High vocational college quality evaluation system construction method based on production and teaching fusion

Abstract

The invention discloses a higher vocational college quality evaluation system construction method based on production and teaching integration, and particularly relates to the technical field of production and teaching integration teaching assistance. Based on enterprise support resources and student actual receiving information, preset and actual knowledge network topological graphs are constructed respectively, and standardized knowledge points are extracted through natural language processing; generating a migration difficulty weight in combination with expert scores and industry standards, and forming a directed weighted graph; a support resource quality coefficient is generated by comparing a topological structure, semantic content and course stage integrating degree of the topological graph, calculating a skill knowledge matching degree, an aging synchronization rate and an operation integrating degree and performing conjoint analysis, and if the coefficient is lower than a threshold value, early warning is triggered to prompt optimization of resource supply; and by taking maximization of a quality coefficient as a target, redividing a course stage boundary based on a dynamic programming algorithm, correcting an operation integrating degree formula in combination with a migration path efficiency coefficient, preferentially adjusting a high migration difficulty stage, and realizing real-time alignment of a teaching rhythm and an industrial demand.

Description

Technical Field

[0001] The present invention relates to the field of industry-education integration teaching auxiliary technology, and more specifically, to a method for constructing a quality evaluation system for higher vocational colleges based on industry-education integration. Background Art

[0002] The integration of industry and education has become the core path for higher vocational colleges to connect with industrial upgrading and cultivate high-quality technical and skilled talents. The collaborative construction of project-based teaching systems between schools and enterprises has become the norm, but the structural contradiction between the quality of teaching resource supply and the speed of industrial technology iteration has become increasingly prominent, and it is urgent to establish a scientific and quantitative dynamic evaluation mechanism to ensure the effectiveness of industry-education collaboration.

[0003] In the actual operation of the industry-education integration project, there is a problem that teaching resources cannot be dynamically adapted to industry needs, which is specifically reflected in: 1) The existing quality evaluation system for higher vocational colleges relies on static indicators and lacks dynamic modeling of the knowledge network topology structure and migration difficulty coefficient, making it difficult to detect the knowledge coverage deviation between the practical training scenarios provided by enterprises and the teaching materials; 2) The course phase adjustment mechanism is lagging behind. The traditional method uses a fixed cycle to reorganize the course modules, which cannot optimize the teaching rhythm in real time according to the changes in industry benchmark working hours; 3) The quality evaluation dimension is single. The existing technology only focuses on the coverage of knowledge points, and has not established a multi-parameter coupling analysis model of skill matching, time synchronization rate and cognitive load gradient, resulting in a lack of data support for resource optimization; These problems seriously restrict the implementation of industry-education integration projects in an environment of rapid technological change. Summary of the invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method for constructing a quality evaluation system for higher vocational colleges based on the integration of industry and education. By constructing a dynamic knowledge network topology map, quantifying the migration difficulty coefficient and dynamically optimizing the course stage boundaries, efficient adaptation and resource optimization of the quality evaluation system for the integration of industry and education are achieved, so as to solve the problems of disconnection between teaching needs and industrial needs and unreasonable resource allocation raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solution: a method for constructing a quality evaluation system for higher vocational colleges based on the integration of industry and education, comprising the following steps: Step 1: Based on the preset requirements of the industry-education integration project, a preset knowledge network topology map is constructed; based on the preset knowledge network topology map, the course stages of the industry-education integration project are set up, the supporting resources of the enterprise in the industry-education integration project are collected, and based on the actual information received by students in the industry-education integration project, an actual knowledge network topology map is constructed; the supporting resources include teaching materials and practical training scenarios; Step 2: By comparing the topological structures, semantic contents, and the degree of fit at each course stage of the preset and actual knowledge network topology diagrams, obtain the skill knowledge matching degree, timeliness synchronization rate, and operation fit degree, and obtain the support resource quality coefficient through weighted summation; If the support resource quality coefficient is lower than the preset value, it indicates that in the industry-education integration project, the quality of the support resources provided by the enterprise is lower than the preset requirements, and a warning is sent out to prompt an improvement in the quality of the support resources; if the support resource quality coefficient is not lower than the preset value, it indicates that the quality of the support resources provided by the enterprise meets the requirements and no adjustment is needed; Step 3: Implement dynamic optimization and feedback calibration based on the support resource quality coefficient. When the support resource quality coefficient is lower than the preset value, with maximizing the support resource quality coefficient as the objective function and the total class hours not exceeding the preset upper limit, re-divide the course stage boundaries based on the dynamic programming algorithm; generate a calibrated preset knowledge network topology diagram in real time to achieve the adaptive alignment of resource supply and teaching needs.

[0006] Preferably, parse the enterprise job demand documents, industry standard documents, and enterprise technical logs through natural language processing technology, extract the standardized knowledge point set and eliminate the ambiguity of synonyms; determine the logical dependency relationship between knowledge points based on the expert scoring method, and combine the enterprise engineer interview data to quantify the weight of the knowledge point migration difficulty, and construct a preset knowledge network topology diagram with knowledge points as nodes, dependency relationships as directed edges, and migration difficulty as edge weights.

[0007] Preferably, by collecting the actual teaching materials of enterprise tutors, analyze and output the knowledge points actually covered by the teaching materials; obtain the knowledge points actually covered by the training scenario by building a digital twin model of the student training scenario; combine the knowledge points actually covered by the teaching materials and the training scenario to obtain the actual knowledge network topology diagram; the digital twin model performs format standardization processing on the data flow of heterogeneous devices through the hardware abstraction layer, uses Fourier time-frequency transformation to eliminate the sensor sampling rate difference, and establishes a skill evaluation benchmark coordinate system for cross-vendor devices.

[0008] Preferably, the weight between knowledge points represents the migration difficulty coefficient between the two, and the calculation formula for the migration difficulty coefficient is: The index marks of knowledge points include i and j. Denote the average working hours required to learn knowledge point D_j after mastering knowledge point D_i as Tavg_ij, and denote the average working hours required to learn knowledge point D_i after mastering knowledge point D_j as Tavg_ji. The edge weight corresponding to the edge where knowledge points i and j are located is calculated by the following formula: Among them, Represents the industry benchmark man-hours; based on the standard operation time, equipment operation records, and employee proficiency curve in the enterprise's real production log, a basic man-hour model is constructed to obtain the industry benchmark man-hours.

[0009] Preferably, the process of obtaining the support resource quality coefficient includes: Step S101: Evaluate the skill knowledge matching degree between the actual knowledge network topology map and the preset knowledge network topology map. The quantification method is as follows: Compare the actual knowledge network topology map with the preset knowledge network topology map. For each knowledge point in the preset knowledge network topology map, calculate the gap between the actual output and the preset output; consider the influence of each knowledge point, weight the gap by the influence, and obtain the comprehensive learning deviation; analyze the complexity of the preset knowledge network topology map to obtain the complexity value; divide the comprehensive learning deviation by the complexity value, put it into the inverse proportional exponential function, and calculate the skill knowledge matching degree. Step S102: The timeliness synchronization rate is used to measure the dynamic matching degree between the preset knowledge network topology map and the actual knowledge network topology map in terms of knowledge point update and industrial demand change; it reflects whether the knowledge points designed in the curriculum can adapt to the changes in industrial demand in a timely manner; compare the update frequency and content changes of the knowledge points in the actual knowledge network topology map and the preset knowledge network topology map, and combine the industrial update cycle to quantify the timeliness synchronization rate. Step S103: Evaluate the operation fit degree between the actual knowledge network topology map and the preset knowledge network topology map, which is used to measure the rhythm consistency between the actual knowledge network and the preset knowledge network. The quantification method is as follows: Divide the preset knowledge network topology map into several curriculum stages in the way of the minimum cumulative migration difficulty coefficient, obtain the corresponding preset knowledge network topology map and actual knowledge network topology map for each curriculum stage, compare the differences between the actual knowledge network topology map and the preset knowledge network topology map of each curriculum stage, and obtain the node difference and edge difference; combine the node deviation and the edge deviation, and calculate the operation fit degree of each curriculum stage in the form of the reciprocal; represent the operation fit degree with the harmonic mean of the deviations of each curriculum stage. Step S104: Jointly analyze the skill knowledge matching degree, timeliness synchronization rate, and operation fit degree to obtain the support resource quality coefficient.

[0010] Preferably, calculate the average migration difficulty of the preset knowledge network topology map and the average migration difficulty of the actual knowledge network topology map ; Based on the difference between the preset and actual average migration difficulties, generate a migration path efficiency coefficient, which is used to measure the difference between the preset average migration efficiency of knowledge points and the actual average migration difficulty. If < 1 indicates that the complexity of the actual teaching path significantly deviates from the preset, triggering the dynamic programming algorithm to preferentially adjust the boundary division of the high-migration-difficulty course stage to achieve global optimization of the teaching rhythm and resource supply; when ≥ 1, it indicates that the path efficiency meets the expectation.

[0011] Preferably, use to represent the node difference between the actual knowledge network topology graph and the preset knowledge network topology graph, to represent the edge difference between the actual knowledge network topology graph and the preset knowledge network topology graph. Use k to represent the course stage index, N to represent the number of course stages, and the operation fitness of the kth course stage is denoted as ; The operation fitness is calculated by the formula: .

[0012] Preferably, the method includes an optimization step for course stages based on buffer nodes, including: Technical mutation probability detection: Based on the variational autoencoder, perform probability distribution modeling on the standardized knowledge point set, and adaptively determine the mutation timing of the preset knowledge network topology graph through the statistical characteristics of the reconstruction error; Buffer node generation and insertion: Extract the standardized knowledge point sets before and after the mutation, and construct a mapping rule library for the old and new technology interfaces; Insert buffer nodes at the mutation points of the preset knowledge network topology graph. The buffer nodes are the course stages connecting the old and new technologies to achieve a smooth transition of the technology gap; Dynamic division of course stages: Dynamically adjust the course stage boundaries according to the buffer node distribution density to make the class hour allocation match the rhythm of industrial technology evolution elastically.

[0013] Preferably, the triggering condition for step three is any of the following: When the support resource quality coefficient is lower than the preset threshold, immediately trigger dynamic optimization; trigger optimization when a single course stage of the industry-education integration project is completed or a fixed time period is reached.

[0014] Preferably, the operation process of step three includes: Construct a state space, divide the courses into several candidate stage boundary nodes, each node corresponding to the cumulative class hours, support resource quality coefficient, and preset knowledge network topology graph. Set the upper limit of the total class hours and the non-decreasing constraint of the quality coefficient, with the goal of maximizing the support resource quality coefficient; According to the state transition rule, transfer from the previous course stage boundary node to the current node. By adjusting the knowledge point set, migration path, and class hour allocation, ensure that the cumulative class hours do not exceed the upper limit and the support resource quality coefficient meets the preset threshold, and use the reverse recursive method to calculate the optimal course stage boundary division sequence; Based on the optimal stage boundary division sequence, dynamically update the nodes, edge weights, and stage markers of the preset knowledge network topology map, and align the support resource supply with the teaching demand through enterprise resource reallocation and teaching plan push; Monitor the quality coefficient of the optimized support resources. If it is still lower than the preset value, trigger the dynamic programming algorithm again, and aggregate the optimization data of multiple institutions through the federated learning framework, and periodically calibrate the industry benchmark working hours to improve the adaptability.

[0015] Preferably, in step one, for the data fusion method of the heterogeneous training scenario, in view of the missing physical feedback data of the digital twin model, deploy a pressure sensor array to collect the force distribution of students' operations; use the cross-modal attention mechanism to align the sensor data stream with the virtual operation log, reconstruct the three-dimensional spatio-temporal feature vector to fill the blind area of the training knowledge point coverage, and generate an enhanced actual knowledge network topology map.

[0016] Preferably, in step one, include the multi-modal data timeliness compensation method, construct a multi-dimensional matrix of enterprise resource timeliness, and use the upload timestamp of teaching materials and the data acquisition cycle of training equipment as the time dimension benchmark; use the time decay factor to exponentially weight the confidence of the covered knowledge points, and at the same time access the industrial policy API to obtain the latest technical standards, and generate a knowledge point patch package with a timestamp to dynamically update the actual knowledge network topology map.

[0017] Preferably, based on the enterprise technology update log, industry standard revision notice, and student training failure rate data, construct an incremental update trigger rule; when it is detected that the equipment parameters, process standards, or skill requirements associated with the knowledge points change, automatically extract the new knowledge point set and mark it as a node to be verified; after passing the expert review and industry coverage verification, insert the new node into the preset knowledge network topology map, and recalculate the transfer difficulty weight based on the student's historical learning path, and synchronously update the actual knowledge network topology map; the incremental update trigger rule dynamically adjusts the calibration period of the industry benchmark working hours through a time series prediction model to ensure that the preset knowledge network topology map is synchronized with the industrial technology iteration in real time.

[0018] Preferably, according to the industry attributes (such as manufacturing, information technology) and stage goals (such as skill strengthening, innovation practice) of the industry-education integration project, construct a dynamic weight configuration matrix; based on the fusion model of the analytic hierarchy process (AHP) and the entropy weight method, calculate the weight ratio of skill knowledge matching degree, timeliness synchronization rate, and operation fit degree in real time; when it is detected that the industrial demand suddenly changes or the ability distribution of the student group shifts, trigger the weight adaptive adjustment mechanism, and preferentially increase the weight ratio of the timeliness synchronization rate to quickly respond to the technology iteration.

[0019] Preferably, in the federated learning framework, each institution locally retains the original teaching data and only uploads the encrypted support resource quality coefficient and migration path efficiency coefficient; through homomorphic encryption and differential privacy technologies, noise injection is performed on the aggregated global model parameters to prevent the leakage of sensitive information; based on the distributed aggregation results of the optimized data of multiple institutions, a consensus value of the industry benchmark working hours is periodically generated, and the decentralized synchronization of the calibration results is achieved through smart contracts, solving the problem of cross-regional data islands.

[0020] Preferably, the method further includes: Step 4: Collect the evaluation data of different ports in each industry-education integration project, summarize the evaluation data of different ports, and output a unified-caliber evaluation data set after summarization and standardization, including: Multi-source heterogeneous data tagging processing: Collect the evaluation data of different ports in the industry-education integration project, and use natural language processing and multi-dimensional label aggregation algorithms to perform word segmentation parsing, entity recognition, and feature annotation on the multi-source heterogeneous data to generate a retrievable tagged data set; Unifying the cross-port index caliber: Based on the field mapping table, perform logical conversion on the differences in the names, units, or measurement methods of the same index in different collection ports in the tagged data set to generate a preliminary unified index set; Calculating the comprehensive deviation factor: For each index in the preliminary unified index set, calculate the numerical deviation and distribution deviation from the reference benchmark value, and generate a comprehensive deviation factor matrix after normalization and weighted summation; Dynamic correction and data set generation: Based on the comprehensive deviation factor matrix, perform piecewise correction on the index values through a dynamic correction function regulated by a non-linear sensitivity parameter to generate a unified-caliber data set.

[0021] The technical effects and advantages of the present invention: (1) The method for constructing a quality evaluation system for higher vocational colleges based on industry-education integration provided by the present invention forms a dynamic modeling of a directed weighted graph by constructing a preset and actual knowledge network topology diagram, combining natural language processing technology to extract knowledge points and generate migration difficulty weights; by comparing the topological structure, semantic content, and course stage fit, calculating the skill knowledge matching degree, timeliness synchronization rate, and operation fit, and jointly analyzing to generate a support resource quality coefficient, and based on the dynamic programming algorithm, the course stage boundary is optimized in real time to achieve the adaptive alignment of the teaching rhythm and industrial needs, solving the problems of lagging resource quality adaptation and slow response to course adjustment caused by the static nature of the traditional evaluation system.

[0022] (2) Through the standardized processing of multi-source heterogeneous data and dynamic correction functions, the present invention performs labeled aggregation and outlier correction on text, numerical, and image data, generates an industry-education integration evaluation coefficient by combining industry benchmark values, realizes segmented calibration of cross-industry indicators relying on natural language processing and multi-dimensional labeling algorithms, and ensures data credibility and traceability through an intelligent correction mechanism, thereby solving the problems of multi-source data format conflicts, indicator caliber differences, and insufficient dynamic adaptation ability of the evaluation system. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a flowchart of the method for constructing a quality evaluation system for higher vocational colleges based on the analysis of support resources according to the present invention.

[0024] Figure 2 It is an optimization diagram of the course stage based on the analysis of support resources according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] Hereinafter, exemplary embodiments of the present disclosure will be described in more detail with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be completely conveyed to those skilled in the art.

[0026] At the same time, it should be understood that, for the sake of description, the dimensions of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

[0027] The following description of at least one exemplary embodiment is merely illustrative in nature and in no way constitutes a limitation to the present application and its application or use.

[0028] For technologies, methods, and devices known to those of ordinary skill in the relevant art, detailed discussion may not be made, but in appropriate cases, the said technologies, methods, and devices should be regarded as part of the specification.

[0029] Example 1, referring to Figure 1 the flowchart of the method for constructing a quality evaluation system for higher vocational colleges based on the analysis of support resources and Figure 2 the optimization diagram of the course stage based on the analysis of support resources, the present invention provides a method for constructing a quality evaluation system for higher vocational colleges based on industry-education integration, including: Step 1: Based on the preset requirements of the industry-education integration project, construct a preset knowledge network topology diagram; set the course stage of the industry-education integration project based on the preset knowledge network topology diagram, collect the support resource situation of enterprises in the industry-education integration project, and construct an actual knowledge network topology diagram based on the actual received information of students in the industry-education integration project; the support resources include teaching materials and training scenarios; Step 2: By comparing the topological structures, semantic contents, and the degree of fit at each course stage of the preset and actual knowledge network topologies, obtain the skill-knowledge matching degree, timeliness synchronization rate, and operation fit degree, and obtain the support resource quality coefficient through weighted summation; Explanation: Measure the similarity between the knowledge actually learned by students through the industry-education integration project and the knowledge network designed in the curriculum, and output the skill-knowledge matching degree. The higher the value, the more matching it is; Measure the degree to which the knowledge network update keeps up with the changes in industrial demands. Through weighted calculation of the update delay and importance, output the timeliness synchronization rate. The higher the value, the stronger the synchronization; The operation fit degree measures whether the teaching execution conforms to the curriculum design rhythm. Through stage-by-stage comparison of the teaching duration and the migration difficulty difference calculation, output the operation fit degree. The higher the value, the higher the fit degree; If the support resource quality coefficient is lower than the preset value, it indicates that in the industry-education integration project, the quality of the support resources provided by the enterprise is lower than the preset requirements, and an early warning is sent out to prompt to improve the quality of the support resources; If the support resource quality coefficient is not lower than the preset value, it indicates that the quality of the support resources provided by the enterprise meets the requirements and no adjustment is needed; Step 3: Implement dynamic optimization and feedback calibration according to the support resource quality coefficient. When the support resource quality coefficient is lower than the preset value, with maximizing the support resource quality coefficient as the objective function and the total class hours not exceeding the preset upper limit, re-divide the course stage boundaries based on the dynamic programming algorithm; Generate a calibrated preset knowledge network topology in real time to achieve the adaptive alignment of resource supply and teaching demand.

[0030] In the embodiments of the present invention, it needs to be further explained that by using natural language processing technology (such as the BERT+CRF model) to parse the enterprise job demand documents and industry standard documents, extract the standardized knowledge point set and eliminate the synonym ambiguity; Based on the expert scoring method, determine the logical dependency relationship between knowledge points, combine the enterprise engineer interview data to quantify the knowledge point migration difficulty weights, and construct an initial project graph with knowledge points as nodes, dependency relationships as directed edges, and migration difficulties as edge weights, that is, the preset knowledge network topology; Use the graph database to realize the visual storage and dynamic update of the preset knowledge network topology, and verify the logical relationship and weight error through the industry standard coverage rate verification and enterprise expert review meeting; Establish a dynamic update mechanism. When it is detected that the industry technical standard is updated or the enterprise demand changes, use incremental updates to adjust the nodes and edges (such as deleting obsolete knowledge points and adding skill nodes), and re-calibrate the weights based on the student training migration failure rate data to ensure that the project graph is synchronized with the industrial demands in real time.

[0031] In the embodiments of the present invention, it needs to be further explained that by collecting the actual teaching materials of enterprise tutors and analyzing them, the knowledge points actually covered by the teaching materials are output; by building a digital twin model of the student training scenario, the knowledge points actually covered by the training scenario are obtained; by synthesizing the knowledge points actually covered by the teaching materials and the training scenario, an actual knowledge network topology map is obtained; the digital twin model standardizes the data flow format of heterogeneous devices through a hardware abstraction layer, eliminates the sensor sampling rate difference by using Fourier time-frequency transformation, and establishes a skill evaluation benchmark coordinate system for cross-vendor devices. In a possible embodiment, in the process of obtaining the knowledge points actually covered by the training scenario, if it is necessary to output knowledge points through operation demonstrations, the limb movements (spatial dimension) and voice annotations (time dimension) during the expert's operation are fused to construct a spatio-temporal coupled unstructured skill database; a graph neural network is used to mine the association rules between the action sequences and the device state changes to form a quantifiable skill subgraph; through hardware-level data format conversion, the interference of different vendor device precision differences on skill evaluation is eliminated to ensure the comparability of heterogeneous device data; a complete conversion chain from limb movements to skill maps and then to teaching standards is realized, breaking through the limitations of traditional text data in skill representation.

[0032] In the embodiments of the present invention, it needs to be further explained that the knowledge network topology map refers to a directed weighted graph of the knowledge network, which consists of a node set, an edge set connecting the nodes, and an edge weight set; each node represents a knowledge point in the knowledge network topology map, and the edge represents the weight between the knowledge points; in the embodiments of the present invention, the weight between the knowledge points represents the migration difficulty coefficient between the two, and the calculation formula for the migration difficulty coefficient is: The index marks of the knowledge points include i and j. The average working hours required to learn knowledge point D_j after mastering knowledge point D_i are denoted as Tavg_ij, and the average working hours required to learn knowledge point D_i after mastering knowledge point D_j are denoted as Tavg_ji. The edge weight corresponding to the edge where the knowledge points i and j are located is calculated through the following formula: Where Indicates the industry benchmark working hours; the quantification of the industry benchmark working hours is achieved by integrating multi-source data of industrial production, educational practice, and industry standards; first, based on the standard operation time, equipment operation records, and employee proficiency curves in the enterprise's real production logs, a basic working hours model is constructed to obtain the industry benchmark working hours; in a possible embodiment, teaching data from vocational colleges (such as the time taken for students to master skills and the effects of teaching interventions) is introduced, and statistical analysis methods are used to divide the industry into three levels: entry, benchmark, and excellent. Then, according to production environment factors such as equipment automation level and material complexity, the corresponding weights of each level are dynamically adjusted, and the weighted average working hours of each level are used to represent the industry benchmark working hours; at the same time, a dynamic feedback mechanism is established. By monitoring the technology update speed, equipment iteration cycle, and students' ability growth curve, the industry benchmark working hours are periodically calibrated to ensure that it not only meets the actual efficiency requirements of the industry but also adapts to the teaching rules.

[0033] Furthermore, dynamically update the industry benchmark working hours and calculate the migration difficulty coefficient.

[0034] In a possible embodiment, the process of obtaining the support resource quality coefficient includes: Step S101: Evaluate the skill knowledge matching degree between the actual knowledge network topology map and the preset knowledge network topology map, which is used to measure the similarity between the actual knowledge network received by students and the preset knowledge network topology map; the way to obtain the skill knowledge matching degree is as follows: compare the actual knowledge network topology map with the preset knowledge network topology map. For each knowledge point in the preset knowledge network topology map, calculate the gap between the actual output and the preset output; consider the influence of each knowledge point (such as its dependence on other knowledge points and importance in the project), weight the gap according to the influence, and obtain the comprehensive learning deviation; analyze the complexity of the preset knowledge network topology map to obtain the complexity value; divide the comprehensive learning deviation by the complexity value and put it into an inverse proportional exponential function (the larger the deviation, the smaller the exponent) to calculate the skill knowledge matching degree; the closer the skill knowledge matching degree is to 1, the closer the knowledge actually received by students is to the requirements of the industry-education integration project; Step S102: The timeliness synchronization rate is used to measure the dynamic matching degree between the preset knowledge network topology map (knowledge structure based on curriculum design) and the actual knowledge network topology map (knowledge structure actually learned by students) in terms of knowledge point update and industrial demand changes; it reflects whether the knowledge points in the curriculum design can timely adapt to the changes in industrial demands (such as the introduction of new technologies and new skills); compare the update frequency and content changes of the knowledge points in the actual knowledge network topology map and the preset knowledge network topology map, and combine the industrial update cycle to quantify the timeliness synchronization rate; Explanation: Define a set of time periods. For each time point, obtain the knowledge point sets of the preset knowledge network topology graph and the actual knowledge network topology graph, calculate the update delay of each knowledge point. The update delay is the difference between the update time of the knowledge point in the preset knowledge network topology graph and the change time of industrial demand; based on the update delay and the reference time period, calculate the synchronization of the knowledge point, which is quantified by an exponential decay function. The smaller the delay, the higher the synchronization; combine and weight according to the importance of the knowledge point in industrial demand, and the importance is determined based on enterprise feedback or job requirements; calculate the timeliness synchronization rate through the weighted average of synchronization and importance, which reflects the dynamic matching degree of the knowledge network topology graph in terms of knowledge point update and industrial demand change; Step S103: Evaluate the operation fit degree between the actual knowledge network topology graph and the preset knowledge network topology graph, which is used to measure the rhythm consistency between the actual knowledge network (the knowledge structure actually learned by students) and the preset knowledge network (the curriculum system design), and reflects whether the teaching execution conforms to the design rhythm. The way to quantify the operation fit degree is as follows: Divide the preset knowledge network topology graph into several curriculum stages in the way with the smallest cumulative migration difficulty coefficient, obtain the corresponding preset knowledge network topology graph and actual knowledge network topology graph for each curriculum stage, compare the differences between the actual knowledge network topology graph and the preset knowledge network topology graph in each curriculum stage, and obtain the node difference and the edge difference; The node difference refers to, for the common knowledge points in the two graphs, comparing the relative deviation of the teaching duration (preset duration and actual duration), and weighting according to the importance of the knowledge point (such as its central position in the curriculum system) to obtain the node deviation; the edge difference refers to, for the common knowledge point relationships in the two graphs, comparing the difference in migration difficulty (preset difficulty and actual difficulty), and weighting according to the flow of the student learning path (such as the frequency of use of this relationship by students) to obtain the edge deviation; Combine the node deviation and the edge deviation (by taking the square root), and calculate the operation fit degree of each curriculum stage in reciprocal form; use the harmonic mean of the deviations of each curriculum stage to represent the operation fit degree. The closer the value is to 1, the more the actual teaching execution conforms to the curriculum design rhythm; Step S104: Jointly analyze the skill knowledge matching degree, the timeliness synchronization rate and the operation fit degree to obtain the support resource quality coefficient.

[0035] In the embodiments of the present invention, it needs to be further explained that when calculating the operation fit degree, calculate the average migration difficulty of the preset knowledge network topology graph and the average migration difficulty of the actual knowledge network topology graph ; Based on the difference between the preset and actual average migration difficulties, generate a migration path efficiency coefficient, which is used to measure the difference between the preset average migration efficiency of the knowledge point and the actual average migration difficulty; If < 1 indicates that the complexity of the actual teaching path significantly deviates from the preset, triggering the dynamic programming algorithm to preferentially adjust the boundary division of the high-migration-difficulty course stage to achieve the global optimization of teaching rhythm and resource supply; When ≥ 1, it indicates that the path efficiency meets the expectations.

[0036] In a possible embodiment, the process of obtaining the support resource quality coefficient includes: represents the learning deviation between the actual knowledge network topology map and the preset knowledge network topology map, represents the complexity value of the preset knowledge network topology map, and the skill knowledge matching degree is calculated by the formula: represents the node difference between the actual knowledge network topology map and the preset knowledge network topology map, represents the edge difference between the actual knowledge network topology map and the preset knowledge network topology map. Let k represent the course stage index and N represent the number of course stages. The operation fitness of the k-th course stage is denoted as ; The operation fitness is calculated by the formula: The support resource quality coefficient is calculated by the formula: the weighted sum of the skill knowledge matching degree, the timeliness synchronization rate, and the operation fitness.

[0037] Furthermore, the triggering condition of step three is any of the following. The present invention does not make specific limitations on this and selects according to the actual situation: 1) When the support resource quality coefficient is lower than the preset threshold, dynamic optimization is immediately triggered; through the real-time monitoring of the support resource quality coefficient, it is detected that the enterprise's support resource supply is insufficient, and optimization is immediately triggered to shorten the resource imbalance window period and prevent the teaching process from deviating from the preset path due to the continuous accumulation of resource defects; 2) Trigger optimization when a single course stage of the industry-education integration project is completed or a fixed time period (such as every 30 days) is reached; Furthermore, for the high-migration-difficulty stage (such as (corresponding to the course stage of ≥1.2), after the end of the stage, prioritize optimizing the subsequent course boundary division, recalculate the minimum cumulative migration path through the dynamic programming algorithm, and reduce the fluctuation of the cognitive leap gradient; Triggered at fixed intervals: Set the calibration period based on the half-life of industry technology, force an update of the industry benchmark working hours, and regenerate the knowledge network topology map to address the issue of the decline in the timeliness synchronization rate caused by the implicit technological iteration of the enterprise (such as the equipment parameter upgrade not being notified to the institution in a timely manner).

[0038] Further, the operation process of step three includes: Step S201: State space modeling Construct a state space, divide the remaining courses into several candidate stage boundary nodes, and each node corresponds to a state, including the cumulative study hours, the support resource quality coefficient, and the preset knowledge network topology map; Set constraint conditions, such as the upper limit of the total study hours and the non-decreasing constraint of the quality coefficient; Define the objective function as maximizing the support resource quality coefficient of the industry-education integration project; Step S202: Course boundary optimization According to the state transition rule, transfer from the previous course stage boundary node to the current node. By adjusting the knowledge point set, migration path, and study hour allocation, ensure that the cumulative study hours do not exceed the upper limit and the support resource quality coefficient meets the preset threshold, and use the backward recursive method to calculate the optimal course stage boundary division sequence; Step S203: Calibration and resource alignment Based on the optimal stage boundary division sequence, dynamically update the nodes, edge weights, and stage markers of the preset knowledge network topology map, and achieve the alignment of support resource supply and teaching needs through enterprise resource reallocation and teaching plan push; Step S204: Closed-loop verification mechanism Monitor the optimized support resource quality coefficient. If it is still lower than the preset value, trigger the dynamic programming algorithm again, and aggregate the optimization data of multiple institutions through the federated learning framework, and periodically calibrate the industry benchmark working hours to improve the adaptability.

[0039] Through backward recursive dynamic programming, iterate from the final course stage to the initial stage, optimize the course stage boundary, adjust the knowledge point set, migration path, and study hours to meet the study hour limit and quality threshold.

[0040] Summary: In the embodiment of the present invention, based on the curriculum stage design of the knowledge network topology diagram, the supporting resources provided by the enterprise (such as training scenarios and teaching materials) and the information actually received by students are integrated to construct a dynamic knowledge network; by analyzing the learning effectiveness of students, the timeliness of knowledge update, and the teaching execution rhythm, the weighted support resource quality coefficient is calculated; then, the matching degree between the coefficient and the preset threshold is judged. If the standard is not met, the warning mechanism is triggered and the resource supply plan is optimized to form a "evaluation - feedback - calibration" closed loop; the process emphasizes the importance of non - profit cooperation goals and balanced resource supply in school - enterprise collaboration. By mapping the topological structure and dynamically adjusting the weights, the problem of dynamically adapting teaching resources to industrial needs is solved, and finally the sustainable development goal of improving education quality and accurately docking with industrial technology needs is achieved.

[0041] Embodiment 2, when industrial technology mutations (such as the version - break upgrade of the numerical control system) cause the preset knowledge network topology diagram to fail frequently, the traditional dynamic programming algorithm causes teaching rhythm oscillations due to the lack of a buffer mechanism; based on this, the embodiment of the present invention includes an optimization step for the curriculum stage based on buffer nodes, including: Technical mutation probability detection: Based on the variational auto - encoder, a probability distribution model is built for the standardized knowledge point set, and the mutation timing of the preset knowledge network topology diagram is adaptively determined through the statistical characteristics (μ + 3σ) of the reconstruction error; μ is the error mean, and σ is the error standard deviation. Buffer node generation and insertion: Extract the standardized knowledge point sets before and after the mutation, and construct a mapping rule library for the old and new technology interfaces. Insert buffer nodes at the mutation points of the preset knowledge network topology diagram. The buffer node is the curriculum stage connecting the old and new technologies, realizing a smooth transition of the technology gap; the buffer node includes: The basic knowledge module of the old technology version (including at least the core instruction set of the old - version numerical control system). The preparatory knowledge module of the new technology version (including at least the description of the differential characteristics of the new - version numerical control system). Cross - version compatibility training levels (debugging tasks that generate mixed - technology instructions based on the interface mapping rules). Dynamic division of the curriculum stage: Dynamically adjust the curriculum stage boundary according to the buffer node distribution density, so that the class hour allocation is elastically matched with the evolution rhythm of industrial technology (such as compressing the basic class hours in the intensive mutation area and enabling the curriculum expansion mode in the non - intensive mutation area: inserting a new technology intensive training stage between buffer nodes).

[0042] Background description: At present, the industry-education integration project of higher vocational colleges is an important way to improve the quality of talent training. By cooperating with enterprises and industry associations, internships, training and other activities are carried out to enhance students' skills; however, the data sources in this process are highly heterogeneous, including enterprise feedback, industry needs, professional qualification certification and teaching materials, etc., and due to differences in regional industrial structures, there are significant differences in evaluation indicators; the existing technology lacks refined data fusion strategies, making it difficult to objectively evaluate teaching quality, and the evaluation system lacks version management and cannot respond to changes in a timely manner. Therefore, building a quality evaluation system that can dynamically adapt to multi-source data, achieve segmented caliber correction, accurate fusion and version traceability has become a key technical problem that needs to be solved in the integration of industry and education. Based on this, Example 3 is set; Embodiment 3: The embodiment of the present invention differs from Embodiment 1 in that it further comprises: Step 4: Collect evaluation data from different ports in each industry-education integration project, summarize the evaluation data from different ports, summarize and standardize them, and output a unified evaluation data set; including: Step 301: constructing a multi-source labeling system. When it is detected that multi-source heterogeneous evaluation data need to be uniformly identified and the original text or numerical file is in a difficult-to-retrieve state and different collection ports, such as enterprise-side feedback and school-side record formats conflict, and the label distribution is highly discrete and there are multi-language annotations or non-standard abbreviations and some fields are temporarily missing, natural language processing and multi-dimensional label aggregation algorithms are used to perform word segmentation, entity recognition and feature annotation on abnormal data, and output a labeled data set; convert the multi-source heterogeneous evaluation data into a labeled data set and change its form from scattered and messy to searchable structured information; Explanation: The original data (including text feedback, numerical indicators, pictures or other unstructured content) is formatted and converted into readable and indexable basic units to form a preprocessed data set; the collected evaluation data is divided into several collection ports according to the data source, industry, time period, enterprise size and other dimensions, and each collection port includes several records; an initial label set is established, and corresponding labels are assigned to each record to form a basic label index table; for text data, business terms, technical terms, etc. are annotated using natural language processing algorithms (such as word segmentation and keyword extraction); for numerical or image data, its meta-information tags (such as shooting time, department to which it belongs) are embedded in the record to ensure retrievability and distinguishability during subsequent analysis; Step 302: When it is detected that the indicator differences of the labeled data set at different collection ports are greater than the set threshold and extreme outlier distribution appears and the scoring rules of the previous and subsequent cycles are obviously inconsistent, the cross-industry scoring deviations are segmented and nonlinearly corrected based on the dynamic correction function and combined with the caliber correction parameter table, so as to generate a unified caliber data set and realize quantifiable traceable difference correction.

[0043] Explanation: After the construction of the multi-source label system, the metric caliber of the labeled dataset is corrected and consistency processed; the goal is to uniformly adjust the value or weight of the evaluation metrics under different industries and different time axes so that they can be fairly compared and aggregated in the subsequent multi-source fusion stage; align the metrics with the same meaning but different forms in the multi-source data in advance to avoid incomparable phenomena during subsequent caliber correction or fusion; quantitatively identify potential abnormal records in the evaluation data through the recorded consistency coefficient analysis model, reduce the possibility of subjective misjudgment, and ensure the credibility of the finally unified caliber dataset.

[0044] In the embodiment of the present invention, it needs to be further explained that the value range of the recorded consistency coefficient is 0 to 1, and the recorded consistency coefficient satisfies the following formula: where the record is a single piece of data in the labeled dataset; is the weighted median calculated for ri; Example: If is a record of the "operation accuracy of numerical control machine tools" of a higher vocational student, is the 90th percentile of the operation accuracy of excellent students in the student's major over the years; represents the reference benchmark value, which can come from the aggregated statistics or manual calibration of the same industry; and are the attenuation coefficient and the exponential factor respectively, used to regulate the sensitivity to the attenuation rate of the recorded consistency coefficient.

[0045] In the embodiment of the present invention, it needs to be further explained that the difference correction process includes: Step S401: For the differences in name, unit, or measurement method of the same indicator at different collection ports (different time dimensions or subject dimensions); based on the field mapping table, independent correction functions are configured for different collection ports to obtain consistency-optimized indicators; the caliber correction parameter table is used to ensure that the field names and units are uniformly converted before subsequent correction operations; for example, if it is found in the "labeled dataset" that the same evaluation indicator "job adaptability" is in a 0-100 point system in some enterprises and is marked as an A / B / C grading system in some industries, then logical unification needs to be carried out with the help of the "field mapping table" to generate a preliminary unified indicator set; Step 402: Obtain the comprehensive deviation term matrix of each indicator at different collection ports; Based on the indicator comprehensive deviation term matrix, a dynamic correction function is constructed to generate a unified caliber dataset and achieve quantifiable traceable difference correction; the implementation steps of constructing the comprehensive deviation term matrix include: Define the reference benchmark: Select an authoritative data source or historical mean as the benchmark value (such as the industry average score); Calculating the deviation term: For each metric of each acquisition port, calculate the deviation from the reference value, including the numerical deviation sp and the distribution deviation fp; after normalizing the numerical deviation sp and the distribution deviation fp, obtain the comprehensive deviation factor through weighted summation, and denote the comprehensive deviation factor of the j-th metric of the i-th port as ; Taking the sequential number of the metric as the index, denote the deviation of the j-th metric of the i-th port as (sp, fp), and obtain the comprehensive deviation term matrix; It needs to be further explained in the embodiments of the present invention that the dynamic correction function satisfies the following formula: Wherein, represents the calibrated value of the j-th metric of the i-th port, represents the corresponding reference value, represents the original value of the j-th metric of the i-th port; represents the global attenuation factor, β represents the non-linear sensitivity parameter. When β > 1, it is sensitive to high deviation values; when β < 1, smoothing correction is performed; Step S403: Based on the calculation result of the dynamic correction function, perform piecewise transformation on all metric values in the consistency optimization metric set, and finally obtain a unified caliber dataset after correcting the outlier.

[0046] It needs to be further explained in the embodiments of the present invention that based on the evaluation dataset with the final unified caliber, the integration evaluation coefficient of industry and education is comprehensively evaluated. If the integration evaluation coefficient of industry and education is lower than the threshold, it indicates that the experience effect of the integration of industry and education is lower than expected.

[0047] Finally: The above is only the preferred embodiment of the present invention and is not used to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for constructing a quality evaluation system for higher vocational colleges based on the integration of industry and education, characterized in that, Including: Step 1: Based on the preset requirements of the industry-education integration project, construct a preset knowledge network topology diagram; Set the course stages of the industry-education integration project based on the preset knowledge network topology diagram, collect the support resource situation of the enterprise in the industry-education integration project, and construct an actual knowledge network topology diagram based on the actual received information of the students in the industry-education integration project; The support resources include teaching materials and training scenarios; Step 2: By comparing the topological structures, semantic contents, and the degree of fit at each course stage of the preset and actual knowledge network topology diagrams, obtain the skill knowledge matching degree, timeliness synchronization rate, and operation fit degree, and obtain the support resource quality coefficient through weighted summation; If the support resource quality coefficient is lower than the preset value, it indicates that in the industry-education integration project, the quality of the support resources provided by the enterprise is lower than the preset requirements, and an early warning is sent out to prompt to improve the quality of the support resources; If the support resource quality coefficient is not lower than the preset value, it indicates that the quality of the support resources provided by the enterprise meets the requirements and no adjustment is needed; Step 3: Implement dynamic optimization and feedback calibration according to the support resource quality coefficient. When the support resource quality coefficient is lower than the preset value, take maximizing the support resource quality coefficient as the objective function, with the total class hours not exceeding the preset upper limit, and re-divide the course stage boundary based on the dynamic programming algorithm; Generate a calibrated preset knowledge network topology diagram in real time to achieve the adaptive alignment of resource supply and teaching needs.

2. The method for constructing a quality evaluation system for higher vocational colleges based on the integration of production and education according to claim 1, characterized in that, Parse the enterprise job requirement documents, industry standard files, and enterprise technical logs through natural language processing technology, extract the standardized knowledge point set and eliminate the ambiguity of synonyms; Based on the expert scoring method, determine the logical dependency relationship between knowledge points, combine the enterprise engineer interview data to quantify the knowledge point migration difficulty weight, and construct a preset knowledge network topology diagram with knowledge points as nodes, dependency relationships as directed edges, and migration difficulty as edge weights.

3. The method for constructing a quality evaluation system for higher vocational colleges based on the integration of production and education according to claim 2, characterized in that, By collecting the actual teaching materials of enterprise tutors and analyzing the knowledge points actually covered by the output teaching materials; by building a digital twin model of the student training scenario, obtain the knowledge points actually covered by the training scenario; combine the knowledge points actually covered by the teaching materials and the training scenario to obtain the actual knowledge network topology diagram; The digital twin model standardizes the data flow format of heterogeneous devices through the hardware abstraction layer, uses Fourier time-frequency transformation to eliminate the sensor sampling rate difference, and establishes a skill evaluation benchmark coordinate system for cross-vendor devices.

4. The method for constructing a quality evaluation system for higher vocational colleges based on the integration of production and education according to claim 2, characterized in that, The weight between knowledge points represents the migration difficulty coefficient between the two, and the calculation formula for the migration difficulty coefficient is: The index marks of knowledge points include i and j. Denote the average working hours required to learn knowledge point D_j after mastering knowledge point D_i as Tavg_ij, and denote the average working hours required to learn knowledge point D_i after mastering knowledge point D_j as Tavg_ji. The edge weight corresponding to the edge where knowledge points i and j are located is calculated through the following formula: Among them, represents the industry benchmark working hours; based on the standard operation time, equipment operation records, and employee proficiency curves in the enterprise's real production logs, a basic working hours model is constructed to obtain the industry benchmark working hours.

5. The method for constructing a quality evaluation system for higher vocational colleges based on the integration of production and education according to claim 3, characterized in that, The process of obtaining the support resource quality coefficient includes: Step S101: Evaluate the skill knowledge matching degree between the actual knowledge network topology diagram and the preset knowledge network topology diagram. The quantification method is as follows: Compare the actual knowledge network topology diagram with the preset knowledge network topology diagram. For each knowledge point in the preset knowledge network topology diagram, calculate the gap between the actual output and the preset output; consider the influence of each knowledge point, weight the gap according to the influence, and obtain the comprehensive learning deviation; analyze the complexity of the preset knowledge network topology diagram to obtain the complexity value; divide the comprehensive learning deviation by the complexity value, put it into the inverse proportional exponential function, and calculate the skill knowledge matching degree. Step S102: The timeliness synchronization rate is used to measure the dynamic matching degree between the preset knowledge network topology diagram and the actual knowledge network topology diagram in terms of knowledge point update and industrial demand change; it reflects whether the knowledge points in the curriculum design can adapt to the changes in industrial demands in a timely manner; compare the update frequency and content changes of the knowledge points in the actual knowledge network topology diagram and the preset knowledge network topology diagram, and combine the industrial update cycle to quantify the timeliness synchronization rate. Step S103: Evaluate the operation fit degree between the actual knowledge network topology diagram and the preset knowledge network topology diagram, which is used to measure the rhythm consistency between the actual knowledge network and the preset knowledge network. The quantification method is as follows: Divide the preset knowledge network topology diagram into several curriculum stages in the way with the minimum cumulative migration difficulty coefficient, obtain the corresponding preset knowledge network topology diagram and actual knowledge network topology diagram for each curriculum stage, compare the differences between the actual knowledge network topology diagram and the preset knowledge network topology diagram for each curriculum stage to obtain the node difference and edge difference; combine the node deviation and the edge deviation, and calculate the operation fit degree for each curriculum stage in the form of a reciprocal; represent the operation fit degree with the harmonic mean of the deviations of each curriculum stage. Step S104: Jointly analyze the skill knowledge matching degree, the timeliness synchronization rate and the operation fit degree to obtain the support resource quality coefficient.

6. The method for constructing a quality evaluation system for higher vocational colleges based on the integration of production and education according to claim 4, wherein, Calculate the average migration difficulty of the preset knowledge network topology diagram and the average migration difficulty of the actual knowledge network topology diagram ; Generate a migration path efficiency coefficient based on the difference between the preset and actual average migration difficulties, which is used to measure the difference between the preset average migration efficiency of knowledge points and the actual average migration difficulty. If < 1 indicates that the complexity of the actual teaching path significantly deviates from the preset, triggering the dynamic programming algorithm to preferentially adjust the boundary division of the high-transfer-difficulty course stage to achieve global optimization of the teaching rhythm and resource supply; when ≥ 1, it indicates that the path efficiency meets the expectations.

7. The method for constructing a quality evaluation system for higher vocational colleges based on the integration of production and education according to claim 1, characterized in that, Include the curriculum stage optimization steps based on buffer nodes, including: Technical mutation probability detection: Based on the variational autoencoder, perform probability distribution modeling on the standardized knowledge point set, and adaptively determine the mutation timing of the preset knowledge network topology diagram through the statistical characteristics of the reconstruction error. Buffer node generation and insertion: Extract the standardized knowledge point sets before and after the mutation, and construct a mapping rule library for the old and new technology interfaces. Insert buffer nodes at the mutation points of the preset knowledge network topology diagram. The buffer nodes are the curriculum stages connecting the old and new technologies to achieve a smooth transition of the technical gap. Dynamic division of curriculum stages: Dynamically adjust the curriculum stage boundaries according to the buffer node distribution density to make the class hour allocation match the rhythm of industrial technology evolution elastically.

8. The method for constructing a quality evaluation system for higher vocational colleges based on the integration of production and education according to claim 1, characterized in that, The triggering condition of Step 3 is any of the following: When the support resource quality coefficient is lower than the preset threshold, immediately trigger dynamic optimization; trigger optimization when a single curriculum stage of the industry-education integration project is completed or a fixed time period is reached.

9. The method for constructing a quality evaluation system for higher vocational colleges based on the integration of production and education according to claim 1, characterized in that, The running process of the said Step 3 includes: Construct the state space, divide the curriculum into several candidate stage boundary nodes, where each node corresponds to the cumulative class hours, the quality coefficient of supporting resources, and the preset knowledge network topology diagram. Set the upper limit of the total class hours and the non-decreasing constraint of the quality coefficient, and aim to maximize the quality coefficient of supporting resources; According to the state transition rule, transfer from the previous curriculum stage boundary node to the current node. By adjusting the knowledge point set, migration path, and class hour allocation, ensure that the cumulative class hours do not exceed the upper limit and the quality coefficient of supporting resources meets the preset threshold, and use the reverse recursive method to calculate the optimal curriculum stage boundary division sequence; Based on the optimal stage boundary division sequence, dynamically update the nodes, edge weights, and stage marks of the preset knowledge network topology diagram, and align the supply of supporting resources with the teaching needs through enterprise resource reallocation and teaching plan push; Monitor the quality coefficient of the optimized supporting resources. If it is still lower than the preset value, trigger the dynamic programming algorithm again, and aggregate the optimization data of multiple institutions through the federated learning framework, and periodically calibrate the industry benchmark working hours to improve the adaptability.

10. The method for constructing a quality evaluation system for higher vocational colleges based on the integration of production and education according to any one of claims 1-9, characterized in that, The method further includes: Step 4: Collect the evaluation data of different ports in each industry-education integration project, summarize the evaluation data of different ports, and output a unified-caliber evaluation data set after summarization and standardization, including: Multi-source heterogeneous data tagging processing: Collect the evaluation data of different ports in the industry-education integration project, and use natural language processing and multi-dimensional label aggregation algorithms to perform word segmentation analysis, entity recognition, and feature annotation on multi-source heterogeneous data to generate a retrievable tagged data set; Unification of cross-port index calibers: Based on the field mapping table, perform logical conversion on the differences in the names, units, or measurement methods of the same index in different collection ports in the tagged data set to generate a preliminary unified index set; Calculation of the comprehensive deviation factor: For each index in the preliminary unified index set, calculate the numerical deviation and distribution deviation from the reference benchmark value, and generate a comprehensive deviation factor matrix after normalization and weighted summation; Dynamic correction and data set generation: Based on the comprehensive deviation factor matrix, perform piecewise correction on the index values through a dynamic correction function regulated by a non-linear sensitivity parameter to generate a unified-caliber data set.

Images