Intelligent higher vocational mathematics teaching decision and evaluation system fused with AI adaptive algorithm

Abstract

The invention provides a higher vocational mathematics intelligent teaching decision and evaluation system fused with an AI adaptive algorithm, which relates to the field of electrical digital data processing and comprises an intelligent resource development module, an adaptive teaching decision module, a multi-dimensional data acquisition module and a comprehensive evaluation optimization module. The intelligent resource development module is responsible for constructing a dynamic mathematical teaching resource library, the self-adaptive teaching decision module dynamically plans a personalized learning path based on an AI algorithm, and the multi-dimensional data acquisition module comprehensively acquires behavior data, result data and feedback data in a teaching process. The comprehensive evaluation optimization module generates a capability map through data analysis; according to the system, intelligent closed-loop management of teaching resources, learning paths and an evaluation system is realized through innovative algorithms such as comprehensive learning condition characteristic value calculation, task income value evaluation and capability index quantification, and the individuation level and the childbearing quality of higher vocational mathematics teaching are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of electronic digital data processing, specifically to an intelligent teaching decision-making and evaluation system for higher vocational mathematics that integrates AI adaptive algorithms. Background Technology

[0002] Vocational college mathematics teaching has long faced prominent challenges such as significant differences in students' foundational knowledge, severe disparities in learning abilities, and the inability of traditional teaching models to cater to individualized needs. With the rapid development of artificial intelligence technology, intelligent teaching systems are gradually becoming an important approach to solving these problems. Intelligent teaching systems, by collecting student learning data, constructing learner models, and dynamically adjusting teaching strategies, achieve a shift from standardized to personalized teaching. Adaptive learning, as a core technology of intelligent education, can dynamically adjust teaching resources and learning paths based on learners' cognitive levels, learning abilities, and knowledge acquisition, thereby improving learning efficiency.

[0003] In the field of adaptive learning, learning path planning algorithms are a key technology for achieving personalized instruction. Currently, there are three main representative algorithms. The first is a rule-based optimization ranking algorithm. This algorithm monitors students' accuracy and response time, assigns priority scores to each question, and dynamically adjusts the presentation order of questions according to preset rules. This algorithm is simple to implement and computationally efficient, but its fixed rules make it difficult to handle complex learning situations and it cannot fully consider the relationships between knowledge points and the multidimensional characteristics of students' abilities. The second is a knowledge graph-based path search algorithm. This algorithm constructs a directed graph of the knowledge space, explicitly models knowledge points and their dependencies, and searches for the optimal learning path in the knowledge graph based on the student's current mastery. This method can better represent the knowledge structure, but its path planning mainly relies on graph search algorithms, lacking the ability to adapt to the dynamic changes in individual students' learning processes and struggling to balance the dual goals of knowledge acquisition and ability development. The third is a collaborative filtering-based recommendation algorithm. This algorithm analyzes the learning trajectories and performance data of similar learners to recommend learning resources and paths to target students. This method can leverage group learning experience, but it suffers from the cold start problem, is less effective in recommending new students or resources, and the recommendation results often lack educational theoretical support, making it difficult to explain the rationality of the recommendations.

[0004] In the relevant technical field abroad, US Patent 7052277B2 discloses an adaptive learning system and method. This patent proposes an adaptive learning technology based on question ranking. The system dynamically modifies the presentation order of subsequent questions by continuously monitoring the speed and accuracy of students' answers. The system uses an optimal ranking algorithm to assign a recurrence priority to each learning item, determining the order of subsequent learning questions based on the student's performance on specific questions. The shortcomings of this patent are that its adaptive mechanism mainly focuses on optimizing the question sequence, lacking deep modeling of learners' ability dimensions, failing to construct a correlation system between knowledge points and professional skills, and its algorithm design does not consider cognitive factors such as forgetting curves and learning cycle fluctuations, resulting in a lack of predictive ability for long-term learning outcomes in learning path planning. Furthermore, the system lacks a dynamic updating mechanism for teaching resources, making it unable to adapt to rapidly changing professional needs and technological iterations, thus limiting its applicability in vocational education scenarios.

[0005] In the relevant technical field in China, Chinese patent CN107507468A discloses an adaptive learning method and system based on directed graphs in a knowledge space. This patent proposes a system architecture including a knowledge recognition device, an ability recognition device, an adaptive device, and an ability tracking device. The knowledge recognition device and the ability recognition device are connected through knowledge vectors, and connections are also established through ability vectors. This patent emphasizes that learning is a process of establishing associations between existing knowledge and abilities and new knowledge, revealing a three-layered interactive structure of knowledge relationships, ability-knowledge relationships, and relationships between abilities. The patent's shortcomings lie in its overly theoretical system architecture, lacking specific algorithmic implementation details, particularly in the quantitative modeling of learning progress diagnosis and path planning. The system does not provide a clear learning path benefit evaluation model, failing to effectively balance the dual goals of knowledge acquisition and ability development. Furthermore, its ability tracking mechanism lacks in-depth data mining of the learning process, making precise dynamic adjustments difficult. In addition, the patent does not involve the fusion technology of multi-dimensional evaluation data, failing to integrate feedback from teachers, peers, and enterprises, resulting in insufficient objectivity and comprehensiveness of the learning evaluation system. In the context of industry-education integration in vocational education, it is difficult to effectively support the optimization and iteration of teaching decisions. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings by proposing an intelligent teaching decision-making and evaluation system for higher vocational mathematics that integrates AI adaptive algorithms.

[0007] The present invention adopts the following technical solution:

[0008] A smart teaching decision-making and evaluation system for higher vocational mathematics that integrates AI adaptive algorithms includes an intelligent resource development module, an adaptive teaching decision-making module, a multi-dimensional data acquisition module, and a comprehensive evaluation optimization module;

[0009] The intelligent resource development module is responsible for building a dynamic mathematics teaching resource library; the adaptive teaching decision module dynamically plans personalized learning paths based on AI algorithms; the multi-dimensional data acquisition module comprehensively acquires behavioral data, outcome data, and feedback data during the teaching process; and the comprehensive evaluation and optimization module generates a capability map through data analysis.

[0010] The intelligent resource development module includes an industry-education integration resource library unit, a knowledge graph construction unit, and a resource dynamic update unit. The industry-education integration resource library unit is responsible for the collaborative development of industry case libraries by schools and enterprises, transforming real professional scenarios into mathematics teaching materials. The knowledge graph construction unit uses digital intelligence technology to sort out the relationship between mathematical knowledge points and professional abilities, and establish a structured knowledge network system. The resource dynamic update unit establishes an automatic update mechanism for the resource library based on changes in enterprise needs and technological iterations.

[0011] The adaptive teaching decision-making module includes a learning situation intelligent diagnosis unit, a learning path planning unit, and a teaching strategy recommendation unit. The learning situation intelligent diagnosis unit is used to analyze students' knowledge mastery level, learning style, and ability characteristics to generate personalized learning situation profiles. The learning path planning unit dynamically adjusts task difficulty, content sequence, and learning pace based on the learning situation diagnosis results. The teaching strategy recommendation unit provides teachers with differentiated teaching method suggestions.

[0012] The multi-dimensional data acquisition module includes a classroom behavior monitoring unit, a project achievement recording unit, and a multi-source feedback integration unit. The classroom behavior monitoring unit collects students' online learning behavior data in real time through an intelligent teaching platform. The project achievement recording unit systematically stores students' works, reports, and practical results generated in inquiry tasks and collaborative projects, establishing a complete learning archive. The multi-source feedback integration unit gathers multi-dimensional evaluation information from student self-evaluation, peer evaluation, teacher evaluation, and enterprise feedback.

[0013] The comprehensive evaluation and optimization module includes a capability map generation unit, a teaching effectiveness analysis unit, and an iterative optimization decision-making unit. The capability map generation unit constructs a digital and intelligent capability map of students based on collected multidimensional data and big data analysis technology. The teaching effectiveness analysis unit quantitatively evaluates the teaching process and results, identifies the strengths and weaknesses in teaching, and the iterative optimization decision-making unit feeds back the evaluation results to the resource development and teaching decision-making stages, driving the continuous improvement and intelligent upgrading of the teaching system.

[0014] Furthermore, the intelligent learning assessment unit includes a data preprocessor, a feature extraction processor, and a profile generation processor. The data preprocessor cleans, standardizes, and processes missing values ​​in the collected student learning data to construct a structured analysis dataset. The feature extraction processor extracts key features from the learning behavior data to generate multi-dimensional feature vectors. The profile generation processor integrates the multi-dimensional feature data to generate a personalized learning profile for each student.

[0015] The profile generation processor generates the student's comprehensive learning characteristic value F(t) at time t according to the following formula:

[0016] ;

[0017] Where D represents the total number of mathematical knowledge points. Let K be the weight of the d-th knowledge point. d (t) represents the mastery of the d-th knowledge point at time t, where t d This indicates the time when the student last studied the d-th knowledge point. Let V be the forgetting time constant for the d-th knowledge point. d This represents the importance of the d-th knowledge point to the student's target career position, where J is the total number of learning behavior dimensions. As the weight of learning behavior, A j (t) represents the student's activity level in the j-th learning behavior dimension, P j Let be the occupational relevance coefficient for the j-th behavioral dimension. Let C(t) represent the student's comprehensive ability index at time t, where C(t) is the ability weight. T is the learning cycle fluctuation coefficient. cyc This is the learning cycle.

[0018] Furthermore, the learning path planning unit includes a goal setting processor, a path algorithm processor, and a dynamic adjustment processor. The goal setting processor sets personalized, phased learning goals and ability achievement standards for each student based on the course teaching objectives, student learning profiles, and professional ability requirements. The path algorithm processor uses an AI adaptive algorithm to plan the optimal learning path from the current level to the target state. The dynamic adjustment processor monitors learning progress in real time and automatically adjusts subsequent learning paths based on task completion quality and ability improvement.

[0019] The path algorithm processor calculates the comprehensive benefit value Q(st, k) of the student choosing learning task k in state st according to the following formula:

[0020] ;

[0021] Among them, R K (st k) represents state s t Knowledge layer benefit of task k, R C (s t k) represents state s t The ability layer benefit of task k, w K and w C For the two-layer objective weights, D k Let k be the capability requirement value. X is the fitness sensitivity coefficient, and X is the nonlinear adjustment index. As a discount factor, This is a path consistency penalty function.

[0022] Furthermore, the capability map generation unit includes a capability index calculation processor, a map modeling processor, and a visualization rendering processor. The capability index calculation processor calculates the quantitative index values ​​of students in multiple capability dimensions based on the results of multi-dimensional data collection. The map modeling processor is used to construct a multi-dimensional capability assessment model to depict the students' capability strengths and weaknesses. The visualization rendering processor presents the capability data in an intuitive graphical way and generates personalized capability development reports.

[0023] Furthermore, the ability index calculation processor calculates the student's comprehensive ability value E in the m-th ability dimension according to the following formula. m (t):

[0024] ;

[0025] Where Smp is the score of the p-th data source for the m-th capability dimension, and P is the total number of data sources. Let p be the weight coefficient of the p-th data source. For the scoring uncertainty, For uncertain adjustment parameters, As an outcome-based evaluation index, For the incentive coefficient of progress, Perform a quality increment on the student's path. The average increment for all students For the group standard deviation, This serves as a verification factor for the consistency of process results.

[0026] The beneficial effects achieved by this invention are:

[0027] This system generates accurate, personalized learning profiles by integrating multiple dimensions such as knowledge mastery, forgetting curves, learning behavior activity, career relevance, and learning cycle fluctuations through a comprehensive learning characteristic value calculation formula. This overcomes the limitations of traditional learning analysis, which only focuses on knowledge mastery, and achieves a comprehensive depiction of students' cognitive states. The task-based comprehensive benefit value calculation model adopted in this module introduces dual-layer target weights for knowledge-level and ability-level benefits, combined with adaptive sensitivity adjustment and path coherence penalty mechanisms. This allows for simultaneous optimization of both knowledge mastery and ability development during dynamic learning path planning. Furthermore, the iterative update mechanism of reinforcement learning gradually converges the Q-value table to the optimal state, ensuring the scientific and efficient nature of the learning path. Compared to existing rule-based fixed algorithms and simple knowledge graph search methods, this significantly improves the adaptive capability of path planning and educational effectiveness. By innovatively introducing a formula for calculating the comprehensive ability value, the system incorporates a scoring uncertainty adjustment mechanism, a progress incentive mechanism, and a process-result consistency verification factor. This approach considers the reliability differences of evaluations from multiple data sources, avoids the one-sidedness of simple horizontal comparisons by comparing students' own progress longitudinally, and ensures the objectivity and fairness of the evaluation by verifying the consistency between process evaluation and outcome evaluation. This effectively solves the problems of strong subjectivity, single dimension, and insufficient incentives in traditional evaluation methods.

[0028] To further understand the features and technical content of the present invention, please refer to the following detailed description and drawings of the present invention. However, the drawings provided are for reference and illustration only and are not intended to limit the present invention. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of the overall structural framework of the present invention;

[0030] Figure 2 This is a schematic diagram of the intelligent resource development module of the present invention;

[0031] Figure 3 This is a schematic diagram of the adaptive teaching decision module of the present invention;

[0032] Figure 4 This is a schematic diagram of the multidimensional data acquisition module of the present invention;

[0033] Figure 5 This is a schematic diagram of the comprehensive evaluation and optimization module of the present invention;

[0034] Figure 6 This is a schematic diagram comparing the core indicators of the present invention with those of the control group;

[0035] Figure 7 This is a schematic diagram comparing the improvement in abilities of students at different levels in this invention and the control group. Detailed Implementation

[0036] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can understand the advantages and effects of the present invention from the content disclosed in this specification. The present invention can be implemented or applied through other different specific embodiments, and various details in this specification can also be modified and changed based on different viewpoints and applications without departing from the spirit of the present invention. Furthermore, the accompanying drawings of the present invention are for simple illustrative purposes only and are not depictions of actual dimensions; this is stated beforehand. The following embodiments will further describe the relevant technical content of the present invention in detail, but the disclosed content is not intended to limit the scope of protection of the present invention.

[0037] Example 1.

[0038] This embodiment provides an intelligent teaching decision-making and evaluation system for higher vocational mathematics that integrates AI adaptive algorithms. Figure 1 It includes an intelligent resource development module, an adaptive teaching decision-making module, a multi-dimensional data acquisition module, and a comprehensive evaluation and optimization module;

[0039] The intelligent resource development module is responsible for building a dynamic mathematics teaching resource library; the adaptive teaching decision module dynamically plans personalized learning paths based on AI algorithms; the multi-dimensional data acquisition module comprehensively acquires behavioral data, outcome data, and feedback data during the teaching process; and the comprehensive evaluation and optimization module generates a capability map through data analysis.

[0040] The intelligent resource development module includes an industry-education integration resource library unit, a knowledge graph construction unit, and a resource dynamic update unit. The industry-education integration resource library unit is responsible for the collaborative development of industry case libraries by schools and enterprises, transforming real professional scenarios into mathematics teaching materials. The knowledge graph construction unit uses digital intelligence technology to sort out the relationship between mathematical knowledge points and professional abilities, and establish a structured knowledge network system. The resource dynamic update unit establishes an automatic update mechanism for the resource library based on changes in enterprise needs and technological iterations.

[0041] The adaptive teaching decision-making module includes a learning situation intelligent diagnosis unit, a learning path planning unit, and a teaching strategy recommendation unit. The learning situation intelligent diagnosis unit is used to analyze students' knowledge mastery level, learning style, and ability characteristics to generate personalized learning situation profiles. The learning path planning unit dynamically adjusts task difficulty, content sequence, and learning pace based on the learning situation diagnosis results. The teaching strategy recommendation unit provides teachers with differentiated teaching method suggestions.

[0042] The multi-dimensional data acquisition module includes a classroom behavior monitoring unit, a project achievement recording unit, and a multi-source feedback integration unit. The classroom behavior monitoring unit collects students' online learning behavior data in real time through an intelligent teaching platform. The project achievement recording unit systematically stores students' works, reports, and practical results generated in inquiry tasks and collaborative projects, establishing a complete learning archive. The multi-source feedback integration unit gathers multi-dimensional evaluation information from student self-evaluation, peer evaluation, teacher evaluation, and enterprise feedback.

[0043] The comprehensive evaluation and optimization module includes a capability map generation unit, a teaching effectiveness analysis unit, and an iterative optimization decision-making unit. The capability map generation unit constructs a digital and intelligent capability map of students based on collected multidimensional data and big data analysis technology. The teaching effectiveness analysis unit quantitatively evaluates the teaching process and results, identifies the strengths and weaknesses in teaching, and the iterative optimization decision-making unit feeds back the evaluation results to the resource development and teaching decision-making stages, driving the continuous improvement and intelligent upgrading of the teaching system.

[0044] The intelligent learning diagnosis unit includes a data preprocessor, a feature extraction processor, and a profile generation processor. The data preprocessor cleans, standardizes, and processes missing values ​​in the collected student learning data to construct a structured analysis dataset. The feature extraction processor extracts key features from the learning behavior data to generate multi-dimensional feature vectors. The profile generation processor integrates the multi-dimensional feature data to generate a personalized learning profile for each student.

[0045] The profile generation processor generates the student's comprehensive learning characteristic value F(t) at time t according to the following formula:

[0046] ;

[0047] Where D represents the total number of mathematical knowledge points. Let K be the weight of the d-th knowledge point. d (t) represents the mastery of the d-th knowledge point at time t, where t d This indicates the time when the student last studied the d-th knowledge point. Let V be the forgetting time constant for the d-th knowledge point. d This represents the importance of the d-th knowledge point to the student's target career position, where J is the total number of learning behavior dimensions. As the weight of learning behavior, A j (t) represents the student's activity level in the j-th learning behavior dimension, P j Let be the occupational relevance coefficient for the j-th behavioral dimension. Let C(t) represent the student's comprehensive ability index at time t, where C(t) is the ability weight. T is the learning cycle fluctuation coefficient.cyc This is the learning cycle.

[0048] The learning path planning unit includes a goal setting processor, a path algorithm processor, and a dynamic adjustment processor. The goal setting processor sets personalized, phased learning goals and ability achievement standards for each student based on the course teaching objectives, student learning profiles, and professional ability requirements. The path algorithm processor uses an AI adaptive algorithm to plan the optimal learning path from the current level to the target state. The dynamic adjustment processor monitors learning progress in real time and automatically adjusts subsequent learning paths based on task completion quality and ability improvement.

[0049] The path algorithm processor calculates the comprehensive benefit value Q(st, k) of the student choosing learning task k in state st according to the following formula:

[0050] ;

[0051] Among them, R K (s t k) represents state s t Knowledge layer benefit of task k, R C (s t k) represents state s t The ability layer benefit of task k, w K and w C For the two-layer objective weights, D k Let k be the capability requirement value. X is the fitness sensitivity coefficient, and X is the nonlinear adjustment index. As a discount factor, This is a path consistency penalty function.

[0052] The competency graph generation unit includes a competency index calculation processor, a graph modeling processor, and a visualization rendering processor. The competency index calculation processor calculates quantitative index values ​​for students in multiple competency dimensions based on multidimensional data collection results. The graph modeling processor is used to construct a multidimensional competency assessment model to depict students' competency strengths and weaknesses. The visualization rendering processor presents competency data in an intuitive graphical way and generates personalized competency development reports.

[0053] The ability index calculation processor calculates the student's comprehensive ability value E in the m-th ability dimension according to the following formula. m (t):

[0054] ;

[0055] Where Smp is the score of the p-th data source for the m-th capability dimension, and P is the total number of data sources. Let p be the weight coefficient of the p-th data source. For the scoring uncertainty, For uncertain adjustment parameters, As an outcome-based evaluation index, For the incentive coefficient of progress, Perform a quality increment on the student's path. The average increment for all students For the group standard deviation, This serves as a verification factor for the consistency of process results.

[0056] Example 2.

[0057] This embodiment includes all the contents of Embodiment 1, and provides an intelligent teaching decision and evaluation system for higher vocational mathematics that integrates AI adaptive algorithms, including an intelligent resource development module, an adaptive teaching decision module, a multi-dimensional data acquisition module, and a comprehensive evaluation optimization module;

[0058] The intelligent resource development module is responsible for building a dynamic mathematics teaching resource library; the adaptive teaching decision module dynamically plans personalized learning paths based on AI algorithms; the multi-dimensional data acquisition module comprehensively acquires behavioral data, outcome data, and feedback data during the teaching process; and the comprehensive evaluation and optimization module generates a capability map through data analysis.

[0059] Combination Figure 2 The intelligent resource development module includes an industry-education integration resource library unit, a knowledge graph construction unit, and a resource dynamic update unit. The industry-education integration resource library unit is responsible for the collaborative development of industry case libraries by schools and enterprises, transforming real professional scenarios into mathematics teaching materials. The knowledge graph construction unit uses digital intelligence technology to sort out the relationship between mathematical knowledge points and professional abilities, and establishes a structured knowledge network system. The resource dynamic update unit establishes an automatic update mechanism for the resource library based on changes in enterprise needs and technological iterations.

[0060] Combination Figure 3 The adaptive teaching decision module includes a learning situation intelligent diagnosis unit, a learning path planning unit, and a teaching strategy recommendation unit. The learning situation intelligent diagnosis unit is used to analyze students' knowledge mastery level, learning style, and ability characteristics to generate personalized learning situation profiles. The learning path planning unit dynamically adjusts task difficulty, content sequence, and learning pace based on the learning situation diagnosis results. The teaching strategy recommendation unit provides teachers with differentiated teaching method suggestions.

[0061] Combination Figure 4The multi-dimensional data acquisition module includes a classroom behavior monitoring unit, a project achievement recording unit, and a multi-source feedback integration unit. The classroom behavior monitoring unit collects students' online learning behavior data in real time through an intelligent teaching platform. The project achievement recording unit systematically stores students' works, reports, and practical results generated in inquiry tasks and collaborative projects, establishing a complete learning archive. The multi-source feedback integration unit gathers multi-dimensional evaluation information from student self-evaluation, peer evaluation, teacher evaluation, and enterprise feedback.

[0062] Combination Figure 5 The comprehensive evaluation and optimization module includes a capability map generation unit, a teaching effectiveness analysis unit, and an iterative optimization decision-making unit. The capability map generation unit constructs a digital and intelligent capability map of students based on collected multidimensional data and big data analysis technology. The teaching effectiveness analysis unit quantitatively evaluates the teaching process and results, identifies the strengths and weaknesses in teaching, and the iterative optimization decision-making unit feeds back the evaluation results to the resource development and teaching decision-making stages, driving the continuous improvement and intelligent upgrading of the teaching system.

[0063] The industry-education integration resource library unit includes an industry case collection processor, a scenario transformation processor, and an ideological and political integration processor. The industry case collection processor is responsible for connecting with partner companies, collecting mathematical application cases in real production scenarios, and establishing a corporate case material library. The scenario transformation processor transforms the original corporate cases for teaching purposes, extracts the mathematical problems, designs task scenarios and problem chains, and transforms them into teaching resources that meet the cognitive level of higher vocational students. The ideological and political integration processor integrates educational elements into the case resources.

[0064] The knowledge graph construction unit includes a knowledge point parsing processor, a relationship mining processor, and a graph visualization processor. The knowledge point parsing processor performs fine-grained decomposition of the vocational college mathematics course content, identifies knowledge elements, and establishes a knowledge point tagging system. The relationship mining processor constructs a multi-dimensional relationship network by analyzing the logical dependencies between knowledge points and the mapping relationship between vocational ability requirements and mathematical knowledge. The graph visualization processor presents the knowledge relationships in a graphical way and generates an interactive knowledge map.

[0065] The resource dynamic update unit includes a demand monitoring processor, a resource matching processor, and a version management processor. The demand monitoring processor tracks industry technology development, changes in professional standards, and updates to enterprise job requirements in real time, and identifies update needs for teaching resources. The resource matching processor automatically retrieves and matches new case materials based on changes in demand, triggering the content supplementation and replacement process of the resource library. The version management processor archives and manages the historical versions of the resource library, recording the time node, changes, and basis for each update.

[0066] The intelligent learning diagnosis unit includes a data preprocessor, a feature extraction processor, and a profile generation processor. The data preprocessor cleans, standardizes, and processes missing values ​​in the collected student learning data to construct a structured analysis dataset. The feature extraction processor extracts key features from the learning behavior data to generate multi-dimensional feature vectors. The profile generation processor integrates the multi-dimensional feature data to generate a personalized learning profile for each student.

[0067] The profile generation processor generates the student's comprehensive learning characteristic value F(t) at time t according to the following formula:

[0068] ;

[0069] Where D represents the total number of mathematical knowledge points. Let K be the weight of the d-th knowledge point. d (t) represents the mastery of the d-th knowledge point at time t, where t d This indicates the time when the student last studied the d-th knowledge point. Let V be the forgetting time constant for the d-th knowledge point. d This represents the importance of the d-th knowledge point to the student's target career position, where J is the total number of learning behavior dimensions. As the weight of learning behavior, A j (t) represents the student's activity level in the j-th learning behavior dimension, P j Let be the occupational relevance coefficient for the j-th behavioral dimension. Let C(t) represent the student's comprehensive ability index at time t, where C(t) is the ability weight. T is the learning cycle fluctuation coefficient. cyc For the learning cycle;

[0070] {K d (t)}、{A j C(t) and C(t) constitute the student's multidimensional feature vector;

[0071] The learning path planning unit includes a goal setting processor, a path algorithm processor, and a dynamic adjustment processor. The goal setting processor sets personalized, phased learning goals and ability achievement standards for each student based on the course teaching objectives, student learning profiles, and professional ability requirements. The path algorithm processor uses an AI adaptive algorithm to plan the optimal learning path from the current level to the target state. The dynamic adjustment processor monitors learning progress in real time and automatically adjusts subsequent learning paths based on task completion quality and ability improvement.

[0072] The path algorithm processor calculates the student's position in s according to the following formula. t The overall benefit value Q(s) of choosing learning task k under the given statet ,k):

[0073] ;

[0074] Among them, R K (s t k) represents state s t Knowledge layer benefit of task k, R C (s t k) represents state s t The ability layer benefit of task k, w K and w C For the two-layer objective weights, D k Let k be the capability requirement value. X is the fitness sensitivity coefficient, and X is the nonlinear adjustment index. As a discount factor, This is a path consistency penalty function;

[0075] The expression for the path coherence penalty function is:

[0076] ;

[0077] Where H represents the student's sequence of historical tasks, S(k, h) represents the similarity of knowledge points between task k and historical task h, and th represents the completion time of historical task h. The path memory decay constant, For continuity weights;

[0078] For the formula of comprehensive income value , representing the expected future return term, which indicates the return after completing the current task k and transitioning to the next state s. t+1 Then, the maximum Q-value that the student can obtain when facing all possible subsequent tasks k' is calculated using an iterative update mechanism: the system maintains a Q-value table, recording the historical Q-values ​​of all state-task pairs, and updates the table when needed. At that time, directly query the state s from the Q value table. t+1 The current Q value of each task is calculated and the maximum value is taken. As students continuously complete learning tasks, the Q value table is continuously updated through incremental learning and gradually converges to a stable state that accurately reflects the value of the task. This part is managed by AI.

[0079] The teaching strategy recommendation unit includes a strategy matching processor, a resource push processor, and an intervention reminder processor. The strategy matching processor intelligently matches suitable teaching methods from the teaching strategy library based on the learning situation diagnosis results and learning path requirements. The resource push processor automatically pushes relevant teaching resources, learning materials, and auxiliary tools to teachers and students according to the matched teaching strategies. The intervention reminder processor identifies abnormal situations and risk signals in the learning process and sends early warning information to teachers in a timely manner.

[0080] The classroom behavior monitoring unit includes a behavior data collector, a behavior feature recognition processor, and a behavior pattern analysis processor. The behavior data collector collects students' online learning behavior data in real time through the teaching platform interface. The behavior feature recognition processor performs feature engineering on the raw behavior data and extracts behavior feature indicators. The behavior pattern analysis processor is used to identify students' learning habits, behavior patterns, and potential problems.

[0081] The project outcome recording unit includes an automatic outcome archiver, an outcome quality assessment processor, and an outcome association processor. The automatic outcome archiver categorizes and stores various learning outputs submitted by students according to timeline and task type, creating electronic learning portfolios. The outcome quality assessment processor performs preliminary quality assessments on outcomes such as text reports, data analysis results, and visualizations, generating outcome feature tags. The outcome association processor establishes a mapping relationship between outcomes and knowledge points and ability objectives, tracking students' ability performance trajectories in different tasks.

[0082] The multi-source feedback integration unit includes a feedback information collector, a feedback data standardization processor, and a feedback fusion processor. The feedback information collector is used to collect evaluation information from students, teachers, peers, and corporate mentors. The feedback data standardization processor converts evaluation data from different sources and in different formats into a unified data structure and establishes a standardized evaluation index system. The feedback fusion processor uses a weighted fusion algorithm to integrate multiple evaluation opinions and generate an objective and comprehensive evaluation result.

[0083] The competency graph generation unit includes a competency indicator calculation processor, a graph modeling processor, and a visualization rendering processor. The competency indicator calculation processor calculates the quantitative indicator values ​​of students in multiple competency dimensions based on multidimensional data collection results. The graph modeling processor is used to construct a multidimensional competency assessment model to depict students' competency strengths and weaknesses. The visualization rendering processor presents competency data in an intuitive graphical way and generates personalized competency development reports.

[0084] The ability index calculation processor calculates the student's comprehensive ability value E in the m-th ability dimension according to the following formula. m (t):

[0085] ;

[0086] Where Smp is the score of the p-th data source for the m-th capability dimension, and P is the total number of data sources. Let p be the weight coefficient of the p-th data source. For the scoring uncertainty, For uncertain adjustment parameters, As an outcome-based evaluation index, For the incentive coefficient of progress, Perform a quality increment on the student's path. The average increment for all students For the group standard deviation, This serves as a verification factor for the consistency of process results.

[0087] The expression for the consistency verification factor of the process results is:

[0088] ;

[0089] Where, N m P represents the total number of evaluation items related to the m-th ability dimension. mn This represents the ranking of the process scores for the nth evaluation item, O mn To rank the results of the nth evaluation item, rank() is a function that returns the rank position. This is the consistency incentive coefficient;

[0090] The teaching effectiveness analysis unit includes an effectiveness index statistical processor, a comparative analysis processor, and an attribution diagnosis processor. The effectiveness index statistical processor performs statistical analysis on teaching process and result data and calculates core effectiveness indicators. The comparative analysis processor identifies the changing trends and group differences in teaching effectiveness by longitudinally comparing students' historical performance and horizontally comparing the overall level of the class. The attribution diagnosis processor is used to analyze the key factors affecting teaching effectiveness and locate the influencing mechanisms of teaching resources, teaching methods, student characteristics, etc.

[0091] The iterative optimization decision unit includes a problem identification processor, an optimization scheme generation processor, and a feedback execution processor. The problem identification processor automatically identifies the weak links and areas for improvement in the teaching system based on the teaching effectiveness analysis results. The optimization scheme generation processor generates optimization suggestions for the identified problems. The feedback execution processor automatically transmits the optimization schemes to the corresponding functional modules.

[0092] In the above text, d, j, p, m, and n are ordinal numbers used to represent sequence numbers and have no actual meaning.

[0093] To verify the beneficial effects of this invention, a comparative experiment was conducted for one semester in mathematics teaching at a vocational college. Two parallel classes of students from the 2023 cohort of the Mechatronics major were selected as the research subjects, with 45 students in each class and similar academic levels. The experimental group used the intelligent teaching decision-making and evaluation system of this invention, while the control group used the traditional teaching model.

[0094] The specific steps of the experiment are as follows: The first stage is the system initialization stage. A mathematics teaching resource library containing 50 real-world enterprise cases is built through the intelligent resource development module, a knowledge graph covering 156 knowledge points is established, and a dynamic resource update mechanism is established with three partner enterprises. The second stage is the learning diagnosis and path planning stage. The system conducts pre-tests on the experimental group students, collecting multi-dimensional data including knowledge mastery, learning behavior characteristics, and ability foundation, generating personalized learning profiles for each student, and planning initial learning paths through AI adaptive algorithms. The third stage is the teaching implementation stage, lasting 16 weeks. The system continuously collects online learning behavior data from the experimental group students, including average weekly login time, number of courseware views, exercise completion status, and frequency of interaction in discussion forums, accumulating over 18,000 data entries. Simultaneously, 526 project reports and data analysis works submitted by students are recorded; and 1,350 multi-source evaluation data entries from student self-evaluation, peer evaluation, teacher evaluation, and feedback from enterprise mentors are integrated. The fourth stage is the dynamic adjustment and optimization stage. Based on students' task completion quality and ability improvement, the system automatically adjusts the learning path every two weeks, with an average of 4.2 adjustments per student. The teaching strategy recommendation unit pushes differentiated teaching suggestions to teachers 128 times, triggering 23 early warning interventions. The fifth stage is the comprehensive evaluation stage. At the end of the semester, the system generates a digital intelligence ability map for each student, including quantitative scores in five dimensions: mathematical modeling ability, data analysis ability, problem-solving ability, collaboration ability, and vocational application ability. The control group uses traditional lecture-based teaching, with uniform textbooks and teaching progress, and is evaluated through assignments, quizzes, and a final exam.

[0095] The data was organized and obtained Figure 6 and Figure 7 .

[0096] Example 3.

[0097] This embodiment provides a specific implementation plan for an intelligent teaching decision-making and evaluation system for higher vocational mathematics that integrates AI adaptive algorithms. The system is deployed on an Alibaba Cloud ECS server cluster, using four 8-core 16GB computing nodes and two independent data storage nodes. The database uses MySQL 8.0, the caching layer uses Redis 6.2, the front-end interactive interface is developed based on the Vue.js 3.0 framework, the back-end service adopts the Spring Boot 2.7 microservice architecture, and the AI ​​algorithm module is implemented using the TensorFlow 2.10 deep learning framework and the Scikit-learn 1.1 machine learning library.

[0098] In this embodiment, the intelligent resource development module has established in-depth cooperation with 12 local manufacturing enterprises and 8 information technology enterprises. The industry case collection processor in the industry-education integration resource library unit is equipped with a dedicated enterprise interface. Through regular interviews and questionnaires, 15 to 20 mathematical application cases in real production scenarios are collected from partner enterprises each month. These cases include process parameter optimization, quality control analysis, production scheduling, cost accounting modeling, etc. The case materials are stored in various formats such as PDF documents, Excel data tables, CAD drawings, and on-site photos, with a total capacity of 180GB. The scenario transformation processor adopts a template-based processing flow, which structures and transforms enterprise cases according to five modules: scenario description, problem statement, knowledge association, task design, and evaluation criteria. The teaching processing cycle of each case is controlled within 3 working days. The difficulty level of the transformed cases is marked from 1 to 5 to suit students of different levels. The ideological and political integration processor embeds educational elements such as craftsmanship, innovation awareness, and professional ethics into case resources. Each case contains at least two points for ideological and political integration. Through the case background, it introduces the company's independent innovation process or the dedication of engineers, and strengthens quality awareness and the concept of striving for excellence in the process of problem solving.

[0099] The knowledge graph construction unit uses Neo4j 4.4 graph database to store knowledge relationships. The knowledge point parsing processor breaks down the content of vocational college mathematics courses into six major knowledge domains: functions, limits, derivatives, integrals, linear algebra, and probability and statistics. Each knowledge domain is further subdivided into 3 to 5 knowledge modules, and each module is further subdivided into 8 to 12 specific knowledge points, identifying a total of 342 knowledge elements. A tag system with 12 attributes, including difficulty level, cognitive level, suggested study time, prerequisites, and subsequent connections, is established for each knowledge point. The relationship mining processor uses a combination of expert interviews and data mining to identify six types of logical dependencies between knowledge points, including prerequisite relationships, application relationships, extension relationships, comparison relationships, comprehensive relationships, and transfer relationships, establishing a total of 1856 connection edges between knowledge points. At the same time, it analyzes the ability requirements of 32 typical occupational positions, establishing 763 sets of mapping relationships between mathematical knowledge points and occupational abilities. For example, it maps the knowledge of function monotonicity to the ability to analyze process curves, and the knowledge of probability distribution to the ability to control quality. The knowledge network visualization processor uses the D3.js visualization library to render knowledge networks. The size of nodes is dynamically adjusted according to the importance of knowledge points, and the thickness of edges reflects the strength of associations. It supports multi-level zooming and interactive exploration. Teachers can rearrange nodes by dragging and dropping them, and students can click on nodes to view details of knowledge points and related learning resources.

[0100] The resource dynamic update unit's demand monitoring processor is configured with an industry news crawler to automatically crawl information such as technology news, standard updates, and job requirements in the manufacturing and information technology fields daily. It uses the TF-IDF algorithm and BERT semantic analysis model to identify technological changes related to mathematics teaching. When the frequency of a keyword in a certain technology field exceeds a threshold of 15 times, it automatically triggers a resource update requirement reminder. The resource matching processor maintains a pool of candidate resources containing over 8,000 enterprise case materials. It uses a cosine similarity algorithm to calculate the matching degree between requirements and resources. When the matching degree exceeds 0.75, it recommends resource supplementation; when the matching degree is between 0.6 and 0.75, it prompts resource modification; and when the matching degree is below 0.6, it initiates a new resource development process. The version management processor generates a version number and update log for each update of the resource library, retaining the five most recent historical versions for retrospective querying. Each version records 12 metadata items, including update time, changed content, update basis, and reviewer, ensuring the traceability of resource evolution. In this embodiment, the resource library underwent 4 large-scale updates and 17 partial updates within an academic year, adding 63 new case resources, replacing 28 outdated cases, and revising case content in 89 places.

[0101] The intelligent learning diagnosis unit in the adaptive teaching decision-making module cleans the collected raw learning data, identifies and processes outliers, duplicates, and missing values. Outliers are detected using box plots; data points exceeding 1.5 times the interquartile range are marked as outliers and removed. Missing values ​​are filled using multiple imputation; data items with a missing rate exceeding 30% are directly deleted. The feature extraction processor extracts 45 behavioral features from the learning behavior data, such as knowledge mastery, practice accuracy, video viewing completion rate, number of comments in discussion forums, and timeliness of assignment submission. From the outcome data, it extracts 18 outcome features, such as report word count, data processing complexity, number of visualization charts, and innovation score. From the feedback data, it extracts 12 evaluation features, such as self-evaluation score, average peer evaluation, teacher comments keywords, and corporate mentor ratings, forming a 75-dimensional student feature vector. The profile generation processor integrates multi-dimensional feature data to generate an electronic learning profile for each student, encompassing five dimensions: basic information, knowledge mastery status, learning behavior characteristics, ability development trajectory, and learning style preferences. The profiles are presented in visual formats such as radar charts, bar charts, and line charts, and are updated every two weeks. Teachers can view an overview of the profiles for the entire class and detailed profiles for individual students through the system.

[0102] The goal-setting processor in the learning path planning unit sets phased learning goals for each student based on the vocational college mathematics curriculum standards, students' current level, and the competency requirements of their target positions. The 16-week teaching cycle of a semester is divided into four learning phases, each with 3-5 knowledge mastery goals and 2-3 competency achievement goals. The difficulty of the goals is differentiated according to students' foundational levels: students with weak foundations focus on understanding basic concepts and practicing basic problems; average students focus on applying knowledge and practicing comprehensive problems; and high-achieving students focus on extended learning and exploring innovative problems. The path algorithm processor uses a deep Q-network algorithm to plan the learning path, encoding the learning task state as a 256-dimensional vector. The action space contains 158 selectable learning tasks. The reward function comprehensively considers the immediate benefits of knowledge mastery and the long-term benefits of competency development. During the algorithm training phase, offline training is conducted using historical teaching data, with a training sample size of 12,000 learning trajectory records. After 5,000 training iterations, the model converged, and the path recommendation accuracy on the test set reached 87.3%. The processor dynamically adjusts its assessment of students' mastery after completing three learning tasks. If the quality of task completion falls below the expected standard for three consecutive times, the difficulty of subsequent tasks is automatically reduced or supplementary learning tasks are inserted. If the quality of task completion exceeds the expected standard for three consecutive times, the difficulty of subsequent tasks is automatically increased or repetitive tasks are skipped. The trigger threshold for the adjustment strategy can be set by the teacher according to the actual teaching situation. The default threshold is ±15% of the expected standard.

[0103] The strategy matching processor in the teaching strategy recommendation unit maintains a strategy library containing 48 teaching methods, covering various teaching models such as lectures, discussions, case studies, project-based learning, flipped classrooms, and blended learning. Each teaching method is labeled with attributes such as applicable scenarios, implementation conditions, and expected effects. Strategy matching uses a combination of rule-based expert systems and case-based reasoning methods. When a student's learning profile shows that their mastery of a certain knowledge point is below 60%, a strategy combining intensive lectures and individual tutoring is recommended. When a student's learning behavior shows that their participation is below the standard value, a strategy combining group discussions and project-driven learning is recommended. The resource push processor automatically pushes relevant teaching materials, teaching videos, question banks, case studies, and other resources to teachers based on the matched teaching strategies, and pushes personalized learning materials, micro-lecture videos, exercises, and extended reading materials to students. The push frequency is 2 to 3 times per week, and the total amount of pushed content is controlled within a range acceptable to students to avoid information overload. The intervention reminder processor sets five types of early warning rules: a learning stagnation warning is triggered when a student does not log in to the system for three consecutive weeks; a knowledge weakness warning is triggered when a student fails a test on a certain knowledge point twice in a row; a learning integrity warning is triggered when the similarity of a student's homework exceeds 80%; an insufficient participation warning is triggered when a student's classroom interaction frequency is less than 50% of the class average; and a developmental stagnation warning is triggered when a student's ability improvement curve shows no significant growth for four consecutive weeks. The warning information is sent to the teacher simultaneously through system messages, emails, and SMS.

[0104] The classroom behavior monitoring unit within the multi-dimensional data acquisition module interfaces with APIs of teaching platforms such as Smart Vocational Education and Chaoxing Learning Platform to collect real-time data on 12 categories of online learning behaviors, including student login duration, page browsing, video viewing, homework submission, quiz answers, and discussion posting. Data is collected every 5 minutes, and the raw data is stored in JSON format in a MongoDB 5.0 NoSQL database, accumulating to 2.3TB. The behavior feature recognition processor uses time-series analysis to extract students' learning rhythm characteristics, identifying eight typical learning time patterns, including focused, intermittent, cramming, and procrastination-based learning. Cluster analysis is used to identify students' resource preference characteristics, revealing six typical resource preference patterns: video-based, text-based, case-based, and practice-based learning. The behavioral pattern analysis processor uses association rule mining algorithms to discover the correlation patterns between learning behaviors and learning outcomes. For example, it found that students who watched videos in full for more than 85% of the time and did synchronized exercises had an average knowledge mastery rate 18 percentage points higher than other students. It also found that students who participated in discussion forums more than 5 times a week had an average problem-solving ability score 12 points higher than other students.

[0105] The project outcome recording unit's automatic archiver uses a combination of file system and database storage. Document-type outcomes are stored in Alibaba Cloud OSS object storage service in PDF format, while data-type outcomes are stored in Excel and CSV formats. Project-type outcomes include various forms such as charts, models, and code. Each outcome file is automatically named in the format "Student ID_Name_Task Name_Submission Time," and an index table is created to record the outcome's metadata information. The outcome quality assessment processor uses natural language processing technology to perform preliminary evaluations of text reports, calculating indicators such as word count, paragraph count, keyword coverage, and language fluency. Image recognition technology is used to perform preliminary evaluations of visual works, identifying features such as chart type, color scheme, and data display completeness. The preliminary evaluation results serve as an auxiliary reference for teachers' manual evaluation, achieving an accuracy rate of 73.5%. The outcome association processor labels each learning outcome with associated knowledge point tags, ability goal tags, and task type tags, establishing a many-to-many mapping relationship between outcomes and course objectives. By tracking students' performance in different tasks, it generates an ability development trajectory map, visually presenting students' growth process in various ability dimensions.

[0106] The feedback information collector in the multi-source feedback integration unit provides a unified evaluation interface. Students conduct self-evaluation and peer evaluation through a mobile app, teachers conduct teaching evaluation and process guidance through a PC system, and industry mentors evaluate students' performance during their internships through a web system. Evaluation dimensions include eight aspects: knowledge application, problem-solving, teamwork, and professional qualities, with each dimension using a 5-level rating system. The feedback data standardization processor converts evaluation data from different sources into a unified data structure. For qualitative evaluations, sentiment analysis technology is used to extract evaluation levels; for quantitative scores, normalization is performed, converting scores of different dimensions into a percentage system; and missing evaluation items are filled with default values ​​or predicted based on historical data. The feedback fusion processor uses a weighted average method to synthesize multiple evaluation opinions. Student self-evaluation is weighted at 0.15, peer evaluation at 0.20, teacher evaluation at 0.45, and industry feedback at 0.20. These weighting parameters can be adjusted according to different course characteristics and evaluation stages. The final comprehensive evaluation result is an important component of the student's grade, accounting for 40% of the total course grade.

[0107] The competency graph generation unit in the comprehensive evaluation and optimization module uses a competency index calculation processor to calculate quantitative indicators for students across five competency dimensions: mathematical modeling, data analysis, problem-solving, collaborative communication, and career application, based on collected multidimensional data. Each competency dimension has 3 to 5 secondary indicators, totaling 18 subdivided competency indicators. The indicator calculation comprehensively considers process data and outcome data, with process data weighted at 60% and outcome data weighted at 40%. Students with significant improvement are given additional incentive points, up to a maximum of 5 points. The graph modeling processor uses the analytic hierarchy process (AHP) to determine the relative importance of each competency indicator. It obtains the indicator weights by constructing a judgment matrix and calculating eigenvectors. Consistency checks ensure the rationality of the weight settings. A consistency verification mechanism between process and result is introduced during the modeling process. When there is a significant deviation between a student's process performance and the final result, the system automatically marks the anomaly and prompts for manual review. The visualization rendering processor uses the ECharts 5.0 visualization library to generate a five-dimensional ability radar chart. The distance from the vertex of each of the five dimensions to the center point in the radar chart represents the ability score. Different colors distinguish the ability status at the beginning and end of the semester, clearly showing the student's ability growth trajectory. At the same time, it generates a personalized ability development report containing four parts: ability score, strengths analysis, weaknesses diagnosis, and improvement suggestions. The report is exported in Word document format for students to download and view.

[0108] The teaching effectiveness analysis unit uses a statistical processor to calculate 15 core indicators, including knowledge mastery rate, ability achievement rate, task completion rate, learning participation, and satisfaction evaluation. Knowledge mastery rate is calculated using the average accuracy rate of knowledge point quizzes; ability achievement rate is calculated as the ratio of ability indicator scores to target scores; task completion rate is calculated based on the proportion of students who complete tasks on time; learning participation comprehensively considers indicators such as attendance rate, interaction frequency, and homework submission rate; and satisfaction evaluation is collected through questionnaires to gather students' subjective evaluations of course content, teaching methods, and learning support. The comparative analysis processor performs longitudinal comparisons of individual students, analyzing ability changes at the beginning, middle, and end of the semester to identify periods of rapid and plateaued ability growth. It also performs horizontal comparisons of the class as a whole, analyzing the ability distribution characteristics of different student groups and identifying the proportion and characteristics of high-achieving students, students with significant progress, and students requiring support. The attribution diagnostic processor uses decision tree algorithm and regression analysis to identify key factors affecting teaching effectiveness. The analysis found that the influence weights of resource quality, path fit, teacher guidance frequency, and student time spent on teaching effectiveness were 0.28, 0.25, 0.23, and 0.24, respectively, providing a data basis for subsequent optimization.

[0109] The problem identification processor in the iterative optimization decision unit uses anomaly detection algorithms to automatically identify weaknesses in the teaching system. When the average mastery rate of a knowledge point is below 75% for two consecutive rounds, it is determined that there is a problem with the teaching resources or teaching strategies for that knowledge point. When the completion quality of a certain type of learning task is significantly lower than that of other tasks, it is determined that there is a problem with the design difficulty or guidance support of that task. The problem identification results are prioritized according to the scope and severity of impact. The optimization solution generation processor generates specific improvement suggestions for the identified problems. For resource quality problems, it suggests updating case content or adding teaching videos; for path adaptation problems, it suggests adjusting the task order or modifying the task difficulty; for teaching strategy problems, it suggests changing teaching methods or adding interactive elements. Each optimization suggestion includes information such as problem description, improvement plan, expected effect, and implementation difficulty, and is implemented after being reviewed by teaching management personnel. The feedback execution processor automatically transmits the approved optimization solutions to the corresponding functional modules to trigger the update process. Resource adjustment suggestions are transmitted to the resource dynamic update unit of the intelligent resource development module, strategy optimization suggestions are transmitted to the teaching strategy recommendation unit of the adaptive teaching decision module, and path improvement suggestions are transmitted to the learning path planning unit, forming a complete closed loop from evaluation feedback to system optimization and then to teaching implementation. In this embodiment, the system performs a total of 23 optimization iterations in one academic year, and the teaching effect continues to improve.

[0110] The content disclosed above is only a preferred and feasible embodiment of the present invention, and is not intended to limit the scope of protection of the present invention. Therefore, all equivalent technical changes made based on the content of the present invention specification and drawings are included within the scope of protection of the present invention. Furthermore, the elements therein can be updated as technology develops.

Claims

1. A vocational college mathematics intelligent teaching decision-making and evaluation system integrating AI adaptive algorithms, characterized in that, It includes an intelligent resource development module, an adaptive teaching decision-making module, a multi-dimensional data acquisition module, and a comprehensive evaluation and optimization module; The intelligent resource development module is responsible for building a dynamic mathematics teaching resource library; the adaptive teaching decision module dynamically plans personalized learning paths based on AI algorithms; the multi-dimensional data acquisition module comprehensively acquires behavioral data, outcome data, and feedback data during the teaching process; and the comprehensive evaluation and optimization module generates a capability map through data analysis. The intelligent resource development module includes an industry-education integration resource library unit, a knowledge graph construction unit, and a resource dynamic update unit. The industry-education integration resource library unit is responsible for the collaborative development of industry case libraries by schools and enterprises, transforming real professional scenarios into mathematics teaching materials. The knowledge graph construction unit uses digital intelligence technology to sort out the relationship between mathematical knowledge points and professional abilities, and establish a structured knowledge network system. The resource dynamic update unit establishes an automatic update mechanism for the resource library based on changes in enterprise needs and technological iterations. The adaptive teaching decision-making module includes a learning situation intelligent diagnosis unit, a learning path planning unit, and a teaching strategy recommendation unit. The learning situation intelligent diagnosis unit is used to analyze students' knowledge mastery level, learning style, and ability characteristics to generate personalized learning situation profiles. The learning path planning unit dynamically adjusts task difficulty, content sequence, and learning pace based on the learning situation diagnosis results. The teaching strategy recommendation unit provides teachers with differentiated teaching method suggestions. The multi-dimensional data acquisition module includes a classroom behavior monitoring unit, a project achievement recording unit, and a multi-source feedback integration unit. The classroom behavior monitoring unit collects students' online learning behavior data in real time through an intelligent teaching platform. The project achievement recording unit systematically stores students' works, reports, and practical results generated in inquiry tasks and collaborative projects, establishing a complete learning archive. The multi-source feedback integration unit gathers multi-dimensional evaluation information from student self-evaluation, peer evaluation, teacher evaluation, and enterprise feedback. The comprehensive evaluation and optimization module includes a capability map generation unit, a teaching effectiveness analysis unit, and an iterative optimization decision-making unit. The capability map generation unit constructs a digital and intelligent capability map of students based on collected multidimensional data and big data analysis technology. The teaching effectiveness analysis unit quantitatively evaluates the teaching process and results, identifies the strengths and weaknesses in teaching, and the iterative optimization decision-making unit feeds back the evaluation results to the resource development and teaching decision-making stages, driving the continuous improvement and intelligent upgrading of the teaching system.

2. The intelligent teaching decision-making and evaluation system for higher vocational mathematics integrating AI adaptive algorithms as described in claim 1, characterized in that, The intelligent learning diagnosis unit includes a data preprocessor, a feature extraction processor, and a profile generation processor. The data preprocessor cleans, standardizes, and processes missing values ​​in the collected student learning data to construct a structured analysis dataset. The feature extraction processor extracts key features from the learning behavior data to generate multi-dimensional feature vectors. The profile generation processor integrates the multi-dimensional feature data to generate a personalized learning profile for each student. The profile generation processor generates the student's comprehensive learning characteristic value F(t) at time t according to the following formula: ； Where D represents the total number of mathematical knowledge points. Let K be the weight of the d-th knowledge point. d (t) represents the mastery of the d-th knowledge point at time t, where t d This indicates the time when the student last studied the d-th knowledge point. Let V be the forgetting time constant for the d-th knowledge point. d This represents the importance of the d-th knowledge point to the student's target career position, where J is the total number of learning behavior dimensions. As the weight of learning behavior, A j (t) represents the student's activity level in the j-th learning behavior dimension, P j Let be the occupational relevance coefficient for the j-th behavioral dimension. Let C(t) represent the student's comprehensive ability index at time t, where C(t) is the ability weight. T is the learning cycle fluctuation coefficient. cyc This is the learning cycle.

3. The intelligent teaching decision-making and evaluation system for higher vocational mathematics integrating AI adaptive algorithms as described in claim 2, characterized in that, The learning path planning unit includes a goal setting processor, a path algorithm processor, and a dynamic adjustment processor. The goal setting processor sets personalized, phased learning goals and ability achievement standards for each student based on the course teaching objectives, student learning profiles, and professional ability requirements. The path algorithm processor uses an AI adaptive algorithm to plan the optimal learning path from the current level to the target state. The dynamic adjustment processor monitors learning progress in real time and automatically adjusts subsequent learning paths based on task completion quality and ability improvement. The path algorithm processor calculates the comprehensive benefit value Q(st, k) of the student choosing learning task k in state st according to the following formula: ； Among them, R K (s t k) represents state s t Knowledge layer benefit of task k, R C (s t k) represents state s t The ability layer benefit of task k, w K and w C For the two-layer objective weights, D k Let k be the capability requirement value. X is the fitness sensitivity coefficient, and X is the nonlinear adjustment index. As a discount factor, This is a path consistency penalty function.

4. The intelligent teaching decision-making and evaluation system for higher vocational mathematics integrating AI adaptive algorithms as described in claim 3, characterized in that, The competency graph generation unit includes a competency index calculation processor, a graph modeling processor, and a visualization rendering processor. The competency index calculation processor calculates quantitative index values ​​for students in multiple competency dimensions based on multidimensional data collection results. The graph modeling processor is used to construct a multidimensional competency assessment model to depict students' competency strengths and weaknesses. The visualization rendering processor presents competency data in an intuitive graphical way and generates personalized competency development reports.

5. The intelligent teaching decision-making and evaluation system for higher vocational mathematics integrating AI adaptive algorithms as described in claim 4, characterized in that, The ability index calculation processor calculates the student's comprehensive ability value E in the m-th ability dimension according to the following formula. m (t): ； Where Smp is the score of the p-th data source for the m-th capability dimension, and P is the total number of data sources. Let p be the weight coefficient of the p-th data source. For the scoring uncertainty, For uncertain adjustment parameters, As an outcome-based evaluation index, For the incentive coefficient of progress, Perform a quality increment on the student's path. The average increment for all students For the group standard deviation, This serves as a verification factor for the consistency of process results.

Images