skill entity and normalize them. Using the feature vectors of any two skill entities as input parameters, a cosine similarity skill entity association strength function is constructed. By calculating the ratio of the inner product of the two feature vectors to the product of the L2 norms of the two feature vectors, the association strength value representing the degree of association between the two skill entities is output.

3. The cloud platform system for integration of production, teaching and evaluation in the automotive industry according to claim 1, characterized in that: The constructed skills demand heat assessment model includes: First, multi-dimensional foundational data supporting the skills demand assessment are determined, including the frequency of skill occurrence in industry demand, citation data of skills-related innovative achievements, the intensity of internal demand for skills by enterprises, and the corresponding commercial value coefficient of skills. Next, a combined weighting calculation mechanism is constructed, combining subjective weights determined by industry experience with objective weights calculated based on the distribution entropy of the foundational data. A coupling algorithm generates a comprehensive weight for each assessment dimension, balancing subjective industry perception with objective data characteristics. Then, the foundational data for each dimension are transformed to form quantitative assessment values ​​for each dimension. The quantitative assessment values ​​for each dimension are then weighted and summed with their corresponding comprehensive weights, mapped to the 0-1 interval using a Sigmoid normalization function. A time decay factor is introduced to adjust the contribution of the demand based on the interval between the foundational data collection time and the current time, suppressing the interference of outdated data on the assessment results. Finally, a standardized skills demand demand value is generated.

4. The cloud platform system for integrating production, education, and assessment in the automotive industry according to claim 1, characterized in that: The adaptive course assembly and virtual training module includes: Based on the skill association strength and skill popularity values ​​output by the industry resource integration and intelligent semantic association module, a directed graph of skill dependencies is constructed. Combined with the learners' initial skill levels, a path optimization algorithm generates personalized learning paths with priority given to high-popularity skills, following a progressive difficulty logic. Teaching resources are matched to the skill difficulty and learners' learning styles, and a dynamic resource adjustment mechanism is established to update with skill popularity. A virtual training environment is built to simulate the characteristics and process specifications of practical operation scenarios. Learners' practical operation data is collected at a preset frequency. The training data is output to the teaching evaluation comprehensive credit model module, and the learning paths and training content are iteratively optimized with skill popularity.

5. The cloud platform system for integrating production, education, and assessment in the automotive industry according to claim 1, characterized in that: The three-dimensional basic evaluation function includes: Based on the practical training data, theoretical test data, and project practice data, standardized preprocessing was used as the input for the functions. An operation proficiency function was constructed: using the number of correct steps and total time consumed in the practical training as basic parameters, advanced operation weights and synergistic gain terms were introduced to quantify the accuracy and efficiency of the operation. A practice innovation function was constructed: using project technical complexity scores, solution uniqueness coefficients, and peer review average scores as variables, multivariate correlations were coupled through a Copula function to suppress extreme value interference. A knowledge mastery function was constructed: integrating the basic knowledge score and applied knowledge score from the theoretical test, with weights determined by Delphi-entropy weight coupling, and incorporating a speed reward factor and complementary terms for basic and applied abilities. The outputs of all three functions were mapped to the [0, 1] interval.

6. The cloud platform system for integrating production, education, and assessment in the automotive industry according to claim 1, characterized in that: The generation of the comprehensive credit score includes: The system obtains quantitative evaluation values ​​in the [0, 1] interval from the output of the three-dimensional basic evaluation functions to ensure the standardization of input data. An attention fusion mechanism is then constructed, using the evaluation values ​​of each dimension as input. A softmax function with temperature parameter adjustment is introduced to calculate the attention weights for each dimension, tilting the weights towards the advantageous dimensions with higher evaluation values, thus achieving differentiated weighted fusion. A weakness penalty factor is then introduced, extracting the minimum value among the three-dimensional evaluation values. A preset penalty formula incorporates the impact of the weakness dimension into the calculation, weakening the evaluation bias caused by a single prominent dimension and obvious weakness. Finally, the attention-weighted fusion result is superimposed with the result after weakness penalty, and linear normalization is applied to ensure that the final comprehensive credit value remains in the [0, 1] interval.

7. The cloud platform system for integrating production, education, and assessment in the automotive industry according to claim 1, characterized in that: The process of constructing the profile includes: The system integrates skill popularity values ​​from the industry resource integration and intelligent semantic association module, student training data from the adaptive course assembly and virtual training module, and comprehensive credit values ​​and electronic certificate status from the teaching evaluation comprehensive credit model module; it is constructed in three dimensions: hard skill indicators, industry practice, and comprehensive credit; and the data in each dimension are standardized to form a structured profile.

8. The cloud platform system for integrating production, education, and assessment in the automotive industry according to claim 1, characterized in that: The matching degree of the core computing skills, industry practices, and scenario adaptation includes: When calculating the core skill matching degree, the core skill vector of the job and the hard indicator vector of the talent profile are used as inputs. The skill association strength output by the industry resource integration and intelligent semantic association module is called as the correlation coefficient. Combined with the industry demand frequency weighted by the skill popularity value, the matching value in the range of [0, 1] is output through the vector similarity algorithm. When calculating the matching degree of industry practice, the data of the industry practice dimension in the job practice requirements and the profile are extracted. The completion rate of the training project and the accuracy of operation are the core parameters. The balance between the two is coupled by the harmonic average and the project result adaptation coefficient is superimposed to generate a quantitative matching result. When calculating the scenario adaptability, the characteristics of the job work environment are associated with the training environment data of Module 2, the environment adaptability rate and the matching degree between skill level and job threshold are quantified, regional adaptability preference and interaction gain are introduced, and the comprehensive adaptability score is output; the three types of matching results are all standardized to the [0, 1] interval.

9. A cloud platform system integrating production, education, and assessment in the automotive industry according to claim 1, characterized in that: The generation of the comprehensive matching score includes: The standardized matching results based on three categories—core skills, industry practice, and scenario adaptation—serve as the basic input for score calculation. A dynamic weight model is constructed, with the similarity between each matching degree result and the ideal threshold of the corresponding job as the adjustment factor, coupled with the basic weights preset by the job type, and the dynamic weights of the three matching degrees are generated after normalization; the matching degree result of each type is multiplied by the corresponding dynamic weight and then summed to obtain the comprehensive matching score in the interval [0, 1]. Based on the scores, four matching levels are defined: high, medium, basic, and unsuitable. The score details and level labels are output simultaneously.

Images