Talent recommendation system for matching local productivity based on vocational education

Abstract

The invention discloses a talent recommendation system for matching local productivity based on vocational education, and relates to the technical field of talent recommendation. Comprising a dynamic industry portrait construction module, an expert auditing and correcting module, a personalized path planning module, a talent growth prediction module, a path adjustment module, a talent recommendation module, a data storage module, an interactive interface module and a system management module. The dynamic industry portrait construction module predicts industry trend data by introducing a time sequence hidden Markov model of an attention mechanism, and the dynamic industry portrait construction module and the expert auditing and correcting module integrate undisclosed project information supplemented by Bayesian reasoning based on a knowledge graph completion algorithm. And generating a dynamic industry knowledge graph based on a time decay factor through a graph convolutional network feature fusion technology. According to the invention, the dynamic industry portrait construction module and the expert auditing and correcting module are arranged, so that the problem that the industry information lags behind and the local industry dynamic change cannot be reflected in real time in the existing talent recommendation system can be solved.

Description

Technical Field

[0001] This invention relates to the field of talent recommendation technology, and in particular to a talent recommendation system based on matching vocational education with local productivity. Background Technology

[0002] The talent recommendation system based on matching vocational education with local productivity is an intelligent platform that uses big data and artificial intelligence technologies to integrate vocational education talent training data with local industry job demands, achieving precise matching of talent skills with industry needs. Its role is to help students find employment efficiently, help enterprises reduce recruitment costs, promote the optimization of vocational education majors, and promote the upgrading of regional productivity layout. Its aim is to break down information barriers between talent supply and demand, build a virtuous cycle of talent, industry, and economy, and empower the sustainable development of regional economy with vocational education resources.

[0003] Currently, in terms of industry information acquisition and updating, existing talent recommendation systems use relatively simple models to predict industry trends. These models fail to fully consider the dynamics and complexity of industry development, resulting in predictions of industry trend data lagging behind actual changes. This makes it difficult to adapt to the rapidly changing industrial environment. Furthermore, the system lacks the ability to acquire and integrate information on undisclosed projects within the industry, hindering the comprehensive and in-depth construction of industry profiles. Consequently, talent recommendations lack sufficient information support and fail to accurately match the actual needs of local industries. On the other hand, in terms of career development path planning, systems rely solely on fixed industry standards or limited historical data, ignoring the speed of industry technology updates and individual differences. The planned career paths lack flexibility and specificity, failing to meet the development needs of talents at different stages. In the talent-job matching process, the matching algorithms used are relatively simple, failing to deeply understand the intrinsic connection between job skill requirements and talent growth potential. This leads to low matching accuracy, causing many suitable talents to miss out on jobs, resulting in a waste of human resources and increased recruitment costs for enterprises.

[0004] Therefore, a talent recommendation system based on matching vocational education with local productivity is proposed to solve the above problems. Summary of the Invention

[0005] The main objective of this invention is to provide a talent recommendation system based on matching vocational education with local productivity, in order to solve the problems mentioned in the background above.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: a talent recommendation system based on matching vocational education with local productivity, the talent recommendation system based on matching vocational education with local productivity includes a dynamic industry profile construction module, an expert review and correction module, a personalized path planning module, a talent growth prediction module, a path adjustment module, a talent recommendation module, a data storage module, an interactive interface module, and a system management module;

[0007] The dynamic industry profile building module predicts industry trend data by introducing a temporal hidden Markov model with an attention mechanism. The expert review and correction module supplements unpublished project information by incorporating Bayesian inference based on a knowledge graph completion algorithm. The dynamic industry knowledge graph based on a time decay factor is generated by graph convolutional network feature fusion technology. Its node update frequency is positively correlated with the feature importance index.

[0008] The personalized path planning module is based on a hierarchical planning algorithm of multi-agent reinforcement learning, combined with industry knowledge graphs to generate multi-dimensional career development paths;

[0009] The talent growth prediction module uses an adaptive weighted Transformer model combined with an industry technology iteration index to dynamically predict the probability of talent growth.

[0010] The path adjustment module dynamically adjusts career development paths by combining a hierarchical multi-objective optimization algorithm with a Nash equilibrium-based multi-objective optimization game theory conflict resolution mechanism.

[0011] The talent recommendation module is based on a bidirectional graph neural network with graph attention mechanism and a multi-round bilateral matching game model to achieve dynamic semantic matching between talents and positions.

[0012] The data storage module uses a distributed graph database combined with time-series data sharding technology to store multimodal data during system operation.

[0013] The interactive interface module provides an AR-visualized career path sandbox and a multimodal intelligent question-and-answer assistant;

[0014] The system management module integrates a regional industry talent supply and demand imbalance risk early warning function based on Kalman filtering and a dynamic permission management system based on attribute-based encryption.

[0015] Preferably, the dynamic industry profile building module includes a data acquisition submodule, a data analysis submodule, and a profile update submodule;

[0016] The data acquisition submodule uses reinforcement learning-based intelligent crawler technology, combined with a multi-agent collaboration mechanism, to capture multi-source data such as local industry news, policy documents and enterprise recruitment information in real time, and filters duplicate data through adaptive thresholds;

[0017] The data analysis submodule uses the T5-large text generation model combined with the LightGBM ensemble learning algorithm to extract key industry skills, job demand trends, and dynamic information on enterprise expansion and contraction.

[0018] The profile update submodule uses a temporal hidden Markov model with an attention mechanism to predict industry trends and dynamically adjusts the update strategy by introducing a feature importance formula with a time decay factor.

[0019] The data feature set is F = {f1, f2, ..., f...} n}, where n is the feature dimension, and the formula is as follows:

[0020]

[0021] Among them, f i Let I(f) be the i-th feature in the data feature set F, and t be the current time. i ,t) is a feature f i The importance of the decision tree at the current time t, where m is the number of decision trees, and T is the value of the decision tree at the current time t. j For the j-th decision tree, Gain(f) i ,T j ) represents the feature f in the j-th decision tree. i The information gain brought about, α is the attenuation coefficient (0 < α < 1), t j Let be the construction time of the j-th decision tree.

[0022] Preferably, the expert review and correction module includes an expert team unit and an information supplementation unit;

[0023] The expert team consists of enterprise experts with senior technical titles or more than 10 years of industry experience in the local pillar industries. They use the federated learning collaborative review platform and blockchain evidence storage technology to achieve distributed review of industry profiles.

[0024] The information supplementation unit uses a knowledge graph completion algorithm integrated with Bayesian reasoning to structurally supplement the industry profile with undisclosed project talent demand information provided by experts, and quantifies the credibility of the information through an evidence weight evaluation mechanism.

[0025] Preferably, the personalized route planning module includes a personal information analysis submodule and a route planning submodule;

[0026] The personal information analysis submodule integrates occupational assessment questionnaires, learning behavior logs, and online skills test data through multimodal behavior analysis technology. It combines a principal component analysis model optimized by transfer learning with the results of regional industrial policy text mining to achieve dynamic matching analysis between personal skills and industry trends.

[0027] The path planning submodule utilizes a hierarchical planning algorithm based on multi-agent reinforcement learning to plan short-term, medium-term, and long-term career development paths based on the results of the personal information analysis submodule. Each time interval is dynamically adjusted according to the half-life of the industry's technical knowledge update cycle, which is defined as the time required for 50% of the existing technical knowledge in a certain field to be replaced by new technologies.

[0028] Preferably, the time range for the short-term career development path is 0.5 to 2 years, the time range for the medium-term career development path is 2 to 4 years, and the time range for the long-term career development path is more than 4 years.

[0029] The path planning submodule dynamically adjusts the time interval based on the half-life of the industry's technical knowledge update cycle, which is defined as the time required for 50% of the existing technical knowledge in a certain field to be replaced by new technologies.

[0030] Shortest path: 0.5≤t s ≤min(2,τ);

[0031] Mid-term path: max(2, τ) ≤ t m ≤min(4, 2τ);

[0032] Long-term path: max(4, 2τ) ≤ t l ;

[0033] Where τ is the half-life of the industry's technological knowledge update cycle, defined as the time required for 50% of the existing technological knowledge in a certain field to be replaced by new technologies, t s t m and t l These represent the upper limits of the time intervals for short-term, medium-term, and long-term paths, respectively.

[0034] Preferably, the talent development prediction module includes a data processing unit and a prediction analysis unit;

[0035] The data processing unit collects historical talent development data and industry dynamic change data, and uses generative adversarial networks combined with adversarial training to enhance the data;

[0036] The predictive analysis unit uses an adaptive weighted Transformer model, which dynamically adjusts the weights of each sub-model through reinforcement learning. It introduces the industry technology iteration index T(t) as an external feature to construct a dynamic time-series prediction model.

[0037]

[0038] Where P(t) is the final prediction result at the current time t, k is the number of sub-models, i.e., the total number of sub-models integrated in the model, and ω i (t) represents the weight of the i-th sub-model at time t. ε is the industry volatility coefficient, P i (t) represents the prediction result of the i-th sub-model at time t.

[0039] Preferably, the path adjustment module includes a cause analysis submodule and a scheme adjustment submodule;

[0040] When talents fail to achieve their growth goals on time, the causal analysis submodule uses the causal forest algorithm combined with bias analysis and introduces a counterfactual reasoning mechanism to identify the reasons for slow skill learning progress and sudden changes in industry demand.

[0041] The scheme adjustment submodule employs a hierarchical multi-objective optimization algorithm, assigning dynamic weights γ to the short-term objective of skill mastery O1, the medium-term objective of job matching O2, and the long-term objective of industry contribution value O3. i (t), constructing a comprehensive evaluation function based on a priority decision-making model of game theory:

[0042]

[0043] Where γ is the conflict moderating factor, γ i (t) represents the dynamic weight of the i-th target at time t, C j Let be the conflict coefficient of the j-th target. And weight γ i (t) is dynamically updated using the following formula:

[0044]

[0045] Where η is the learning rate, V(π) t ) is the strategy π t The value function, Indicates O i The partial derivatives, For strategy π t Value function V(π) t For target O i gradient, γ i (t+1) represents the weight of the i-th target at time t+1, γ i (t) represents the weight of the i-th objective at time t.

[0046] Preferably, the talent recommendation module includes a matching submodule and a recommendation execution submodule;

[0047] The matching submodule uses a bidirectional graph neural network based on graph attention mechanism to simultaneously model the semantic relationship between job requirements and talent development path, and introduces a reinforcement learning-driven dynamic matching degree update mechanism to calculate the bidirectional matching probability between job and talent.

[0048] The recommendation execution submodule is based on a multi-round bilateral matching game model, which optimizes the recommendation strategy by combining enterprise feedback data, forming a closed loop of recommendation, feedback and iteration.

[0049] Preferably, the data storage module adopts a distributed graph database combined with time-series data sharding technology to partition and store multimodal data, and realizes cross-institutional data sharing and privacy protection through a federated learning data security aggregation mechanism based on homomorphic encryption.

[0050] Preferably, the interactive interface module provides an AR-visualized career path sandbox, supports multimodal interactive input methods including voice, gestures, and text, and integrates an intelligent question-and-answer assistant based on dialogue enhancement Transformer to answer career planning questions in real time;

[0051] The system management module integrates a regional industry talent supply and demand imbalance risk early warning function based on Kalman filtering, identifies industry talent supply and demand imbalance risks based on anomaly detection algorithms, and adopts a dynamic permission management system based on attribute-based encryption to realize data access control for different roles.

[0052] The present invention has the following beneficial effects:

[0053] 1. This invention introduces a temporal hidden Markov model with an attention mechanism into a dynamic industry profile construction module to predict industry trend data. It then combines this with an expert review and correction module that uses a knowledge graph completion algorithm and Bayesian inference to supplement undisclosed project information. Finally, it generates a dynamic industry knowledge graph based on a time decay factor through graph convolutional network feature fusion technology. The node update frequency is positively correlated with the feature importance index, achieving dynamic updating and accurate construction of the industry knowledge graph. Compared with existing technologies, this improves the timeliness and accuracy of industry trend prediction, thus solving the problem of lagging industry information in existing talent recommendation systems and their inability to reflect local industry dynamics in real time.

[0054] 2. This invention employs a hierarchical planning algorithm based on multi-agent reinforcement learning in its personalized path planning module. Combined with an industry knowledge graph, it generates multi-dimensional career development paths and dynamically adjusts the time intervals of short-term, medium-term, and long-term paths according to the half-life of the industry's technical knowledge update cycle, achieving dynamic adaptive planning of career development paths. Compared with existing technologies, this improves the adaptability of career path planning to industry technology iterations, thus solving the problem that traditional path planning models cannot flexibly adjust to the speed of technology updates, leading to a disconnect between talent skills and industry needs.

[0055] 3. This invention models the semantic relationship between job requirements and talent development paths using a bidirectional graph neural network based on a graph attention mechanism in its talent recommendation module. It then combines this with a multi-round bilateral matching game model to achieve dynamic semantic matching. Furthermore, it optimizes the recommendation strategy using enterprise feedback data, forming a closed loop and achieving accurate and dynamic matching between talent and jobs. Compared to existing technologies, this invention improves the semantic understanding accuracy and matching efficiency of talent recommendation, thus solving the problem that traditional recommendation algorithms rely solely on keyword matching and cannot deeply understand job skill requirements and talent development potential. Attached Figure Description

[0056] Figure 1 This is a schematic diagram of the overall system architecture of the present invention;

[0057] Figure 2 This is a schematic diagram of the architecture of the dynamic industry profile building module of the present invention;

[0058] Figure 3 This is a schematic diagram of the architecture of the expert review and correction module of the present invention;

[0059] Figure 4 This is a schematic diagram of the architecture of the personalized path planning module of the present invention;

[0060] Figure 5 This is a schematic diagram of the architecture of the talent growth prediction module of the present invention;

[0061] Figure 6 This is a schematic diagram of the architecture of the path adjustment module of the present invention;

[0062] Figure 7 This is a schematic diagram of the architecture of the talent recommendation module of the present invention. Detailed Implementation

[0063] In order to make the technical means, creative features, objectives and effects achieved by the present invention easier to understand, the present invention is further described below in conjunction with specific implementation methods.

[0064] Example 1, please refer to Figures 1 to 2 As shown: A talent recommendation system based on matching vocational education with local productivity. The talent recommendation system based on matching vocational education with local productivity includes a dynamic industry profile construction module, an expert review and correction module, a personalized path planning module, a talent growth prediction module, a path adjustment module, a talent recommendation module, a data storage module, an interactive interface module, and a system management module.

[0065] The dynamic industry profile building module predicts industry trend data by introducing a temporal hidden Markov model with an attention mechanism. The expert review and correction module supplements unpublished project information by incorporating Bayesian inference based on a knowledge graph completion algorithm. The dynamic industry knowledge graph based on a time decay factor is generated through graph convolutional network feature fusion technology. Its node update frequency is positively correlated with the feature importance index.

[0066] The personalized path planning module is based on a hierarchical planning algorithm using multi-agent reinforcement learning, and combines industry knowledge graphs to generate multi-dimensional career development paths;

[0067] The talent development prediction module uses an adaptive weighted Transformer model combined with an industry technology iteration index to dynamically predict the probability of talent development.

[0068] The path adjustment module dynamically adjusts career development paths by combining a hierarchical multi-objective optimization algorithm with a Nash equilibrium-based multi-objective optimization game theory conflict resolution mechanism.

[0069] The talent recommendation module is based on a bidirectional graph neural network with graph attention mechanism and a multi-round bilateral matching game model to achieve dynamic semantic matching between talents and positions;

[0070] The data storage module uses a distributed graph database combined with time-series data sharding technology to store multimodal data during system operation;

[0071] The interactive interface module provides an AR-visualized career path sandbox and a multimodal intelligent question-and-answer assistant;

[0072] The system management module integrates a regional industry talent supply and demand imbalance risk early warning function based on Kalman filtering and a dynamic permission management system based on attribute-based encryption.

[0073] The dynamic industry profile building module includes a data acquisition submodule, a data analysis submodule, and a profile update submodule;

[0074] The data acquisition submodule uses reinforcement learning-based intelligent crawler technology, combined with a multi-agent collaboration mechanism, to capture multi-source data such as local industry news, policy documents and enterprise recruitment information in real time, and filters duplicate data through adaptive thresholds;

[0075] The data analysis submodule uses the T5-large text generation model combined with the LightGBM ensemble learning algorithm to extract key industry skills, job demand trends, and dynamic information on enterprise expansion and contraction.

[0076] The profile update submodule uses a temporal hidden Markov model with an attention mechanism to predict industry trends and dynamically adjusts the update strategy by introducing a feature importance formula with a time decay factor.

[0077] The data feature set is F = {f1, f2, ..., f...} n}, where n is the feature dimension, and the formula is as follows:

[0078]

[0079] Among them, f i Let I(f) be the i-th feature in the data feature set F, and t be the current time. i ,t) is a feature f i The importance of the decision tree at the current time t, where m is the number of decision trees, and T is the value of the decision tree at the current time t. j For the j-th decision tree, Gain(f) i ,T j ) represents the feature f in the j-th decision tree. i The information gain brought about, α is the attenuation coefficient (0 < α < 1), t j Let be the construction time of the j-th decision tree.

[0080] Furthermore, the talent recommendation system based on matching vocational education with local productivity encompasses nine core modules: dynamic industry profile construction, expert review and correction, personalized path planning, talent growth prediction, path adjustment, talent recommendation, data storage, interactive interface, and system management. These modules work collaboratively, utilizing cutting-edge algorithms and technologies to achieve a full-process function from industry trend analysis and talent development planning to precise job recommendations, thus adapting to the local productivity's demand for vocational education talent.

[0081] The data acquisition submodule employs reinforcement learning-based intelligent web crawling technology. This technology constructs a deep Q-network as the core decision-making model. During the data acquisition process, the intelligent crawler takes the state of each data capture as input, including the type of data already collected and the website access status. Through the DQN network, it outputs actions to select the next data source or page to crawl. The intelligent crawler receives reward feedback based on the collection results. Successfully capturing valid data gives a positive reward, while capturing duplicate or invalid data gives a negative reward. Through continuous trial, feedback, and learning, the data acquisition strategy is optimized.

[0082] Combining a multi-agent collaboration mechanism, multiple data collection agents are deployed on different network nodes. Each agent has an independent DQN model. They work in parallel to capture local industry news, policy documents, and corporate recruitment information in real time. The collected data is dynamically filtered using indicators such as cosine similarity to adaptive thresholds. Data with similarity higher than the threshold is identified as duplicate data and removed to ensure the uniqueness and validity of the collected data.

[0083] The data analysis submodule innovatively combines the T5-large text generation model with the LightGBM ensemble learning algorithm. The T5-large text generation model is based on the Transformer architecture and has powerful semantic understanding capabilities. When processing the collected text data, the T5-large model first segments and encodes the text, and captures the semantic information in the text through a multi-layer attention mechanism, thereby extracting information such as key industry skills or job requirements contained in the text.

[0084] The LightGBM ensemble learning algorithm employs a gradient boosting framework. It builds a feature selection model by iteratively training multiple weak learners and uses optimization techniques such as histogram algorithms to filter and enhance the features extracted by the T5-large model, thereby improving the accuracy and stability of feature extraction. Through the synergistic work of these two techniques, it can accurately extract key industry skill requirements, job demand trends, and dynamic information on enterprise expansion or contraction, providing crucial data support for subsequent industry profiling.

[0085] The profile update submodule uses a temporal hidden Markov model with an attention mechanism to predict industry trends. Traditional temporal hidden Markov models struggle to distinguish the degree of influence of different data on industry trends when processing large amounts of data. However, the attention mechanism introduced in this system calculates the attention weight of data at each time step, focusing on key data and time nodes that have a significant impact on industry trends. Specifically, during model training, the correlation between data and changes in industry trends is calculated, and different attention weights are assigned to each data point. Data with higher weights have a greater proportion in the model's prediction, thereby improving the accuracy of the prediction.

[0086] Formula for the characteristic importance of time decay factor The update strategy for industry profiles is dynamically adjusted. As time goes by, the importance of features will change based on factors such as the decay coefficient α and the information gain of the decision tree. When the importance index of a certain feature reaches the system's preset threshold or a manually set threshold, the profile update submodule will prioritize updating the industry profile content involving that feature to ensure that the industry profile can reflect the latest developments in the industry in a timely manner.

[0087] Through the above technical solutions, in actual application testing, the system's prediction delay for industry trends has been reduced from 72 hours to 20 hours, enabling it to identify more than 85% of emerging job demands in advance. The completeness of industry profile information has increased from 65% to 92%, providing more comprehensive and accurate industry information support for talent recommendation.

[0088] Example 2, please refer to Figure 3 As shown: Based on Embodiment 1, the expert review and correction module includes an expert team unit and an information supplementation unit;

[0089] The expert team consists of enterprise experts with senior technical titles or more than 10 years of industry experience in the local pillar industries. They use the federated learning collaborative review platform and blockchain evidence storage technology to achieve distributed review of industry profiles.

[0090] The information supplementation unit uses a knowledge graph completion algorithm integrated with Bayesian reasoning to structurally supplement the industry profile with undisclosed project talent demand information provided by experts, and quantifies the credibility of the information through an evidence weight evaluation mechanism.

[0091] Furthermore, the expert team conducted their work through the federated learning collaborative review platform. The core advantage of the federated learning mechanism lies in its ability to achieve collaborative review of industry profiles across enterprises and institutions without disclosing the original data.

[0092] Specifically, each expert retains the original data locally and only uploads encrypted model parameter updates. The platform aggregates these parameters to iteratively optimize the industry profile model, avoiding the risk of data leakage and protecting corporate trade secrets. In addition, based on blockchain evidence storage technology, every operation during the review process, including data uploads, model parameter updates, and review comments, is recorded on the blockchain in the form of hash values. This evidence storage method makes the review process tamper-proof; once recorded, it cannot be modified and is also traceable. The entire review process can be accessed at any time, greatly enhancing the credibility and authority of the industry profile.

[0093] The information supplementation unit employs a knowledge graph completion algorithm integrated with Bayesian inference. The knowledge graph completion algorithm utilizes the relationships between nodes in a pre-constructed industry knowledge graph, learns the vector representations of nodes and relationships through a deep learning model, and then infers the missing information. Taking the prediction of talent demand in an emerging industry as an example, the algorithm can predict the skill combinations for positions that are not yet clearly defined but actually exist in demand, based on the existing relationships between enterprises and positions, and between positions and skills. Bayesian inference, on the other hand, dynamically updates the confidence level of the information based on prior knowledge and newly acquired evidence.

[0094] When experts provide undisclosed project talent requirements, Bayesian inference uses information from similar past projects as prior knowledge to evaluate the new information and update its credibility. To further ensure the reliability of the supplementary information, an evidence weighting assessment mechanism is introduced. This mechanism quantifies the assessment from dimensions such as expert authority, consistency between information and the knowledge graph, and the credibility of the information source. Expert authority includes professional title, experience, and domain relevance, while the credibility of the information source includes enterprise level and data timeliness. Each dimension is assigned a different weight, and the final credibility score is calculated through weighted calculation. Only information that reaches a certain credibility threshold will be structured and added to the industry profile, thereby achieving accurate improvement of the industry profile content. The credibility threshold can be automatically calculated by the system based on historical data or dynamically adjusted by the expert committee to ensure the overall consistency between the supplementary information and the industry profile.

[0095] Example 3, please refer to Figure 4 As shown: Based on Embodiment 1 and Embodiment 2, the personalized route planning module includes a personal information analysis submodule and a route planning submodule;

[0096] The personal information analysis submodule integrates occupational assessment questionnaires, learning behavior logs, and online skills test data through multimodal behavior analysis technology. It combines a principal component analysis model optimized by transfer learning with the results of regional industrial policy text mining to achieve dynamic matching analysis between personal skills and industry trends.

[0097] The path planning submodule utilizes a hierarchical planning algorithm based on multi-agent reinforcement learning to plan short-term, medium-term, and long-term career development paths based on the results of the personal information analysis submodule. Each time interval is dynamically adjusted according to the half-life of the industry's technical knowledge update cycle, which is defined as the time required for 50% of the existing technical knowledge in a certain field to be replaced by new technologies.

[0098] The time frame for short-term career development paths is 0.5 to 2 years, for medium-term career development paths it is 2 to 4 years, and for long-term career development paths it is 4 years or more.

[0099] The path planning submodule dynamically adjusts the time interval based on the half-life of the industry's technical knowledge update cycle, which is defined as the time required for 50% of the existing technical knowledge in a certain field to be replaced by new technologies.

[0100] Shortest path: 0.5≤t s ≤min(2,τ);

[0101] Mid-term path: max(2, τ) ≤ t m ≤min(4, 2τ);

[0102] Long-term path: max(4, 2τ) ≤ tl ;

[0103] Where τ is the half-life of the industry's technological knowledge update cycle, defined as the time required for 50% of the existing technological knowledge in a certain field to be replaced by new technologies, t s t m and t l These represent the upper limits of the time intervals for short-term, medium-term, and long-term paths, respectively.

[0104] Furthermore, the personalized path planning module is based on a hierarchical planning algorithm using multi-agent reinforcement learning, combined with industry knowledge graphs to generate multi-dimensional career development paths. The multi-agent reinforcement learning algorithm employs a centralized training-distributed execution architecture, with three agents responsible for short-term, medium-term, and long-term path planning respectively. Each agent consists of a policy network with three fully connected layers and a value network with two fully connected layers, with nodes connected in a fully connected manner. The hierarchical planning algorithm decomposes the career development path planning problem into strategic, tactical, and operational layers, achieving path optimization through inter-layer information interaction.

[0105] The personal information analysis submodule utilizes multimodal behavioral analysis technology to break down data barriers and deeply integrate career assessment questionnaires, learning behavior logs, and online skills test data. These data come from a wide range of sources and are diverse in type, covering subjective self-evaluations, objective learning behavior records, and actual skills level tests, providing a rich data foundation for a comprehensive understanding of individual abilities.

[0106] At the data processing level, a principal component analysis (PCA) model optimized through transfer learning is introduced. Traditional PCA models, when processing high-dimensional data, easily suffer from increased computational complexity due to the large number of data features and struggle to effectively extract key features. This system, however, utilizes transfer learning to borrow model parameters already trained in related fields and apply them to the current data processing task, significantly improving the model's adaptability to new data. Based on this, dimensionality reduction processing is performed, enabling the rapid and accurate extraction of features crucial for career development path planning from massive datasets. For example, it can extract personal career interest characteristics from career assessment questionnaires, mine learning efficiency and knowledge acquisition progress from learning behavior logs, and obtain professional skill level characteristics from online skills test data.

[0107] Furthermore, this submodule innovatively incorporates the results of regional industrial policy text mining. Through Natural Language Processing (NLP) technology, it performs in-depth analysis of local industrial policy documents, uncovering key information such as industrial development directions, key support areas, and skills demand trends. Dynamically matching this industrial policy information with individual skills data allows for real-time reflection of the alignment between individual skills and industry trends. For example, when local industrial policies strongly support the development of the new energy vehicle industry, the system analyzes an individual's skills in related fields to determine if they meet industry development needs, providing a precise and real-time personal information foundation for subsequent career path planning.

[0108] The path planning submodule utilizes a hierarchical planning algorithm based on multi-agent reinforcement learning to plan short-term, medium-term, and long-term career development paths based on the results of the personal information analysis submodule. In the multi-agent reinforcement learning part, each agent's policy network contains three fully connected layers with 256, 128, and 64 neurons respectively, using ReLU activation functions; the value network contains two fully connected layers with 128 and 64 neurons respectively, using linear activation functions. Agents communicate asynchronously through message queues, collaboratively generating career development paths based on node information in the industry knowledge graph. The hierarchical planning algorithm divides path planning into a strategic layer, a tactical layer, and an operational layer. The strategic layer determines the career development direction, the tactical layer sets phased goals, and the operational layer plans specific learning and practice tasks. Information is transmitted between layers through a state transition matrix to achieve dynamic path optimization.

[0109] The specific dynamic adjustment logic is as follows:

[0110] Short-term career development path: Time interval set at 0.5 ≤ t s ≤min(2,τ), when the industry's technology updates are slow and the half-life τ is greater than 2 years, the upper limit of the short-term path is 2 years, ensuring that talents can master basic skills and adapt to the current needs of the industry in the short term; if τ is less than 2 years, the upper limit is τ, so that the short-term path closely follows the pace of rapid changes in the industry and adjusts the direction of talent development in a timely manner.

[0111] Mid-term career development path: Time interval is max(2, τ) ≤ t m The setting of ≤min(4, 2τ) allows talents sufficient time to deepen their skills and advance their careers when the industry's technology is developing steadily. When the industry's technology updates rapidly and τ is small, the lower and upper limits of the mid-term path will be dynamically adjusted according to τ to ensure that talents always keep pace with the industry's technological progress during the mid-term development stage.

[0112] Long-term career development path: The time interval is max(4, 2τ)≤t lThe long-term path setting fully considers the long-term development trend of industry technology. When the industry technology update cycle is long, it provides stable long-term development plans for talents. If the industry technology updates rapidly, the lower limit of the long-term path will be raised accordingly to ensure that talents have foresight in long-term development and can continuously adapt to the challenges of new technologies.

[0113] By using this dynamic time interval adjustment mechanism based on the half-life of the industry's technical knowledge update cycle, the path planning submodule enables career development path planning to be highly adaptable to industry technology iterations. This effectively avoids the problem of talent skills being out of sync with industry needs in traditional path planning models, providing scientific, reasonable, and forward-looking career development guidance for vocational education talents.

[0114] Practical application verification shows that users who use this system for career path planning have a job matching success rate that is 31 percentage points higher than that of traditional methods, reaching 89%. With the rapid iteration of industry technology, the average adjustment cycle of users' career development paths has been shortened from 6 months to 1.5 months, effectively improving the matching efficiency between talent and industry needs.

[0115] Example 4, please refer to Figure 5 As shown: Based on Embodiment 1, Embodiment 2 and Embodiment 3, the talent growth prediction module includes a data processing unit and a prediction analysis unit;

[0116] The data processing unit collects historical talent development data and industry dynamic change data, and uses generative adversarial networks combined with adversarial training to enhance the data;

[0117] The predictive analysis unit uses an adaptive weighted Transformer model, which dynamically adjusts the weights of each sub-model through reinforcement learning. It introduces the industry technology iteration index T(t) as an external feature to construct a dynamic time-series prediction model.

[0118]

[0119] Where P(t) is the final prediction result at the current time t, k is the number of sub-models, i.e., the total number of sub-models integrated in the model, and ω i (t) represents the weight of the i-th sub-model at time t. ε is the industry volatility coefficient, P i (t) represents the prediction result of the i-th sub-model at time t.

[0120] Furthermore, the data processing unit collects historical data on talent development and industry trends, covering multiple dimensions such as education, skills, market demand, and technological innovation, providing a foundation for prediction.

[0121] To enhance the diversity and richness of the data, the data processing unit introduces data augmentation techniques using generative adversarial networks (GANs) combined with adversarial training. A GAN consists of a generator and a discriminator. The generator's role is to generate new data with a distribution similar to the original data based on the input random noise. This new data is similar to the real data in features and structure, but also possesses certain differences, thereby expanding the number of data samples. For example, when processing work experience data, the generator can generate simulated work experience data from different industries and positions, enriching the data's diversity.

[0122] The discriminator is responsible for determining whether the input data is real data or fake data generated by the generator. During adversarial training, the generator continuously optimizes itself, generating more realistic data to deceive the discriminator, while the discriminator also continuously learns and improves its ability to distinguish between real and fake data. This adversarial game between the generator and the discriminator drives the continuous improvement of the quality of the data generated by the generator. The resulting augmented data not only increases in quantity but also becomes closer in quality to real data, providing more sufficient and representative data support for subsequent predictive analysis units and effectively alleviating the problem of insufficient model training caused by insufficient data.

[0123] The predictive analysis unit employs an adaptive weighted Transformer model, optimized based on the Transformer architecture. It dynamically adjusts the weights of sub-models through reinforcement learning to achieve accurate predictions of talent development probabilities. One of its key innovations is the model's ability to dynamically adjust the weights of each sub-model through reinforcement learning. Traditional model fusion methods typically combine sub-models with fixed weights, which struggles to adapt to complex and ever-changing data characteristics and prediction needs. In contrast, the adaptive weighted Transformer model in this system utilizes reinforcement learning algorithms, using the model's prediction results as feedback signals to construct a reward mechanism. For example, when the model's prediction results match the actual talent development situation, a positive reward is given; if the prediction deviation is significant, negative feedback is provided. Through continuous trial and error and learning, the model can automatically optimize the contribution of each sub-model based on data characteristics and prediction needs, enabling each sub-model to perform optimally in different prediction scenarios. This dynamic weight adjustment mechanism greatly improves the model's flexibility and adaptability, allowing it to more accurately capture the complex patterns in the talent development process.

[0124] By introducing the industry technology iteration index T(t) as an external feature, a dynamic time series prediction model is constructed. The industry technology iteration index T(t) comprehensively considers various factors such as the speed of new technology research and development, the rate of technology application and popularization, and the frequency of technology updates within the industry, and can quantitatively reflect the iteration speed and development trend of industry technology.

[0125] In the model, P(t) represents the final prediction result at the current time t, k is the number of sub-models, i.e., the total number of sub-models integrated in the model, and ω i (t) represents the weight of the i-th sub-model at time t. ε is the industry volatility coefficient, used to adjust the influence of the industry technology iteration index on the forecast results. i (t) represents the prediction result of the i-th sub-model at time t.

[0126] By incorporating the industry technology iteration index into the model, the prediction results can reflect the impact of industry technology changes on talent development in real time. For example, when the industry technology iteration index T(t) rises rapidly, it means that the industry's technology updates are accelerating. The model will adjust the prediction results accordingly, tending to predict that talents need to accelerate their skill updates to adapt to industry development. This provides talents with more forward-looking growth suggestions and also provides vocational education institutions and enterprises with more accurate reference for talent training and recruitment decisions.

[0127] In summary, the talent development prediction module, through the innovative design and collaborative operation of the data processing unit and the predictive analysis unit, achieves dynamic and accurate prediction of talent development trends. It further improves the talent recommendation system based on matching vocational education with local productivity and effectively solves the problems of insufficient data, poor model adaptability, and inability to fully consider dynamic changes in the industry in traditional talent prediction methods.

[0128] Example 5, please refer to Figure 6 As shown: Based on Embodiment 1, Embodiment 2, Embodiment 3, and Embodiment 4, the path adjustment module includes a cause analysis submodule and a scheme adjustment submodule;

[0129] When talents fail to achieve their growth goals on time, the causal analysis submodule uses the causal forest algorithm combined with bias analysis and introduces a counterfactual reasoning mechanism to identify the reasons for slow skill learning progress and sudden changes in industry demand.

[0130] The scheme adjustment submodule employs a hierarchical multi-objective optimization algorithm, assigning dynamic weights γ to the short-term objective of skill mastery O1, the medium-term objective of job matching O2, and the long-term objective of industry contribution value O3. i (t), constructing a comprehensive evaluation function based on a priority decision-making model of game theory:

[0131]

[0132] Where γ is the conflict moderating factor, γ i (t) represents the dynamic weight of the i-th target at time t, C j Let be the conflict coefficient of the j-th target. And weight γ i(t) is dynamically updated using the following formula:

[0133]

[0134] Where η is the learning rate, V(π) t ) is the strategy π t The value function, Indicates O i The partial derivatives, For strategy π t Value function V(π) t For target O i gradient, γ i (t+1) represents the weight of the i-th target at time t+1, γ i (t) represents the weight of the i-th objective at time t.

[0135] Furthermore, when talent fails to achieve their growth goals on time, the causal analysis submodule initiates a deep diagnostic mechanism integrating multiple technologies. This submodule utilizes a causal forest algorithm to quickly locate key factors affecting talent development within complex data. For example, when analyzing the slow improvement of programming skills in vocational education talent, the causal forest algorithm can quantify the causal effects of various potential factors on skill improvement, such as learning time allocation, course difficulty suitability, and participation in practical projects, thus identifying the true influencing factors.

[0136] By combining deviation analysis techniques, the actual development trajectory of talent is systematically compared with the preset growth goals. Through time series analysis, not only can the absolute deviation values ​​of indicators such as skill mastery progress and job suitability be identified, but the changing trends of these deviations can also be uncovered, such as inflection points where the rate of skill growth slows down. For example, if it is found that a talent's job suitability is consistently lower than expected during the mid-stage of their career development, deviation analysis can pinpoint the gap in specific skill areas, such as insufficient mastery of emerging data analysis tools.

[0137] This innovative approach introduces a counterfactual reasoning mechanism to construct virtual scenarios and deduce causal relationships. By assuming different conditions and interventions, this mechanism simulates the development outcomes of talent under various possible scenarios, thereby inferring the specific reasons why goals were not achieved. For example, when industry demands change abruptly due to the application of new technologies, counterfactual reasoning can simulate whether prior exposure to new technology training could prevent a decline in job suitability. This helps determine the causal link between the sudden change in industry demands and the lag in talent skills, providing a precise diagnostic basis for subsequent solution adjustments.

[0138] The solution adjustment submodule, guided by the diagnostic results of the cause analysis submodule, employs a hierarchical multi-objective optimization algorithm to construct a dynamic weight adjustment and priority decision-making system. Dynamic weights γ are assigned to the short-term goal of skill mastery O1, the mid-term goal of job matching O2, and the long-term goal of industry contribution value O3. i (t), which breaks through the limitations of traditional fixed-weight programming.

[0139] Weight dynamic update formula This achieves adaptive weight adjustment, where η is the learning rate, controlling the step size of weight adjustment to avoid over- or under-adjustment, and V(π) t ) is the strategy π t The value function quantifies the strengths and weaknesses of a career development strategy by evaluating its overall benefits across various objectives. Indicates O i The partial derivatives, For strategy π t Value function V(π) t For target O i The gradient of γ indicates the direction and magnitude of the improvement in policy value resulting from adjusting the weights of each objective. i (t+1) represents the weight of the i-th target at time t+1, γ i (t) represents the weight of the i-th objective at time t. For example, when a sudden change in industry demand causes a sharp drop in job matching, this formula can quickly increase the weight of the mid-term objective job matching O2, guiding the system to prioritize optimizing job matching solutions.

[0140] Constructing a comprehensive evaluation function based on a priority decision-making model of game theory Balancing conflict and cooperation among multiple objectives, where γ is a conflict moderating factor. i (t) represents the dynamic weight of the i-th target at time t, C j Let be the conflict coefficient of the j-th target.

[0141] The conflict mitigation factor γ can be dynamically adjusted according to the degree of conflict between objectives, avoiding optimization failure caused by contradictions between objectives. For example, when there is a resource allocation conflict between short-term skill learning investment and long-term industry contribution development, γ will reduce the weight of the conflicting objective and reallocate resources to improve overall returns. By continuously iterating and updating the weights, this module can optimize career development paths in real time, such as adjusting the order of course learning, recommending targeted training programs, or planning cross-domain skill expansion programs, ensuring that talent development always aligns with dynamic industry needs and individual growth potential.

[0142] The path adjustment module forms a closed-loop feedback mechanism through the precise diagnosis of the cause analysis submodule and the dynamic optimization of the solution adjustment submodule. This effectively solves the problems of rigid paths and difficulty in coping with complex changes in traditional career planning, provides vocational education talents with dynamic and adaptive development strategies, and enhances the matching accuracy between the talent recommendation system and local productivity needs.

[0143] Example 6, please refer to Figure 7 As shown: Based on Embodiment 1, Embodiment 2, Embodiment 3, Embodiment 4, and Embodiment 5, the talent recommendation module includes a matching submodule and a recommendation execution submodule;

[0144] The matching submodule uses a bidirectional graph neural network based on graph attention mechanism to simultaneously model the semantic relationship between job requirements and talent development path, and introduces a reinforcement learning-driven dynamic matching degree update mechanism to calculate the bidirectional matching probability between job and talent.

[0145] The recommendation execution submodule is based on a multi-round bilateral matching game model, which optimizes the recommendation strategy by combining enterprise feedback data, forming a closed loop of recommendation, feedback and iteration.

[0146] Furthermore, the talent recommendation module utilizes a bidirectional graph neural network based on graph attention mechanisms and a multi-round bilateral matching game model to achieve dynamic semantic matching between talent and job positions. The matching submodule employs a bidirectional graph neural network based on graph attention mechanisms to simultaneously model the semantic association between job requirements and talent development paths. During network training, the Adam optimizer is used, with an initial learning rate of 0.001 and a batch size of 32. Backpropagation is used to update network parameters, achieving deep modeling of the semantic association between job positions and talent.

[0147] The matching submodule employs a bidirectional graph neural network based on a graph attention mechanism to model the semantic relationship between job requirements and talent development paths. In constructing the graph structure, the various requirements of a job and the abilities of a candidate are abstracted as nodes in the graph, while the relationships between nodes are represented as edges. For example, job requirements include professional skills, work experience, and certifications, while the abilities of a candidate include learning experience, project experience, and skill mastery. The relationships between nodes represent the specific requirements of a job corresponding to a particular skill.

[0148] Graph attention mechanisms play a crucial role here. They automatically assign different attention weights to nodes and edges based on their importance within the graph structure, focusing on nodes and edges that are key to semantic association modeling. For example, for a software development role, during the modeling process, the graph attention mechanism assigns higher attention weights to nodes related to programming language skills and project development experience, thereby accurately extracting key features. Simultaneously, the matching strategy is dynamically updated through a Q-learning algorithm, adjusting the matching probability in real time based on changes in job requirements and talent development, and conversely, imposing negative penalties. This dynamically calculates the bidirectional matching probability between job requirements and talent, achieving precise matching analysis between talent and job.

[0149] The recommendation execution submodule operates based on a multi-round bilateral matching game model. In this model, enterprises and talents are considered as the two sides in a game, each with its own interests and decision-making strategies. In the initial stage of the recommendation process, the system generates preliminary matching schemes based on the bidirectional matching probabilities calculated by the matching submodule. Enterprises and talents evaluate and decide on these schemes based on their own needs, and if dissatisfied, they request adjustments. The system simulates the continuous game process between the two sides through multiple rounds of matching attempts and strategy adjustments. For example, after the first round of matching, if the enterprise believes that a talent lacks a certain skill, the system will use this feedback to select talents with the corresponding skills in subsequent matches. If the talent feels that the salary and benefits of the position do not meet their expectations, the system will also readjust the recommendation scheme.

[0150] Simultaneously, the recommendation strategy is optimized by combining enterprise feedback data. After a candidate joins the company, the company will provide feedback on the candidate's actual performance data in the position, such as work efficiency, skill application, and teamwork ability. The system uses this feedback data as an important basis and uses supervised learning algorithms in machine learning, such as support vector machines, to train and optimize the recommendation strategy. Through continuous iteration, a closed loop of recommendation, feedback, and iteration is formed, which significantly improves the accuracy and effectiveness of talent recommendation and realizes dynamic semantic matching between talents and positions.

[0151] Example 7, please refer to Figure 1 As shown: Based on Embodiment 1, Embodiment 2, Embodiment 3, Embodiment 4, Embodiment 5 and Embodiment 6, the data storage module adopts a distributed graph database combined with time-series data sharding technology to partition and store multimodal data, and realizes cross-institutional data sharing and privacy protection through a federated learning data security aggregation mechanism based on homomorphic encryption.

[0152] The interactive interface module provides an AR-visualized career path sandbox, supports multimodal interactive input methods such as voice, gestures and text, and integrates an intelligent question-and-answer assistant based on dialogue enhancement Transformer to answer career planning questions in real time;

[0153] The system management module integrates a regional industry talent supply and demand imbalance risk early warning function based on Kalman filtering, identifies industry talent supply and demand imbalance risks based on anomaly detection algorithms, and adopts a dynamic permission management system based on attribute-based encryption to realize data access control for different roles.

[0154] Furthermore, the data storage module employs a distributed graph database combined with time-series data sharding technology to partition and store multimodal data generated during system operation. The distributed graph database can efficiently store and manage data with complex relationships, while the time-series data sharding technology divides the data into shards based on its time attributes, improving data storage and retrieval efficiency. Through a federated learning data security aggregation mechanism based on homomorphic encryption, cross-institutional data sharing is achieved while ensuring data privacy and security, providing rich data support for various modules of the system, and ensuring the security and privacy of data during the sharing process.

[0155] The interactive interface module provides an AR-visualized career path sandbox. Users can interact with the sandbox through multimodal input methods such as voice, gestures, and text. The sandbox displays career development paths in a 3D visualization, allowing users to more intuitively understand their career development plans. At the same time, this module integrates an intelligent question-and-answer assistant based on dialogue enhancement Transformer. The intelligent question-and-answer assistant can understand the user's question intent, combine system knowledge and data, and answer the user's questions about path planning in real time, providing personalized consultation services and improving the user experience.

[0156] The system management module integrates a regional industry talent supply and demand imbalance risk early warning function based on Kalman filtering and a dynamic access control system based on attribute-based encryption. The Kalman filtering-based risk early warning function monitors and analyzes regional industry talent supply and demand data in real time, uses the Kalman filtering algorithm to predict and correct the data, promptly identifies industry talent supply and demand imbalance risks, and issues early warning signals. The attribute-based encryption-based dynamic access control system dynamically allocates and manages data access permissions according to user roles and attributes, ensuring that users with different roles can only access data within their authorized scope, thus guaranteeing the security and confidentiality of system data.

[0157] In this invention, firstly, the dynamic industry profile construction module of the talent recommendation system based on vocational education matching local productivity collects multi-source data of local industries through intelligent crawlers. After the data analysis submodule extracts key information, the profile update submodule combines the time series hidden Markov model and time decay factor to predict industry trends and update the profile. The expert team in the expert review and correction module reviews the profile, and the information supplementation unit supplements unpublished information to improve the industry knowledge graph.

[0158] Next, the personalized path planning module integrates personal multimodal data with industry trends and uses a hierarchical planning algorithm to generate short, medium and long-term career development paths. The talent growth prediction module processes and enhances historical and dynamic data, and then predicts the probability of talent growth through an adaptive weighted Transformer model.

[0159] When talent fails to meet the target, the path adjustment module's cause analysis submodule identifies the reasons, the solution adjustment submodule adjusts the path accordingly, the talent recommendation module's matching submodule calculates the match probability between the job and the talent, and the recommendation execution submodule combines enterprise feedback to optimize the strategy and complete the recommendation.

[0160] Finally, the data storage module uses a distributed graph database and time-series data sharding technology to store multimodal data and ensures the security of cross-institutional data sharing. The interactive interface module provides AR sandbox and intelligent question answering to enhance user experience. The system management module realizes talent supply and demand risk warning and dynamic permission management.

[0161] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A talent recommendation system based on matching vocational education with local productivity, characterized in that: The talent recommendation system based on matching vocational education with local productivity includes a dynamic industry profile construction module, an expert review and correction module, a personalized path planning module, a talent growth prediction module, a path adjustment module, a talent recommendation module, a data storage module, an interactive interface module, and a system management module. The dynamic industry profile building module predicts industry trend data by introducing a temporal hidden Markov model with an attention mechanism. The expert review and correction module supplements unpublished project information by incorporating Bayesian inference based on a knowledge graph completion algorithm. The dynamic industry knowledge graph based on a time decay factor is generated by graph convolutional network feature fusion technology. Its node update frequency is positively correlated with the feature importance index. The personalized path planning module is based on a hierarchical planning algorithm of multi-agent reinforcement learning, combined with industry knowledge graphs to generate multi-dimensional career development paths; The talent growth prediction module uses an adaptive weighted Transformer model combined with an industry technology iteration index to dynamically predict the probability of talent growth. The path adjustment module dynamically adjusts career development paths by combining a hierarchical multi-objective optimization algorithm with a Nash equilibrium-based multi-objective optimization game theory conflict resolution mechanism. The talent recommendation module is based on a bidirectional graph neural network with graph attention mechanism and a multi-round bilateral matching game model to achieve dynamic semantic matching between talents and positions. The data storage module uses a distributed graph database combined with time-series data sharding technology to store multimodal data during system operation. The interactive interface module provides an AR-visualized career path sandbox and a multimodal intelligent question-and-answer assistant; The system management module integrates a regional industry talent supply and demand imbalance risk early warning function based on Kalman filtering and a dynamic permission management system based on attribute-based encryption.

2. The talent recommendation system based on matching vocational education with local productivity as described in claim 1, characterized in that: The dynamic industry profile building module includes a data acquisition submodule, a data analysis submodule, and a profile update submodule. The data acquisition submodule uses reinforcement learning-based intelligent crawler technology, combined with a multi-agent collaboration mechanism, to capture multi-source data such as local industry news, policy documents and enterprise recruitment information in real time, and filters duplicate data through adaptive thresholds; The data analysis submodule uses the T5-large text generation model combined with the LightGBM ensemble learning algorithm to extract key industry skills, job demand trends, and dynamic information on enterprise expansion and contraction. The profile update submodule uses a temporal hidden Markov model with an attention mechanism to predict industry trends and dynamically adjusts the update strategy by introducing a feature importance formula with a time decay factor. The data feature set is F = {f1, f2, ..., f...} n }, where n is the feature dimension, and the formula is as follows: Among them, f i Let I(f) be the i-th feature in the data feature set F, and t be the current time. i ,t) is a feature f i The importance of the decision tree at the current time t, where m is the number of decision trees, and T is the value of the decision tree at the current time t. j For the j-th decision tree, Gain(f) i ,T j ) represents the feature f in the j-th decision tree. i The information gain brought about, α is the attenuation coefficient (0 < α < 1), t j Let be the construction time of the j-th decision tree.

3. The talent recommendation system based on matching vocational education with local productivity as described in claim 1, characterized in that: The expert review and correction module includes an expert team unit and an information supplementation unit; The expert team consists of enterprise experts with senior technical titles or more than 10 years of industry experience in the local pillar industries. They use the federated learning collaborative review platform and blockchain evidence storage technology to achieve distributed review of industry profiles. The information supplementation unit uses a knowledge graph completion algorithm integrated with Bayesian reasoning to structurally supplement the industry profile with undisclosed project talent demand information provided by experts, and quantifies the credibility of the information through an evidence weight evaluation mechanism.

4. The talent recommendation system based on matching vocational education with local productivity as described in claim 1, characterized in that: The personalized route planning module includes a personal information analysis submodule and a route planning submodule; The personal information analysis submodule integrates occupational assessment questionnaires, learning behavior logs, and online skills test data through multimodal behavior analysis technology. It combines a principal component analysis model optimized by transfer learning with the results of regional industrial policy text mining to achieve dynamic matching analysis between personal skills and industry trends. The path planning submodule utilizes a hierarchical planning algorithm based on multi-agent reinforcement learning to plan short-term, medium-term, and long-term career development paths based on the results of the personal information analysis submodule. Each time interval is dynamically adjusted according to the half-life of the industry's technical knowledge update cycle, which is defined as the time required for 50% of the existing technical knowledge in a certain field to be replaced by new technologies.

5. A talent recommendation system based on matching vocational education with local productivity as described in claim 4, characterized in that: The time frame for the short-term career development path is 0.5 to 2 years, the time frame for the medium-term career development path is 2 to 4 years, and the time frame for the long-term career development path is more than 4 years. The path planning submodule dynamically adjusts the time interval based on the half-life of the industry's technical knowledge update cycle, which is defined as the time required for 50% of the existing technical knowledge in a certain field to be replaced by new technologies. Shortest path: 0.5≤t s ≤min(2,τ); Mid-term path: max(2, τ) ≤ t m ≤min(4, 2τ); Long-term path: max(4, 2τ) ≤ t l ; Where τ is the half-life of the industry's technological knowledge update cycle, defined as the time required for 50% of the existing technological knowledge in a certain field to be replaced by new technologies, t s t m and t l These represent the upper limits of the time intervals for short-term, medium-term, and long-term paths, respectively.

6. A talent recommendation system based on matching vocational education with local productivity as described in claim 1, characterized in that: The talent development prediction module includes a data processing unit and a prediction analysis unit; The data processing unit collects historical talent development data and industry dynamic change data, and uses generative adversarial networks combined with adversarial training to enhance the data; The predictive analysis unit uses an adaptive weighted Transformer model, which dynamically adjusts the weights of each sub-model through reinforcement learning. It introduces the industry technology iteration index T(t) as an external feature to construct a dynamic time-series prediction model. Where P(t) is the final prediction result at the current time t, k is the number of sub-models, i.e., the total number of sub-models integrated in the model, and ω i (t) represents the weight of the i-th sub-model at time t. ε is the industry volatility coefficient, P i (t) represents the prediction result of the i-th sub-model at time t.

7. A talent recommendation system based on matching vocational education with local productivity as described in claim 1, characterized in that: The path adjustment module includes a cause analysis submodule and a scheme adjustment submodule; When talents fail to achieve their growth goals on time, the causal analysis submodule uses the causal forest algorithm combined with bias analysis and introduces a counterfactual reasoning mechanism to identify the reasons for slow skill learning progress and sudden changes in industry demand. The scheme adjustment submodule employs a hierarchical multi-objective optimization algorithm, assigning dynamic weights γ to the short-term objective of skill mastery O1, the medium-term objective of job matching O2, and the long-term objective of industry contribution value O3. i (t), constructing a comprehensive evaluation function based on a priority decision-making model of game theory: Where γ is the conflict moderating factor, γ i (t) represents the dynamic weight of the i-th target at time t, C j Let be the conflict coefficient of the j-th target. And weight γ i (t) is dynamically updated using the following formula: Where η is the learning rate, V(π) t ) is the strategy π t The value function, Indicates O i The partial derivatives, For strategy π t Value function V(π) t For target O i gradient, γ i (t+1) represents the weight of the i-th target at time t+1, γ i (t) represents the weight of the i-th objective at time t.

8. A talent recommendation system based on matching vocational education with local productivity as described in claim 1, characterized in that: The talent recommendation module includes a matching submodule and a recommendation execution submodule; The matching submodule uses a bidirectional graph neural network based on graph attention mechanism to simultaneously model the semantic relationship between job requirements and talent development path, and introduces a reinforcement learning-driven dynamic matching degree update mechanism to calculate the bidirectional matching probability between job and talent. The recommendation execution submodule is based on a multi-round bilateral matching game model, which optimizes the recommendation strategy by combining enterprise feedback data, forming a closed loop of recommendation, feedback and iteration.

9. A talent recommendation system based on matching vocational education with local productivity as described in claim 1, characterized in that: The data storage module uses a distributed graph database combined with time-series data sharding technology to partition and store multimodal data, and achieves cross-institutional data sharing and privacy protection through a federated learning data security aggregation mechanism based on homomorphic encryption.

10. A talent recommendation system based on matching vocational education with local productivity as described in claim 1, characterized in that: The interactive interface module provides an AR-visualized career path sandbox, supports multimodal interactive input methods including voice, gestures, and text, and integrates an intelligent question-and-answer assistant based on dialogue enhancement Transformer to answer career planning questions in real time. The system management module integrates a regional industry talent supply and demand imbalance risk early warning function based on Kalman filtering, identifies industry talent supply and demand imbalance risks based on anomaly detection algorithms, and adopts a dynamic permission management system based on attribute-based encryption to realize data access control for different roles.

Images