A method for dynamic adjustment and optimization of higher vocational major groups based on industrial talent demand big data

Abstract

The application relates to the technical field of vocational education informatization and big data analysis, and discloses a high vocational major group dynamic adjustment and optimization method based on industrial talent demand big data, multi-source heterogeneous industrial data acquisition and standardized processing, construction of a five-dimensional production-education connection knowledge graph, industrial talent demand prediction based on an LSTM-Attention hybrid model, gap analysis between major groups and industrial demand, calculation of a gap comprehensive index ZH, analysis of the calculated gap comprehensive index ZH and corresponding adjustment and optimization processing; the application can predict the number of talents required for each industrial core post, the skill structure change trend and the salary level change direction in advance by 1 to 3 years through real-time acquisition of multi-source heterogeneous industrial data through a customized interface and time sequence prediction through a hybrid model, the prediction error is lower, real-time perception and quantitative prediction of industrial demand changes are realized, and the problem that production-education connection lags behind industrial changes is solved.

Description

Technical Field

[0001] This invention relates to the field of vocational education informatization and big data analysis technology, and more specifically to a method for dynamic adjustment and optimization of higher vocational professional groups based on big data of industry talent demand. Background Technology

[0002] The second phase of the "Double High Plan" (2025-2029) takes "high-level school management capabilities and high-quality industry-education integration" as its core orientation. It requires higher vocational colleges to improve the operation and management mechanism of professional groups and the dynamic adjustment mechanism of professional groups, and to respond quickly to the latest requirements of industrial transformation. The Ministry of Education has clearly required that "professional dynamic adjustments be carried out to adapt to economic structure, industrial needs and social needs." However, the current dynamic adjustment of professional groups in higher vocational colleges generally faces the following technical challenges. In existing technologies, the collection of industry talent demand mainly relies on periodic manual surveys. Data updates are lagging and coverage is limited, making it impossible to capture in real time the impact of rapid changes in industry structure and technological iterations on job competency requirements. There is a lack of quantitative assessment methods for the overall health of professional clusters, and the gaps in various dimensions such as scale supply and demand, competency training, curriculum system and faculty structure during the construction of professional clusters cannot be systematically identified and quantified. Existing technical solutions only provide data display and lack the ability to automatically convert from diagnosis to prescription, resulting in a gap between predicted results and specific adjustment measures. Summary of the Invention

[0003] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide a method for dynamic adjustment and optimization of higher vocational professional groups based on big data of industry talent demand, so as to solve the technical problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a method for dynamically adjusting and optimizing vocational college professional clusters based on big data on industry talent demand, comprising the following steps: Step S1: Collection and standardization of multi-source heterogeneous industrial data; Step S2: Construct a five-dimensional knowledge graph for industry-education integration; Step S3: Predict industry talent demand based on the LSTM-Attention hybrid model; Step S4: Analyze the gap between professional clusters and industry needs, and calculate the comprehensive gap index ZH; Step S5: Analyze the calculated gap index ZH and make corresponding adjustments and optimizations.

[0005] In a preferred embodiment, during step S1, when collecting data, a customized data interface is used to connect with government industry databases, recruitment information platforms, industry research report platforms, and vocational school management systems to collect four types of data: industry economic indicators, job demand data, technology development dynamics data, and education and school management data. The collected multi-source heterogeneous data is then standardized, including data cleaning, data format unification, data encoding mapping, and data quality verification, to form a standardized data stream.

[0006] In a preferred embodiment, in step S2, when constructing the five-dimensional industry-education integration knowledge graph, natural language processing technology is used to extract core entities from unstructured recruitment texts, policy documents, and technical documents. The core entities include industry entities, job entities, competency entities, professional entities, and course entities. The relationships between entities are constructed based on the rules of vocational education and the co-occurrence frequency of entities. Numerical industry data is bound to the corresponding entities as attributes to form an industry-education integration knowledge graph containing semantic associations and quantitative information.

[0007] In a preferred embodiment, the relationship between entities in step S2 is as follows: establishing a mapping relationship between industry classification and job directory, extracting the core competency requirements of each job to establish a job-competency matrix, mapping competency units to corresponding professional training objectives, and establishing a support relationship between professional curriculum system and competency training. The knowledge graph established in step S2 supports dynamic updates and automatically adjusts the association weights between entities when industry data changes.

[0008] In a preferred embodiment, when performing demand prediction in step S3, an LSTM-Attention hybrid model is first established. The establishment steps are as follows: Step S31: Organize the industrial economic indicator data and job talent demand data collected in Step S1 according to time series to construct a training dataset; Step S32: Use a bidirectional LSTM network to capture the bidirectional evolution characteristics of the time series and extract the temporal dependency of talent demand; Step S33: Introduce an attention mechanism to assign differentiated weights to key factors affecting talent demand and strengthen the characteristic expression of key elements. Step S34: Simultaneously output two results through multi-task learning: prediction of job demand quantity and prediction of job skill structure. Step S35: Use reinforcement learning algorithm to dynamically optimize model parameters and control prediction error within 10%; Based on the models established in steps S31-S35, the talent demand, skill structure change trends, and salary level changes for core positions in various industries are predicted for the next 1 to 3 years.

[0009] In a preferred embodiment, during step S4, the gap analysis first calculates the comprehensive gap index ZH between the professional cluster and industry demand. The formula for calculating the comprehensive gap index ZH is as follows: In the formula, GM is the scale gap index, NL is the capability gap index, KC is the curriculum gap index, SZ is the teacher gap index, k1, k2, k3, and k4 are the weights corresponding to each index, and k1+k2+k3+k4=1, XS is the shortcoming penalty coefficient, and BZ is the standard deviation of the four gap indices.

[0010] In a preferred embodiment, in step S5, the calculated gap comprehensive index ZH is compared with its internal first threshold Y1 and second threshold Y2, and the first threshold Y1 < the second threshold Y2. When the gap comprehensive index ZH < the first threshold Y1, it is judged as healthy level; when the first threshold Y1 ≤ the gap comprehensive index ZH < the second threshold Y2, it is judged as sub-healthy level; and when the gap comprehensive index ZH ≥ the second threshold Y2, it is judged as dangerous level.

[0011] In a preferred embodiment, when step S5 is determined to be at a healthy level, the skill demand vector for job skills predicted in step S3 for the next 1 to 3 years is compared with the skill demand vector for the current year, and the skill change rate is calculated. When the skill change rate is greater than a preset change threshold, a forward-looking risk warning signal is triggered, and a corresponding forward-looking fine-tuning strategy suggestion is matched simultaneously. The forward-looking fine-tuning strategy suggestion is: while maintaining the overall structure of the existing professional group, embed short-term experiential training courses corresponding to the predicted new skills into the existing courses.

[0012] In a preferred embodiment, when step S5 determines the condition to be sub-healthy, the contribution percentage of each of the scale gap index GM, ability gap index NL, curriculum gap index KC, and teacher structure gap index SZ to the comprehensive gap value ZH is calculated. The four dimensions are then sorted from largest to smallest contribution percentage, and the corresponding courses are added in order.

[0013] In a preferred embodiment, when step S5 is determined to be at a dangerous level, the scale gap index GM is retrieved and a corresponding deep intervention strategy is generated based on its direction. When the scale gap index ≥ 0, it indicates that the industry has a demand for this direction. At this time, it is further determined whether the college has the teaching capacity for new skills. If it does, a light intervention to expand enrollment is implemented. If it does not, an application for a new major direction is implemented. When the scale gap index GM < 0, it indicates that the industry's demand for this direction tends to shrink. At this time, it is further determined whether the social demand for this direction is sufficient. If there is basically no demand, a phased exit and resource restructuring plan is implemented. If there is still a small amount of demand, a contraction plan to reduce enrollment and appropriately reassign teachers is implemented.

[0014] The technical effects and advantages of this invention are as follows: 1. This invention collects multi-source heterogeneous industry data in real time through a customized interface, and performs time-series prediction using an LSTM-Attention hybrid model. It can predict the talent demand, skill structure change trends, and salary level changes of core positions in various industries 1 to 3 years in advance, with lower prediction error. It realizes real-time perception and quantitative prediction of changes in industry demand, and solves the problem of industry-education integration lagging behind industry changes. 2. This invention calculates quantitative indicators in four dimensions: scale gap index, capability gap index, curriculum gap index, and faculty structure gap index, and introduces a shortcoming penalty item to integrate them into a comprehensive gap index ZH. This quantifies the overall health status of professional groups into comparable and ranking values. Based on the comprehensive gap index ZH, professional groups are divided into three levels: healthy, sub-healthy, and dangerous. Differentiated diagnostic depth and strategy types are configured for each level, realizing a multi-dimensional and systematic quantitative assessment and graded early warning of the health of professional groups. 3. This invention uses a rule engine to automatically map the gap analysis results into hierarchical strategy suggestions: a healthy level triggers a forward-looking fine-tuning strategy, a sub-healthy level triggers a targeted repair strategy, and a dangerous level triggers a deep intervention strategy. It automatically generates specific adjustment plans, including professional structure optimization, curriculum system adjustment, faculty configuration optimization, and practical training resource updates, thus achieving an automated closed loop from gap analysis to executable adjustment plans. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the dynamic adjustment and optimization method of the present invention. Detailed Implementation

[0016] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. In addition, the forms of the various structures described in the following embodiments are merely illustrative. The method for dynamic adjustment and optimization of higher vocational professional groups based on big data of industry talent demand involved in the present invention is not limited to the structures described in the following embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Reference Figure 1 This invention provides a method for dynamically adjusting and optimizing vocational college professional clusters based on big data on industry talent demand, comprising the following steps: Step S1: Collection and standardization of multi-source heterogeneous industrial data; Step S2: Construct a five-dimensional knowledge graph for industry-education integration; Step S3: Predict industry talent demand based on the LSTM-Attention hybrid model; Step S4: Analyze the gap between professional clusters and industry needs, and calculate the comprehensive gap index ZH; Step S5: Analyze the calculated gap index ZH and make corresponding adjustments and optimizations.

[0018] In this embodiment, an end-to-end automated processing chain is constructed, consisting of "industry data collection → knowledge graph construction → demand forecasting → gap quantification → hierarchical diagnosis → strategy generation". This enables the changes in industry talent demand to be fully integrated from perception to response within the same technical framework. Compared with existing technologies where data collection, demand forecasting, and professional adjustment are fragmented and rely on manual coordination, this application achieves a fully automated closed loop, solving the technical problems of poor information flow and broken decision-making chains in the dynamic adjustment of higher vocational professional groups.

[0019] Reference Figure 1 In step S1, during data collection, a customized data interface is used to connect with government industry databases, recruitment information platforms, industry research report platforms, and vocational school management systems to collect four types of data: industry economic indicators, job talent demand data, technology development dynamics data, and education and school operation data. The collected multi-source heterogeneous data is then standardized, including data cleaning, data format unification, data encoding mapping, and data quality verification, to form a standardized data stream.

[0020] In this embodiment, the application addresses the problems of long cycles, narrow coverage, and inconsistent data formats associated with traditional manual surveys by using customized interfaces and standardized processing of multi-source heterogeneous data. The targeted collection of four types of data ensures comprehensive capture of signals from industry policies, market demands, and technological iterations. The four standardization processes guarantee that data from different sources maintain consistent format specifications and quality levels before entering subsequent analysis models, providing a data foundation for the accurate training of the LSTM-Attention prediction model and the precise construction of the industry-education integration knowledge graph.

[0021] Reference Figure 1In step S2, when constructing the five-dimensional industry-education integration knowledge graph, natural language processing technology is used to extract core entities from unstructured recruitment texts, policy documents, and technical documents. The core entities include industry entities, job entities, competency entities, professional entities, and course entities. Based on the rules of vocational education and the co-occurrence frequency of entities, the association between entities is constructed. Numerical industry data is used as attributes and bound to the corresponding entities to form an industry-education integration knowledge graph containing semantic associations and quantitative information. In step S2, the association between entities is as follows: establishing a mapping relationship between industry classification and job directory, extracting the core competency requirements of each job to establish a job-competency matrix, mapping competency units to the corresponding professional training objectives, and establishing a supporting relationship between professional curriculum system and competency training. The knowledge graph established in step S2 supports dynamic updates and automatically adjusts the association between entities when industry data changes.

[0022] In this embodiment, the application addresses the semantic gap between industry needs and professional training by constructing a five-dimensional knowledge graph. This five-dimensional graph semantically links the industry and education sectors through capabilities, enabling automatic tracing of corresponding capability units, professional directions, and course modules when industry changes. This transforms qualitative analysis, previously reliant on expert experience, into a computable and reasonable structured knowledge system. Through precise definitions of four types of relationships, a full-link traceability system from macro-industry to micro-course is constructed. When industry data is updated, the system automatically triggers recalculation of association weights along the "industry → position → capability → major → course" link, ensuring the graph remains synchronized with the latest industry developments and achieving continuous self-evolution of the industry-education integration knowledge system.

[0023] Reference Figure 1 In step S3, when performing demand prediction, an LSTM-Attention hybrid model is first established. The establishment steps are as follows: Step S31: Organize the industrial economic indicator data and job talent demand data collected in Step S1 according to time series to construct a training dataset; Step S32: Use a bidirectional LSTM network to capture the bidirectional evolution characteristics of the time series and extract the temporal dependency of talent demand; Step S33: Introduce an attention mechanism to assign differentiated weights to key factors affecting talent demand and strengthen the characteristic expression of key elements. Step S34: Simultaneously output two results through multi-task learning: prediction of job demand quantity and prediction of job skill structure. Step S35: Use reinforcement learning algorithm to dynamically optimize model parameters and control prediction error within 10%; Based on the models established in steps S31-S35, the talent demand, skill structure change trends, and salary level changes for core positions in various industries are predicted for the next 1 to 3 years.

[0024] In this embodiment, the application utilizes an LSTM-Attention hybrid model to achieve accurate time-series prediction of industry talent demand. The bidirectional LSTM captures time-series features from both forward and backward directions, avoiding the limitation of traditional unidirectional LSTMs that can only utilize historical information while ignoring future trend features. The attention mechanism automatically identifies and focuses on key factors affecting talent demand, avoiding the decrease in prediction accuracy caused by treating all features equally. Multi-task learning simultaneously outputs two dimensions: quantity prediction and skill structure prediction. This ensures that the prediction results include both scale information ("how many people are needed") and structural information ("what abilities are needed"), providing complete input for subsequent gap analysis and solution generation. Reinforcement learning dynamic optimization makes the model adaptive, controlling the prediction error within 10% and achieving higher accuracy.

[0025] Reference Figure 1 In step S4, during the gap analysis, the comprehensive gap index ZH between the professional cluster and industry demand is first calculated. The formula for calculating the comprehensive gap index ZH is as follows: In the formula, GM is the scale gap index, NL is the capability gap index, KC is the curriculum gap index, SZ is the teacher gap index, k1, k2, k3, and k4 are the weights corresponding to each index, and k1+k2+k3+k4=1 is the shortcoming penalty coefficient. It adopts a method that combines historical data-driven and expert experience verification: First, based on the college's historical professional construction data and industrial development data over the past 5 years, different shortcoming penalty coefficients XS are set, and the comprehensive gap value ZH of each professional group in each year is calculated. The consistency between the value and the actual professional adjustment decision is backtested and verified. With the goal of maximizing the consistency, the optimal shortcoming penalty coefficient XS is determined, and BZ is the standard deviation of the four gap indices. The formula for calculating the size gap index (GM) is: In the formula, Sigmoid is the sigmoid activation function, n is the total number of core positions, i is the position number, wi is the weight coefficient of the i-th position, which reflects the importance of the i-th position in the industrial chain and is determined by combining the analytic hierarchy process with the proportion of industrial output value and the proportion of employees in the position, Di is the predicted annual demand for the i-th position, and S is the annual supply of graduates in the corresponding major of the college. A small constant to prevent division by zero errors; The formula for calculating the Capability Gap Index (NL) is: In the formula, m represents the total number of job competency items, and j represents the competency item number. Let be the weight coefficient of the j-th ability, reflecting its importance in the job competency evaluation, and determined by the entropy weight method. Let represent the industry requirement achievement level for the j-th capability, which is obtained by quantitative mapping from the industry job skill standards predicted in step S3. 1 indicates that the industry requirement for this capability is at the highest level, and 0 indicates that there is no requirement. The achievement rate of the j-th ability in the college's training program is generated by the college's score, where 1 indicates full achievement and 0 indicates no training at all. It is a positive gap operator; The formula for calculating the curriculum gap index KC is: In the formula, XY is the set of existing course knowledge points, and XQ is the set of industry demand knowledge points; The formula for calculating the teacher gap index SZ is as follows: In the formula, ReLU is a linear rectified function, q is the total number of skill categories required for the new industry direction, and j is the skill category number. The importance coefficient for skill type j reflects its criticality in industrial upgrading and is determined by a combination of technology roadmap analysis and expert scoring. The number of teachers required for skill type j. This refers to the number of teachers in the college who actually possess the teaching ability for the j-th type of skill.

[0026] In this embodiment, the comprehensive gap index ZH, which is a four-dimensional weighted fusion plus a shortcoming penalty, quantifies the complex multi-dimensional gap between professional groups and industry needs into a single, comparable value. The introduction of the shortcoming penalty item reflects the "barrel effect" principle. When the gap in a certain dimension is particularly prominent, even if other dimensions perform well, the overall gap will be reasonably amplified, thereby avoiding the averaging effect from masking serious structural shortcomings. This design enables the comprehensive gap index ZH to reflect the overall health level of professional groups and to provide a reasonable punitive assessment for "uneven" professional groups, providing a more accurate quantitative basis for subsequent health level classification and strategy matching.

[0027] Reference Figure 1 In step S5, the calculated gap comprehensive index ZH is compared with its internal first threshold Y1 and second threshold Y2, and the first threshold Y1 < the second threshold Y2. When the gap comprehensive index ZH < the first threshold Y1, it is judged as healthy level; when the first threshold Y1 ≤ the gap comprehensive index ZH < the second threshold Y2, it is judged as sub-healthy level; when the gap comprehensive index ZH ≥ the second threshold Y2, it is judged as dangerous level.

[0028] In this embodiment, the application uses a three-level health grading mechanism to transform continuously changing comprehensive gap values ​​into discrete management decision levels, enabling professional group managers to quickly and intuitively judge the health status of the professional group. The three levels correspond to different treatment depths: a healthy level means no major intervention is needed, a sub-healthy level means targeted repair is needed, and a dangerous level means "deep intervention is necessary." This grading diagnosis logic avoids the crude mode of treating all problems the same or relying solely on intuition in traditional management, thus achieving refined and differentiated management of professional groups.

[0029] Reference Figure 1 When step S5 is determined to be at the healthy level, the skill demand vector for job skills predicted in step S3 for the next 1 to 3 years is obtained and compared with the skill demand vector for the current year. The skill change rate is calculated. When the skill change rate is greater than the preset change threshold, a forward-looking risk warning signal is triggered, and a corresponding forward-looking fine-tuning strategy suggestion is matched simultaneously. The forward-looking fine-tuning strategy suggestion is: on the premise of maintaining the overall structure of the existing professional group unchanged, embed short-term experiential training courses corresponding to the predicted new skills into the existing courses.

[0030] In this embodiment, the application introduces a forward-looking early warning mechanism at the health level to realize the transformation from post-remediation to pre-prevention. The application uses the predictive ability of step S3 to actively scan the future trend of skill change. When it is predicted that the skill structure will change significantly in the next 1 to 3 years, an early warning signal of impending decoupling is issued in advance and a fine-tuning strategy is matched, so that the college can start a smooth transition before the actual changes in industry demand occur.

[0031] Reference Figure 1 When step S5 determines the condition to be sub-healthy, the contribution percentage of each of the scale gap index GM, ability gap index NL, curriculum gap index KC, and teacher structure gap index SZ to the comprehensive gap value ZH is calculated. The four dimensions are then sorted from largest to smallest contribution percentage, and the corresponding courses are added in order.

[0032] In this embodiment, the applicant is in a sub-healthy state, and there are different degrees of gaps in all four dimensions. After sorting, the corresponding courses are added in order to target the repair methods, so as to ensure that the limited school resources are invested in the direction with the greatest improvement leverage effect and avoid the problem of waste due to the average distribution of resources.

[0033] Reference Figure 1When step S5 is judged to be at a dangerous level, the scale gap index GM is retrieved and a corresponding deep intervention strategy is generated according to its direction. When the scale gap index ≥ 0, it indicates that the industry has a demand for this direction. At this time, it is further judged whether the college has the teaching ability for new skills. If it does, a light intervention to expand enrollment is implemented. If it does not, an application for a new major direction is implemented. When the scale gap index GM < 0, it indicates that the industry's demand for this direction tends to shrink. At this time, it is further judged whether the social demand for this direction is sufficient. If there is basically no demand, a phased exit and resource restructuring plan is implemented. If there is still a small amount of demand, a contraction plan to reduce enrollment and appropriately transfer teachers is implemented.

[0034] In this embodiment, a two-step decision tree is used with the scale gap index GM as the first judgment variable. When the scale gap index GM ≥ 0 and the college has teaching capabilities, it indicates that the college is in a favorable position and should adopt an expansion strategy. When the scale gap index GM ≥ 0 but the college does not have teaching capabilities, it indicates that the favorable position has arrived but the college is not prepared and needs structural transformation. When the scale gap index GM < 0 and there is basically no demand, it indicates that this direction is outdated and a decisive exit is needed. When the scale gap index GM < 0 but there is still a small amount of demand, it indicates that "there is still one last wave" and a contraction rather than a complete exit is needed. The precise matching of these four strategies avoids erroneous decisions by managers in dangerous situations and reduces the policy risks and economic costs of professional group adjustments.

[0035] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. The units and algorithm steps of the various examples described in the embodiments can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0036] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0037] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0038] 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. A method for dynamically adjusting and optimizing vocational college professional clusters based on big data on industry talent demand, characterized in that: Includes the following steps: Step S1: Collection and standardization of multi-source heterogeneous industrial data; Step S2: Construct a five-dimensional knowledge graph for industry-education integration; Step S3: Predict industry talent demand based on the LSTM-Attention hybrid model; Step S4: Analyze the gap between professional clusters and industry needs, and calculate the comprehensive gap index ZH; Step S5: Analyze the calculated gap index ZH and make corresponding adjustments and optimizations.

2. The method for dynamic adjustment and optimization of higher vocational professional clusters based on big data of industry talent demand as described in claim 1, characterized in that: In step S1, during data collection, a customized data interface is used to connect with government industry databases, recruitment information platforms, industry research report platforms, and vocational school management systems to collect four types of data: industry economic indicators, job talent demand data, technology development dynamics data, and education and school operation data. The collected multi-source heterogeneous data is then standardized, including data cleaning, data format unification, data encoding mapping, and data quality verification, to form a standardized data stream.

3. The method for dynamic adjustment and optimization of higher vocational professional clusters based on big data of industry talent demand as described in claim 1, characterized in that: In step S2, when constructing the five-dimensional industry-education integration knowledge graph, natural language processing technology is used to extract core entities from unstructured recruitment texts, policy documents, and technical documents. The core entities include industry entities, job entities, competency entities, professional entities, and course entities. The relationships between entities are constructed based on the rules of vocational education and the co-occurrence frequency of entities. Numerical industry data is bound to the corresponding entities as attributes to form an industry-education integration knowledge graph containing semantic associations and quantitative information.

4. The method for dynamic adjustment and optimization of higher vocational professional clusters based on big data of industry talent demand, as described in claim 3, is characterized in that: In step S2, the relationships between entities are as follows: establishing a mapping relationship between industry classification and job directory, extracting the core competency requirements of each job to establish a job-competency matrix, mapping competency units to corresponding professional training objectives, and establishing a support relationship between professional curriculum system and competency training. The knowledge graph established in step S2 supports dynamic updates and automatically adjusts the association weights between entities when industry data changes.

5. The method for dynamic adjustment and optimization of higher vocational professional clusters based on big data of industry talent demand as described in claim 1, characterized in that: When performing demand prediction in step S3, the LSTM-Attention hybrid model is first established. The establishment steps are as follows: Step S31: Organize the industrial economic indicator data and job talent demand data collected in Step S1 according to time series to construct a training dataset; Step S32: Use a bidirectional LSTM network to capture the bidirectional evolution characteristics of the time series and extract the temporal dependency of talent demand; Step S33: Introduce an attention mechanism to assign differentiated weights to key factors affecting talent demand and strengthen the characteristic expression of key elements. Step S34: Simultaneously output two results through multi-task learning: prediction of job demand quantity and prediction of job skill structure. Step S35: Use reinforcement learning algorithm to dynamically optimize model parameters and control prediction error within 10%; Based on the models established in steps S31-S35, the talent demand, skill structure change trends, and salary level changes for core positions in various industries are predicted for the next 1 to 3 years.

6. The method for dynamic adjustment and optimization of higher vocational professional clusters based on big data of industry talent demand as described in claim 1, characterized in that: In step S4, during the gap analysis, the comprehensive gap index ZH between the professional cluster and industry demand is first calculated. The formula for calculating the comprehensive gap index ZH is as follows: In the formula, GM is the scale gap index, NL is the capability gap index, KC is the curriculum gap index, SZ is the teacher gap index, k1, k2, k3, and k4 are the weights corresponding to each index, and k1+k2+k3+k4=1, XS is the shortcoming penalty coefficient, and BZ is the standard deviation of the four gap indices.

7. The method for dynamic adjustment and optimization of higher vocational professional clusters based on big data of industry talent demand as described in claim 1, characterized in that: In step S5, the calculated gap comprehensive index ZH is compared with its internal first threshold Y1 and second threshold Y2, and the first threshold Y1 < the second threshold Y2. When the gap comprehensive index ZH < the first threshold Y1, it is judged as healthy level; when the first threshold Y1 ≤ the gap comprehensive index ZH < the second threshold Y2, it is judged as sub-healthy level; when the gap comprehensive index ZH ≥ the second threshold Y2, it is judged as dangerous level.

8. The method for dynamic adjustment and optimization of higher vocational professional clusters based on big data of industry talent demand, as described in claim 7, is characterized in that: When step S5 is determined to be at the healthy level, the skill demand vector for job skills predicted in step S3 for the next 1 to 3 years is compared with the skill demand vector for the current year, and the skill change rate is calculated. When the skill change rate is greater than the preset change threshold, a forward-looking risk warning signal is triggered, and a corresponding forward-looking fine-tuning strategy suggestion is matched simultaneously. The forward-looking fine-tuning strategy suggestion is: while maintaining the overall structure of the existing professional group, embed short-term experiential training courses corresponding to the predicted new skills into the existing courses.

9. The method for dynamic adjustment and optimization of higher vocational professional clusters based on big data of industry talent demand, as described in claim 7, is characterized in that: When step S5 determines the condition to be sub-healthy, the contribution percentage of each of the scale gap index GM, ability gap index NL, curriculum gap index KC, and teacher structure gap index SZ to the comprehensive gap value ZH is calculated. The four dimensions are then sorted from largest to smallest contribution percentage, and the corresponding courses are added in order.

10. The method for dynamic adjustment and optimization of higher vocational professional clusters based on big data of industry talent demand, as described in claim 7, is characterized in that: When step S5 is determined to be at a dangerous level, the scale gap index GM is retrieved and a corresponding deep intervention strategy is generated based on its direction. When the scale gap index ≥ 0, it indicates that the industry has a demand for this direction. At this time, it is further determined whether the college has the teaching capacity for new skills. If it does, a light intervention to expand enrollment is implemented. If it does not, an application for a new major direction is implemented. When the scale gap index GM < 0, it indicates that the industry's demand for this direction tends to shrink. At this time, it is further determined whether the social demand for this direction is sufficient. If there is basically no demand, a phased exit and resource restructuring plan is implemented. If there is still a small amount of demand, a contraction plan to reduce enrollment and appropriately reassign teachers is implemented.

Images