Student comprehensive ability evaluation system based on artificial intelligence

Abstract

The invention discloses a student comprehensive ability evaluation system based on artificial intelligence, and belongs to the technical field of artificial intelligence student learning evaluation. Aiming at the problems of single dimension and poor evaluation effect in the existing student evaluation system, the system not only can provide personalized feedback and improvement suggestions for students and help the students to improve team cooperation, communication and problem solving capabilities, but also can provide accurate teaching evaluation and decision support for teachers and education managers, so that the students can have a better teaching effect. And intelligent upgrading of a teaching mode is promoted. By promoting digital transformation of education evaluation, the system provides an innovative solution for constructing a more scientific, fair and efficient learning evaluation system, and assists high-quality development of education.

Description

Technical Field

[0001] The present invention belongs to the technical field of artificial intelligence-based student learning evaluation, and particularly relates to a comprehensive student ability evaluation system based on artificial intelligence. Background Art

[0002] In recent years, the development of pedagogy, psychology, learning science, neuroscience, and information technology has provided a new theoretical basis and technical support for the innovation of the student comprehensive ability evaluation system. Currently, the student-centered education concept emphasizes the importance of teamwork, communication, and problem-solving abilities in the all-round development of students. However, the existing education evaluation system still mainly focuses on the result-oriented evaluation mode, making it difficult to comprehensively and objectively evaluate these core abilities of students, especially in the integrated analysis of cognitive and non-cognitive factors, where there are significant limitations.

[0003] The deficiencies of the traditional learning evaluation system are mainly reflected in: First, the existing evaluation methods mainly focus on the knowledge learning situation of students, but insufficiently monitor the development process of teamwork, communication, and problem-solving abilities, making it difficult to accurately identify the ability performance of students in actual learning and practice; Second, the existing systems often lack effective evaluation of non-cognitive factors (such as cooperation awareness, expression ability, thinking flexibility, etc.), resulting in an incomplete evaluation of the overall ability of students; In addition, the problem of data islands is serious, and the current system is difficult to effectively integrate multi-source heterogeneous data, restricting the in-depth mining and value-added evaluation of the comprehensive ability development of students. Summary of the Invention

[0004] To make up for these deficiencies, the present invention proposes a comprehensive student ability evaluation system based on artificial intelligence. This system integrates artificial intelligence technology, educational big data analysis, neuroscience, and multi-modal data processing technology to build a comprehensive, systematic, and digital intelligent evaluation system. By using AI to monitor the learning process of students in real time, the system can accurately analyze their teamwork ability (such as cooperation attitude, role division, group decision-making), communication ability (such as language expression, listening comprehension, view elaboration), and problem-solving ability (such as problem analysis, solution formulation, practical application). At the same time, the system adopts multi-source data fusion technology to integrate multi-modal information such as classroom interaction, practical performance, physiological data, and emotional characteristics to achieve a deep evaluation of the ability development of students.

[0005] The "comprehensive student ability evaluation system based on artificial intelligence" of the present invention covers two core subsystems: the learning situation evaluation subsystem and the comprehensive ability evaluation subsystem.

[0006] (1) The learning situation evaluation subsystem includes: a learning foundation analysis and learning interest mining module, a learning method identification and learning style diagnosis module, a learning habit monitoring and learning progress tracking module, and a learning efficiency evaluation.

[0007] 1.1 The learning foundation analysis module: Based on an artificial intelligence-based learning foundation evaluation model, analyze the historical learning data of students to evaluate the current knowledge reserve of students. The implementation method is as follows: Use the historical score data S of students i and the completion situation A of usual homework i to be combined through a weighted formula: F i =α·normalize(S i )+β·normalize(A i ). Further, introduce a knowledge graph to analyze the knowledge coverage of students, and construct an embedding vector k of students' knowledge reserve and higher-order thinking i =GNN(KG,S i ). Among them, GNN is a graph neural network (Graph Neural Network) used to capture the relationships between knowledge points to more accurately evaluate the scope of students' knowledge mastery and thinking quality. Based on an artificial intelligence-based interest mining evaluation model, analyze the interest tendencies of students in different disciplines and track their participation in different courses and contents. The interest mining model will generate an interest vector I for each student i , representing their interest tendencies in different courses; each dimension of the vector represents the interest intensity of the student in a subject area, and the formula is: I i =softmax(E i ); E i represents the participation degree. Through the softmax function, ensure that the values of the interest vector are within the range of [0,1] to reflect the relative intensity of the students' interest in different disciplines. Further, calculate the participation degree according to the participation behaviors of students in different courses: Among them, T i represents the duration of participating in the class, D i represents the number of times of participating in discussions, T max and D max are the maximum values of the duration of participating in the class and the number of times of participating in discussions respectively, and λ1 and λ2 are weight parameters.

[0008] 1.2 The learning method identification module: Based on an artificial intelligence-based learning method identification model, identify the learning method of students according to the learning interaction data of students. The implementation method is as follows: Extract the data features of students' participation in discussions, interaction with peers, and independent thinking during the learning process, and generate a learning method feature vector: f i =[Discussion i ,Interaction i ,Duration i , where, Discussion i represents the number of times of students' participation in discussions, Interactioni Indicates the duration of a student's interaction with peers, Duration i is the duration of a student's independent study. Further, the decision tree model can automatically identify a student's learning style based on learning behavior characteristics: way i = DecisionTree(f i ), where way i represents a student's learning style (such as preferring self-study, cooperative learning, etc.). The learning style diagnosis model based on artificial intelligence uses a support vector machine to identify the type of a student's learning style in order to provide personalized teaching resources. The implementation method is as follows: Construct a learning style feature vector based on a student's learning foundation, interests, and learning style, etc.: m i = [v i ,a i ,t i ,o i ,i i ,], where v i represents the usage frequency of videos, a i represents the usage frequency of audio, t i represents the usage frequency of text, o i represents the frequency of hands-on operations, i i represents the frequency of integrated learning. Furthermore, use a support vector machine (Support Vector Machine, SVM) to diagnose a student's learning style: style i = SVM(m i ), where style i is the category of a student's learning style.

[0009] The learning habit monitoring module described in 1.3: The learning habit monitoring model based on artificial intelligence uses time series analysis technology to evaluate a student's time arrangement and learning frequency, and monitors whether their learning habits are suitable for their learning goals. Among them, set the learning duration of a student in a day to be T i (t), the learning frequency to be f i (t), and extract the main factors of the learning frequency through Fourier transform: Among them, is used to analyze the periodicity of the learning frequency and time arrangement. Further, use the K-means clustering algorithm to classify learning habits: habit i = K-means(T i ,F i ), where habit i is the learning habit of a student. The learning progress tracking model based on artificial intelligence analyzes a student's course learning situation and assignment submission situation, and generates a real-time progress report. The calculation formula is: Among them, Ccorsei is the number of completed courses, C total is the total number of courses. The formula for calculating the assignment submission status is: where c is the specific course, J c,total is the total number of assignments that the student needs to complete for this course, J c,submitted is the number of assignments that have been submitted for this course, P c,homework is the assignment submission progress for this course. Further, analyze the time taken by the student to complete different tasks to identify the bottlenecks restricting them in the learning process. The implementation method is: The time taken by the student to complete the i-th task is T i , by comparing with the average completion time T avg , define the speed score S speed (i) as: where T avg is the average completion time of this task; S speed (i) > 1 indicates that the student completes the task faster than the average level, and S speed (i) < 1 indicates that the student is slower.

[0010] The learning efficacy evaluation module described in 1.4: An artificial intelligence-based learning efficacy evaluation model to evaluate the matching degree of the methods, strategies and goals adopted by students in different learning tasks, as well as the efficiency and effect of achieving the expected learning goals. To evaluate the matching degree of the methods, strategies and goals, define a matching degree scoring function M score (i) to measure the matching degree of the methods and strategies adopted by the student in the i-th learning task with the goal of this task. Assume that the matching degree of the student in the i-th task is represented by M score (i), and the difficulty of the task goal is represented by D target (i), then define the score after adjusting the matching degree and difficulty for each task as: M adj (i) is higher, indicating that the student's learning methods and strategies are more matched with the task goals, and the performance of this strategy for tasks with different difficulties is better.

[0011] Further, to evaluate the efficacy of a single task, define the ratio of the actual completion time T actual (i) and the expected completion time T expected (i) of the student in the task as the efficiency of completing the task: where Q result (i) represents the result quality of the student in the i-th task, and T expected (i) is the reasonable time estimated by the system for completing this task. The efficiency score E efficiency(i) The higher it is, the higher the efficiency of the student in achieving the learning goal. To further comprehensively evaluate the overall effectiveness of the student in a certain type of task, we combine the matching degree and the effectiveness score to define the comprehensive learning effectiveness score, which is used to evaluate the performance of the student in task category C. j in the performance: Among them, α and β are weight parameters used to balance the proportion of the matching degree and the effectiveness score in the comprehensive effectiveness, and N j is the number of tasks in task category C j . Through this comprehensive score, the overall learning effectiveness of the student in a specific task category can be evaluated.

[0012] Furthermore, by using the task bottleneck identification technology in combination with the task type (such as comprehension tasks, memory tasks, application tasks, etc.), analyze in which types of tasks the student encounters bottlenecks. The learning tasks are divided into different task categories C j through cluster analysis, and the average speed of the student in each category is calculated: Among them, N j is the number of tasks in task category C j , and S speed (i) is the speed score of the student for completing the i-th task. By comparing the speed scores of different categories, the system can identify in which types of tasks the student performs poorly, and then find out the possible task bottlenecks.

[0013] (2) The comprehensive ability evaluation subsystem described above includes: a teamwork ability evaluation module, a communication ability evaluation module, and a problem-solving ability evaluation module.

[0014] The teamwork ability evaluation module described in 2.1: In a teamwork environment, introduce artificial intelligence-based intervention measures to improve team performance, and use the difference-in-differences model (DID) to analyze the changes in the collaboration performance of different teams before and after the introduction of the intervention measures to estimate the causal effect of the intervention. The implementation steps are as follows:

[0015] The first step is to preprocess the data, including: standardization of the collaboration score, segmentation of the time variable, and division of the treatment group and the control group. Among them, the standardization of the collaboration score: Due to differences in task complexity, scale, etc. among different teams, the absolute values of the collaboration scores may vary greatly. Before using the difference-in-differences model, the collaboration scores are standardized to ensure the comparability of the scores among different teams. The standardization formula is expressed as: Among them, Y' it is the standardized collaboration score, is the average collaboration score of team i, and σ iis the standard deviation of the collaboration score of Team i; Segmentation of the time variable: To better capture the changes in collaboration ability before and after the intervention, it is necessary to segment the time variable. Select a reasonable time window (such as weekly or monthly) for data segmentation to obtain multi-period data before and after the intervention measures; Post t represents the time dummy variable; Division of the treatment group and the control group: In the team, some teams are selected as the intervention group and other teams as the control group; The intervention group variable Treated i is used to indicate whether the team has received the intervention measure;

[0016] Second, use the difference-in-differences model (DID) to evaluate the changes in the collaboration performance of different teams before and after the intervention measures to estimate the causal effect of the intervention, expressed as: Y it = α + β1·Post t + β2·Treated i + δ·(Post t × Treated i ) + ∈ it , where Y it represents the collaboration ability score of Team i at time t (for example, the score obtained through models such as TACM, IQFM, RILM, etc.); α represents the intercept term in the model, that is, when all independent variables (such as Post t and Treated i ) are zero, the benchmark value of the team collaboration ability score; β1 represents the coefficient of the time dummy variable Post t . It reflects the impact of time factors on the collaboration ability score in the absence of intervention measures; β2 represents the coefficient of the treatment group dummy variable Treated i . It reflects the baseline difference between the teams receiving the intervention measure and the control group when there is no change in time factors; Post t is a time dummy variable indicating whether time t is after the implementation of the intervention measure. If it is, Post t = 1, and if it is not, Post t = 0; Treated i is a treatment group dummy variable indicating whether Team i has received the intervention measure. If it has, Treated i = 1, otherwise Treated i = 0; Post t × Treated i is the interaction term that captures the impact of the intervention on the collaboration ability; δ is the difference-in-differences estimator that reflects the actual impact of the intervention measure on the collaboration ability score; ∈ it is the error term;

[0017] Step 3, Estimation of Intervention Effect: Evaluate the effect of the intervention measure through the interaction term coefficient δ in the difference-in-differences model; this coefficient reflects the actual impact of the intervention measure on the team collaboration ability; if δ > 0, it indicates that the intervention measure has effectively improved the team collaboration ability; if δ ≤ 0, the effect of the intervention measure is not obvious, and there may even be a negative impact; the estimation formula for the intervention effect is as follows: Among them, represents the estimated value of the actual impact of the intervention measure on the collaboration ability score. It is the coefficient estimate of the interaction term Post t ×Treated i in the difference-in-differences model, evaluating the net effect of the intervention measure; N T and N C are the numbers of teams in the treatment group and the control group respectively; Y i,Post and Y i,Pre represent the collaboration ability scores after and before the intervention respectively;

[0018] Step 4, Visualization of Intervention Effect: Include the collaboration score trend chart and the effect comparison bar chart. Collaboration score trend chart: Plot the collaboration ability scores of the intervention group and the control group over time as a line chart to observe the trend of score changes before and after the intervention, and intuitively display the effect of the intervention; Effect comparison bar chart: Draw the bar chart of the changes in the collaboration ability scores of the intervention group and the control group before and after the intervention, and further display the effect of the intervention by comparing the differences between groups;

[0019] Step 5, Robustness Analysis of Intervention Effect: Include parallel trends test, sensitivity analysis, placebo test, and the specific steps are as follows: Parallel trends test: The difference-in-differences model assumes that the trends of collaboration score changes in the treatment group and the control group before the intervention are the same. Verify this assumption by testing whether the trends of score changes in the two groups before the intervention are consistent; if the collaboration score trends of the two groups before the intervention are significantly different, the results of the difference-in-differences model are unreliable, which means that in this case, the conclusion about the impact of the intervention measure on the collaboration ability obtained based on this model may be biased. At this time, it is possible to consider re-screening the samples and selecting teams with more similar collaboration score change trends before the intervention as the treatment group and the control group, or further adjusting the data, such as adopting a more refined time segmentation method, to make the collaboration score change trends of the two groups tend to be consistent before the intervention as much as possible, so as to ensure the reliability of the results of the difference-in-differences model.

[0020] Sensitivity analysis: Observe whether the intervention effect changes significantly by adjusting different time windows or different collaboration score standardization methods to verify the robustness of the model;

[0021] Placebo test: Without actual intervention, randomly select some teams as a virtual intervention group, run the difference-in-differences model, and check whether a significant effect appears. If the effect of the virtual intervention group is not significant, it indicates that the actual intervention effect is relatively credible.

[0022] Step 6, Final intervention effect score: After completing the model analysis and robustness test, obtain the final effect score of the intervention measure on the team collaboration ability. This score is used to measure the effectiveness of the intervention measure and provide decision-making support for future optimization. The formula for the final intervention effect score is: Among them, is the intervention effect estimated by the difference-in-differences method, Robustness check is the result of the robustness test, and λ1 and λ2 are weight parameters used to balance the intervention effect and robustness.

[0023] The communication ability evaluation module described in 2.2: Use the structural equation of interdisciplinary communication ability to analyze the comprehensive impact of different evaluation indicators (such as term coverage rate, communication fluency, emotional stability, etc.) on interdisciplinary communication ability, and improve the accuracy of evaluation through artificial intelligence technology. The implementation steps are as follows:

[0024] Step 1, Definition of model index variables: Term coverage rate - central idea, written genre, disciplinary norms; Communication fluency - purpose of communication, quality of arguments, supporting materials; Emotional stability - word choice, language organization, communication skills.

[0025] Step 2, The structural equation analysis of interdisciplinary communication ability includes both a measurement model and a structural model. The measurement model defines the relationship between the latent variable η and the observed variables of term coverage rate X1, communication fluency X2, and emotional stability X3:

[0026] X1 = λ1·η + ∈1

[0027] X2 = λ2·η + ∈2

[0028] X3 = λ3·η + ∈3

[0029] Among them, λ1, λ2, and λ3 are loading coefficients, indicating the contribution degree of each observed index to the latent variable, and ∈1, ∈2, and ∈3 represent error terms.

[0030] The structural model defines the relationship between latent variables: Taking interdisciplinary communication ability as a latent variable, the structural model is expressed as: η = γ1·ζ, where ζ is an external interference factor (such as personal experience, team cooperation efficiency, etc.), and γ1 is the influence coefficient of the external interference factor, indicating the impact of external factors on communication ability.

[0031] Step 3, Final Comprehensive Scoring Formula: According to the structural equation analysis of interdisciplinary communication ability, the final scoring formula for interdisciplinary communication ability is expressed as: CommunicationScore = λ1·X1 + λ2·X2 + λ3·X3 + γ1·ζ

[0032] Term Coverage Rate: It represents the breadth of students' use of relevant field terms in interdisciplinary communication. The term coverage rate can reflect the breadth of students' knowledge and their mastery of multi-field knowledge in interdisciplinary communication. Through the loading coefficient, we can determine the impact degree of the term coverage rate on interdisciplinary communication ability.

[0033] Communication Fluency: Evaluate students' expression fluency through factors such as speech rate, pauses, and voice stability. The loading coefficient can quantify the contribution of communication fluency to the overall communication ability.

[0034] Emotional Stability: Judge the emotional fluctuations of students during the communication process through voice emotion analysis. The loading coefficient is used to evaluate the impact of emotional stability on communication ability.

[0035] External Interference Factors (γ): These factors may include students' personal backgrounds, the quality of project cooperation, learning environments, etc. Through the coefficient γ1, we can measure the overall impact of these external factors on interdisciplinary communication ability.

[0036] The Problem-Solving Ability Evaluation Module Described in 2.3: Use the multi-dimensional problem-solving ability linear fusion model to evaluate students' problem-solving ability more carefully and comprehensively. Through the non-linear and linear fusion of multiple dimensional characteristics of students' problem-solving (such as problem decomposition ability, strategy generation ability, optimization degree of solutions, diversity, and innovation, etc.), the final comprehensive score is obtained; the core idea of this module is to map the characteristics of each dimension through multi-layer linear mapping, and perform weighted summation to gradually increase the complexity of the feature representation, and finally output a global comprehensive evaluation. First, perform linear transformation on each dimensional characteristic to generate new feature representations; through the multi-layer linear regression model for layer-by-layer combination, finally obtain the comprehensive score of students' problem-solving ability, and at the same time use artificial intelligence technology to achieve intelligent and visual comprehensive analysis. The implementation steps are as follows:

[0037] Step 1, Define Feature Vectors: The feature set of students in solving a certain problem is X = [x1, x2, …, x n , and each x i represents the characteristics of different dimensions; x1: Problem Decomposition Ability Score; x2: Strategy Generation Ability Score; x3: Solution Optimization Score; x4: Diversity Score; x5: Innovation Score;

[0038] Step 2, Multilayer Linear Mapping and Feature Weighting. Perform linear mapping on the features of each dimension to generate new feature representations. The formula for the first-layer linear transformation is as follows: where and are the weight matrix and bias term of the i-th feature respectively;

[0039] Through the above linear mapping, the initial feature vector X is transformed into the first-layer hidden representation Introduce multilayer linear transformations to capture the interactions and relationships between features. The formula for the l-th layer of linear transformation is: where and are the weight matrix and bias term of the (l + 1)-th layer respectively. Through layer-by-layer linear combination, a global comprehensive representation, the global feature vector

[0040] Step 3, Comprehensive Score Calculation: After multilayer linear mapping, map the final global feature vector to a comprehensive score S final , representing the overall problem-solving ability of the student; the formula is as follows: where and b (L) are the weights and bias terms of the last layer. Through this formula, the model can effectively fuse the ability scores of different dimensions to obtain a global comprehensive evaluation result;

[0041] Step 4, Model Training and Optimization. Train the multi-dimensional problem-solving ability linear fusion model by minimizing the mean squared error (MSE) loss function. When the true ability score of the student is y true , and the score predicted by the model is S final , then the loss function is: where N is the number of samples, and θ contains all the model parameters.

[0042] Compared with the prior art, the present invention has the following advantages:

[0043] The present invention has significant advantages in terms of comprehensiveness, personalization, and intelligence. First of all, through the learning situation evaluation subsystem of the present invention, seven modules are used to comprehensively evaluate the learning situation of students from multiple dimensions. These modules not only focus on the academic performance of students, but also include non-cognitive factors such as emotions, attitudes, and social abilities. Through this multi-dimensional evaluation, this system can more accurately reflect the all-round development of students, breaking through the limitation of traditional evaluation methods that only focus on academic achievements.

[0044] Secondly, the present invention introduces a variety of artificial intelligence evaluation models to accurately evaluate students' performance in teamwork, interdisciplinary communication, problem-solving and other abilities, filling the gaps in these fields in the traditional evaluation system. In terms of teamwork, the system uses a double-difference sub-model to compare and analyze the changes in team performance before and after the intervention measures, evaluate the causal effect of the intervention measures, and help improve students' teamwork ability. Through this dynamic feedback, teachers can timely adjust teaching strategies and improve students' learning efficiency and cooperation ability. In the evaluation of interdisciplinary communication ability, the present invention adopts the structural equation analysis method to comprehensively analyze the influence of different evaluation indicators on students' interdisciplinary communication ability. This method can comprehensively evaluate students' communication and cooperation ability in a multi-disciplinary background and improve the accuracy of evaluation. At the same time, the multi-dimensional problem-solving ability linear fusion model enables the system to evaluate students' problem-solving ability more meticulously and comprehensively, ensuring that the problem-solving ability is analyzed and improved in all aspects.

[0045] Finally, the learning situation evaluation subsystem and the comprehensive ability evaluation subsystem of the present invention provide accurate learning progress tracking functions through mutual cooperation, real-time monitoring and dynamic integration of heterogeneous data. This not only helps teachers comprehensively understand students, but also provides personalized learning feedback for students and improves comprehensive ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0047] Figure 1 It is a schematic diagram of the system structure of the present invention;

[0048] Figure 2 It is a system implementation operation flowchart of the present invention;

[0049] Figure 3 It is a schematic diagram of the operation result of the learning situation evaluation in the embodiment of the present invention;

[0050] Figure 4 It is an analysis schematic diagram of the problem-solving ability of the comprehensive ability in the embodiment of the present invention;

[0051] Figure 5 It is an analysis result schematic diagram of the comprehensive ability in the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0052] To gain a deep understanding of the present invention, we will provide a comprehensive and detailed description of it. However, the present invention has multiple implementation manners and is not limited to the specific examples listed herein. The presentation of these examples aims to deepen the overall understanding of the disclosed content of the present invention.

[0053] (I) Data collection

[0054] The present invention collects and integrates multiple heterogeneous data sources through artificial intelligence technology to ensure the comprehensiveness and accuracy of the data, providing a reliable basis for analysis and evaluation. The specific acquisition methods are as follows:

[0055] 1.1 Data sources. The system obtains objective data such as students' exam scores, assignment submissions, course participation behaviors, experimental reports, and project achievements by accessing standardized data interfaces of multiple learning management systems, online course platforms, exam systems, project management platforms, school portals, etc.; subjective data is collected through questionnaires, behavioral observations, and physiological measurements.

[0056] 1.2 Data types. (1) Objective data: Information obtained through measurable and verifiable means, usually not affected by personal emotions, biases, or subjective judgments, emphasizing factuality and accuracy, and capable of obtaining consistent results through repeated measurements or verifications. Such as multiple-choice questions, fill-in-the-blank questions, true or false questions, etc., including question content, students' answer records, and correct answers. (2) Subjective data: Data generated based on personal feelings, opinions, understandings, beliefs, or judgments, which vary from person to person. Such as short-answer questions, essay questions, course papers, project reports, etc., including text data (students' answer content), chart data (experimental results or analysis charts), and code data (programming tasks). (3) Numerical data: Data recorded and expressed in numerical form, capable of reflecting the quantitative characteristics, size relationships, or change trends of things. (4) Text data: Represented in the form of words, symbols, or images, usually used to describe the nature or category of things. (5) Behavioral data: Data obtained by observing, recording, and analyzing the behavioral patterns, behavioral frequencies, behavioral habits, and behavioral results shown by individuals in specific situations or daily life. Such as course video viewing duration, discussion participation times, assignment submission frequencies, learning time distribution, and task completion times. (6) Psychological data: Data related to an individual's mental state, mental processes, behavioral tendencies, or psychological characteristics, reflecting information about the individual in aspects such as cognition, emotion, motivation, personality, and behavior.

[0057] 1.3 Data Formats. (1) Historical Data: Data collected and recorded at a certain past time point or within a certain time period, reflecting past events, states, or behaviors, and usually used for analyzing historical trends, patterns, or regularities. Such as students' exam scores, homework completion status, classroom participation, online learning behavior data, etc.; (2) Real-time Data: Data generated, collected, and processed at the current moment or very close to the current moment, emphasizing the immediacy and timeliness of the data, and usually used in scenarios that require quick responses. During the learning process of students, the system records in real time the task completion time, learning behavior logs, discussion and collaboration records, and course learning progress to ensure the dynamic update of the data.

[0058] (II) System Operation

[0059] The learning situation evaluation subsystem described in 2.1 includes: a learning foundation analysis and learning interest mining module, a learning method identification and learning style diagnosis module, a learning habit monitoring and learning progress tracking module, and a learning efficiency evaluation.

[0060] 2.1.1 Analysis Process of the Learning Foundation Analysis and Learning Interest Mining Module: Combine the historical performance data and homework completion status of students through a weighted formula: F i = α·normalize(S i ) + β·normalize(A i ). Then introduce a knowledge graph to analyze the knowledge coverage of students, and construct an embedding vector ki = GNN(KG, Si) of students' knowledge reserves. Result Generation: Based on the knowledge reserves and performance of students, the system will generate a detailed learning foundation feedback report, indicating in which aspects students are weak and providing corresponding learning improvement plans. Then, calculate the participation degree according to the participation behaviors of students in different courses: Furthermore, generate an interest vector I i for each student. The formula is: I i = softmax(E i );, indicating their participation interest tendency in different courses; each dimension of the vector represents the interest intensity in a subject area. Through the softmax function, ensure that the values of the interest vector are within the range of [0, 1], reflecting the relative intensity of different subject interests. Result Generation: Based on the interest vector, the system will generate personalized learning resource recommendations to help students deeply learn in the fields they are interested in.

[0061] 2.1.2 Analysis Process of the Learning Method Identification and Learning Style Diagnosis Module: Give the interactive data of students' learning to generate a learning method feature vector: f i = [Discussion i , Interaction i,Duration i ; Further, use the decision tree classification algorithm to automatically classify the learning styles of students: way i = DecisionTree(f i ). Result generation: After identifying the learning styles of students, the system will automatically generate personalized feedback. For example, students who prefer cooperative learning will receive more recommendations for group projects or discussion groups. Construct a learning style feature vector based on the student's learning foundation, interests, habits, and styles, etc.: m i = [v i ,a i ,t i ,o i ,i i ,]. Furthermore, use the support vector machine to identify the learning styles of students: style i = SVM(m i ). Result generation: Based on the classification results of the learning styles, the system will recommend learning resources that match the student's style. For example, visual learners will receive more video resource recommendations, while auditory learners will receive more audio resource recommendations.

[0062] Analysis process of the learning habit analysis and learning progress tracking module described in 2.1.3: Assume that the learning duration of a student in a day is T i (t), and the learning frequency is f i (t). Through Fourier transform: Analyze the periodicity of the learning frequency; further, classify the learning habits through the clustering algorithm: [habit i = K-means(T i ,F i ). Result generation: The system generates learning strategy suggestions that adapt to the student's habits according to the learning habit categories of the students. For example, for "low-frequency long-duration learners", it is recommended to optimize the learning time allocation and improve learning efficiency, etc. Calculate the course learning progress of student A and the homework submission situation Furthermore, analyze the time for the student to complete different tasks to identify the bottlenecks in the learning process. The implementation method is: The time for the student to complete the i-th task is T i , by comparing with the average completion time T avg , define the speed score S speed (i) as: Result generation: Generate a student learning progress chart and table, and provide personalized learning plan adjustment suggestions, such as reminding the student to speed up the progress or adjust the learning plan.

[0063] Analysis process of the learning efficiency analysis module described in 2.1.4: First, define a matching degree scoring function Mscore (i) Measure the degree of match between the methods and strategies adopted by the student in the i-th learning task and the objectives of that task. Assume that the degree of match of the student in the i-th task is represented by M score (i), and the difficulty of the task objective is represented by D target (i). Then, the score adjusted by the degree of match and difficulty for each task can be defined as: M adj (i). The higher M efficiency (i) is, the more closely the student's learning methods and strategies match the task objectives, and the better the performance of the strategy for tasks of different difficulties. Secondly, evaluate the efficiency and effectiveness of the student in achieving the objectives. Define the efficiency score E actual (i) of the student in completing the task, which can be based on the ratio of the actual completion time T expected (i) of the student in the task and the expected completion time T result (i), as well as the quality Q of the task result. The efficiency score formula can be written as: result where Q actual (i) represents the quality of the student's result in the i-th task, T expected (i) is the time actually taken by the student to complete the task, and T efficiency (i) is the reasonable time estimated by the system for completing the task. The higher the efficiency score E

[0064] (i), the higher the efficiency of the student in achieving the learning objective. Further, comprehensively evaluate the overall effectiveness of the student in a certain type of task. We combine the degree of match and efficiency score, and define the comprehensive learning effectiveness score L efficiency (C j ) to evaluate the performance of the student in the task category C j : Through this comprehensive score, the overall learning effectiveness of the student in a specific task category can be evaluated. Further, bottleneck analysis. The task bottleneck identification technology will analyze in which types of tasks the student encounters bottlenecks by combining task types (such as comprehension tasks, memory tasks, application tasks, etc.). The learning tasks are divided into different categories C j through cluster analysis, and the average speed of the student under each category is calculated: By comparison, the system can identify in which types of tasks the student performs poorly, and then find out the possible task bottlenecks.

[0065] Result generation: Through the comprehensive learning effectiveness score L efficiency (C j ), the overall learning effectiveness of the student in a specific task category can be clearly evaluated. When L efficiency (C j) When the value is high, it indicates that the learning efficiency of the student in this task category is in a good state, and both the matching degree of the learning method and the goal and the efficiency and effect of achieving the goal are excellent; if L efficiency (C j ) has a low value, it means that there may be problems such as inappropriate learning methods and low efficiency for the student in this task category.

[0066] The comprehensive ability evaluation subsystem described in 2.2 specifically includes a teamwork ability evaluation module, a communication ability evaluation module, and a problem-solving ability evaluation module.

[0067] The analysis process of the teamwork ability evaluation module described in 2.2.1: First step, preprocess the data, including: standardization of collaboration scores, segmentation of time variables, and division of treatment groups and control groups;

[0068] Standardization of collaboration scores: Due to differences in task complexity, scale, etc. among different teams, there may be large differences in the absolute values of collaboration scores. Before using the difference-in-differences model, standardize the collaboration scores to ensure the comparability of scores between different teams. The standardization formula is expressed as: Segmentation of time variables: To better capture the changes in teamwork ability before and after the intervention, it is necessary to segment the time variable. Select a reasonable time window (such as weekly or monthly) for data segmentation to obtain multi-period data before and after the intervention measures; Post t represents the time dummy variable; Division of treatment groups and control groups: In the team, some teams are selected as the intervention group and other teams as the control group; the intervention group variable Treated i is used to indicate whether the team has received the intervention measure;

[0069] Second step, use the difference-in-differences model (DID) to evaluate the change in the collaboration performance of different teams before and after the intervention measure to estimate the causal effect of the intervention, expressed as: Y it = α + β1·Post t + β2·Treated i + δ·(Post t × Treated i ) + ∈ it , where Y it represents the collaboration ability score of team i at time t (such as the score obtained through models such as TACM, IQFM, RILM, etc.); α represents the intercept term in the model, that is, when all independent variables (such as Post t and Treated i ) are zero, the benchmark value of the teamwork ability score; β1 represents the time dummy variable Post tThe coefficient of it reflects the impact of time factors on the collaborative ability score in the absence of intervention measures; β2 represents the treatment group dummy variable Treated i The coefficient of it reflects the baseline difference between the teams receiving the intervention and the control group when there is no change in time factors; Post t is a time dummy variable indicating whether time t is after the implementation of the intervention measure. If it is, Post t = 1; if it is not, Post t = 0; Treated i is a treatment group dummy variable indicating whether team i has received the intervention measure. If it has, Treated i = 1; otherwise, Treated i = 0; Post t ×Treated i is the interaction term that captures the impact of the intervention on the collaborative ability; δ is the difference-in-differences estimator that reflects the actual impact of the intervention measure on the collaborative ability score; ∈ it is the error term;

[0070] Third step, estimation of the intervention effect: Evaluate the effect of the intervention measure through the coefficient δ of the interaction term in the difference-in-differences model; this coefficient reflects the actual impact of the intervention measure on the team's collaborative ability. If δ > 0, it indicates that the intervention measure has effectively improved the team's collaborative ability. If δ ≤ 0, the effect of the intervention measure is not obvious, and it may even have a negative impact. The estimation formula for the intervention effect is: Among them, represents the estimated value of the actual impact of the intervention measure on the collaborative ability score. It is the coefficient estimated value of the interaction term Post t ×Treated i in the difference-in-differences model, which evaluates the net effect of the intervention measure; N T and N C are the numbers of teams in the treatment group and the control group respectively; Y i,Post and Y i,Pre represent the collaborative ability scores after and before the intervention respectively;

[0071] Fourth step, visualization of the intervention effect: including the collaborative score trend chart and the effect comparison bar chart. Collaborative score trend chart: Plot the collaborative ability scores of the intervention group and the control group over time as a line chart to observe the trend of score changes before and after the intervention, and intuitively display the effect of the intervention. Effect comparison bar chart: Draw bar charts showing the changes in the collaborative ability scores of the intervention group and the control group before and after the intervention, and further display the effect of the intervention by comparing the differences between groups;

[0072] Step 5, Robustness Analysis of Intervention Effects: It includes parallel trend test, sensitivity analysis, and placebo test. The specific steps are as follows: Parallel trend test: The difference-in-differences model assumes that the change trends of the collaboration scores of the treatment group and the control group before the intervention are the same. This hypothesis is verified by testing whether the change trends of the scores of the two groups before the intervention are consistent; if the collaboration score trends of the two groups before the intervention are significantly different, the results of the difference-in-differences model are unreliable, which means that in this case, the conclusion about the impact of the intervention measure on the collaboration ability obtained based on this model may be biased. At this time, it is possible to consider re-screening the samples and selecting teams with more similar change trends of collaboration scores before the intervention as the treatment group and the control group, or further adjusting the data, such as using a more refined time segmentation method, to make the change trends of the collaboration scores of the two groups before the intervention tend to be consistent as much as possible, so as to ensure the reliability of the difference-in-differences model results.

[0073] Sensitivity analysis: By adjusting different time windows or different methods of standardizing collaboration scores, observe whether the intervention effect changes significantly to verify the robustness of the model;

[0074] Placebo test: In the absence of actual intervention, randomly select some teams as the virtual intervention group, run the difference-in-differences model, and check whether a significant effect appears; if the effect of the virtual intervention group is not significant, it means that the actual intervention effect is relatively credible;

[0075] Step 6, Final Intervention Effect Score: After completing the model analysis and robustness test, obtain the final effect score of the intervention measure on the team collaboration ability; this score is used to measure the effectiveness of the intervention measure and provide decision-making support for future optimization. The formula for the final intervention effect score is: where, is the intervention effect estimated by the difference-in-differences, Robustness check is the result of the robustness test, and λ1 and λ2 are weight parameters used to balance the intervention effect and robustness.

[0076] Result generation: Intuitive presentation of intervention effects: Through the collaboration score trend chart and the effect comparison bar chart, relevant personnel such as team managers and educators can intuitively obtain the change situation of the team collaboration ability. For example, in the collaboration score trend chart, if the broken line of the intervention group shows an obvious upward trend after the intervention, while the control group is relatively stable, it can clearly show the positive effect of the intervention measure on the improvement of the team collaboration ability; in the effect comparison bar chart, the difference in scores before and after the intervention of the intervention group forms a sharp contrast with the control group, further highlighting the intervention effect.

[0077] Quantify the intervention effect score: The final intervention effect score S TCIEM provides a quantitative basis for the effectiveness of the intervention measure. When S TCIEMWhen the value is relatively high, it indicates that the intervention measure has significantly improved the team collaboration ability, and the intervention measure can be considered for promotion or optimized application in the future; if S TCIEM When the value is relatively low or even negative, it shows that the intervention measure is not effective, and it is necessary to re-examine the intervention plan and adjust the intervention strategy, such as changing the intervention method, adjusting the intervention time, etc., to provide strong decision-making support for subsequent measures to improve team collaboration ability.

[0078] Analysis process of the communication ability evaluation module described in 2.2.2: First step, definition of model index variables: Term coverage rate - central idea, written genre, disciplinary norms; Communication fluency - communication purpose, quality of arguments, supporting materials; Emotional stability - word choice, language organization, communication skills;

[0079] Second step, the structural equation analysis of interdisciplinary communication ability includes both a measurement model and a structural model. The measurement model defines the relationship between the latent variable η and the observed variables of term coverage rate X1, communication fluency X2, and emotional stability X3:

[0080] X1 = λ1·η + ∈1

[0081] X2 = λ2·η + ∈2

[0082] X3 = λ3·η + ∈3

[0083] Among them, λ1, λ2, λ3 are loading coefficients, indicating the contribution degree of each observed index to the latent variable, and ∈1, ∈2, ∈3 represent error terms;

[0084] The structural model defines the relationship between latent variables: Taking interdisciplinary communication ability as the latent variable, the structural model is expressed as: η = γ1·ζ, where ζ is an external interference factor (such as personal experience, team cooperation efficiency, etc.), and γ1 is the influence coefficient of the external interference factor, indicating the influence of external factors on communication ability;

[0085] Third step, the final comprehensive scoring formula: According to the structural equation analysis of interdisciplinary communication ability, the final scoring formula of interdisciplinary communication ability is expressed as: CommunicationScore = λ1·X1 + λ2·X2 + λ3·X3 + γ1·ζ.

[0086] Term coverage: It represents the breadth of students' use of domain-related terms in interdisciplinary communication. Term coverage can reflect the breadth of students' knowledge and their mastery of multi-domain knowledge in interdisciplinary communication. Through the load factor, we can determine the degree of influence of term coverage on interdisciplinary communication ability. Communication fluency: The expression fluency of students is evaluated by factors such as speech rate, pauses, and voice stability. The load factor can quantify the contribution of communication fluency to the overall communication ability. Emotional stability: The emotional fluctuations of students during communication are judged through voice emotion analysis. The load factor is used to evaluate the influence of emotional stability on communication ability.

[0087] External interference factors (ζ): These factors may include students' personal backgrounds, the quality of project cooperation, learning environments, etc. Through the coefficient γ1, we can measure the overall influence of these external factors on interdisciplinary communication ability.

[0088] Result generation:

[0089] Quantitative scoring of communication ability: The score of interdisciplinary communication ability obtained according to the final comprehensive scoring formula accurately quantifies the ability level of students in interdisciplinary communication. This score can be used for horizontal comparison among students, helping teachers understand the strengths and weaknesses of different students in terms of term coverage, communication fluency, and emotional stability. It can also be used for students' own longitudinal comparison to track the development and changes of their interdisciplinary communication ability at different stages.

[0090] Analysis feedback on ability structure: By measuring the load coefficients λ1, λ2, λ3 in the measurement model and structural model, as well as the influence coefficient γ1 of external interference factors, we can deeply analyze the degree of influence of each observed variable and external factors on interdisciplinary communication ability. If the load coefficient λ1 of term coverage is large, it indicates that when improving students' interdisciplinary communication ability, we should focus on expanding the breadth of students' knowledge and enriching their reserves of domain-related terms. If the influence coefficient γ1 of external interference factors is large, it prompts us to pay attention to and optimize external factors such as personal experience and team cooperation efficiency, providing targeted improvement directions for enhancing students' communication ability.

[0091] Analysis process of the problem-solving ability evaluation module described in 2.2.3: First step, define the feature vector: The feature set of students in solving a certain problem is X = [x1, x2, …, x n , and each x i represents features in different dimensions; x1: Score of problem decomposition ability; x2: Score of strategy generation ability; x3: Score of solution optimization; x4: Diversity score; x5: Innovation score;

[0092] Second step, multi-layer linear mapping and feature weighting. Perform linear mapping on the features in each dimension to generate new feature representations. The formula for the first-layer linear transformation is as follows: Among them, and are the weight matrix and bias term of the i-th feature respectively;

[0093] Through the above linear mapping, the initial feature vector X is transformed into the first-layer hidden representation Introduce multiple layers of linear transformations to capture the interactions and relationships between features. The linear transformation formula for the l-th layer is: Among them, and are the weight matrix and bias term of the (l + 1)-th layer respectively. Through layer-by-layer linear combination, a global comprehensive representation, the global feature vector

[0094] The third step, comprehensive score calculation: After multiple layers of linear mapping, map the final global feature vector to a comprehensive score S final , representing the overall problem-solving ability of the student; the formula is as follows: Among them, and b (L) are the weights and bias terms of the last layer. Through this formula, the model can effectively fuse the ability scores of different dimensions to obtain a global comprehensive evaluation result;

[0095] The fourth step, training and optimization of the model. Train the multi-dimensional problem-solving ability linear fusion model by minimizing the mean square error (MSE) loss function. When the true ability score of the student is t true , and the score predicted by the model is S final , then the loss function is: Among them, N is the number of samples, and θ contains all the model parameters.

[0096] Result generation: Comprehensive ability score output: The comprehensive score S output by the model final Comprehensively reflects the problem-solving ability of the student. Teachers judge the comprehensive performance of students in aspects such as problem decomposition, strategy generation, solution optimization, diversity, and innovation based on this score, providing an objective basis for the evaluation of students' learning achievements; in the scenarios of enterprise talent selection or employee training effect evaluation, this score can also be used to screen out talents with strong problem-solving abilities or evaluate the ability improvement of employees after training.

[0097] Model Optimization and Capacity Enhancement: During the model training and optimization process, by continuously adjusting parameters to minimize the mean squared error loss function, not only the accuracy of scoring is improved, but also the potential relationships between features in each dimension can be deeply explored. Based on the optimized model, it is possible to more accurately identify the weak links in students' problem-solving abilities, and then design targeted training programs or learning plans. For example, for students with weak problem decomposition abilities, more case studies and exercises on problem decomposition are provided to help students improve their problem-solving abilities.

[0098] (III) Result Report

[0099] 3.1 Report Form. (1) Chart-based Display: The system uses intuitive visualization methods to convert complex data into concise and clear charts. For example, a radar chart is used to display students' learning interest areas, learning habit characteristics, and learning style categories; a grid chart is used to present the coverage of knowledge points; a flow chart is used to analyze the paths of knowledge transfer and application. These charts can help users quickly understand the data results and extract key information. (2) Text Description: The system provides detailed and clear text explanations to professionally interpret each data result. The text description not only explains the meaning behind the data but also combines the actual situations of different users to report targeted conclusions and suggestions. This combination of graphics and text not only meets the need for intuitive understanding of data but also provides users with in-depth analysis and improvement directions. (3) Hierarchical Reporting: The system generates reports at different levels according to the needs of users, including students' personalized learning and comprehensive ability reports, teachers' overall assessment reports of class teaching, school administrators' global analysis reports, and integrated reports on students' learning quality required for third-party evaluations, etc.

[0100] 3.2 Report Content. (1) Report for students: The system focuses on demonstrating the specific performance of their teamwork ability, communication ability, and problem-solving ability, and provides suggestions for optimizing the learning path; the system also generates an encouraging summary to stimulate students' learning motivation. (2) Report for teachers: The system summarizes the overall performance of the class students' abilities and marks the ability distribution of the whole class; combined with data analysis, it generates scientific suggestions for teaching improvement, such as "strengthening the promotion of project-based learning models" or "optimizing the curriculum content of a certain grade, paying attention to the coherence of ability cultivation, and strengthening the integration of interdisciplinary knowledge". (3) Report for administrators: The system provides school-level or regional-level analysis, shows the differences in comprehensive abilities among disciplines and grades, and proposes directions suitable for overall teaching improvement, such as "increasing practical links for application-oriented tasks" or "strengthening the relevant teaching of interdisciplinary abilities", so as to provide a basis for the optimal allocation of educational resources. (4) Report for parents: Based on the 10 modules of the system, it intuitively shows the strengths and weaknesses of students through radar charts, bar charts, flowcharts, etc.; it helps parents clearly understand the specific development of students in aspects such as communication ability and digital literacy, and provides specific suggestions for parents to support the all-round development of their children. (5) Report for third parties: This system provides third parties with comprehensive data-driven reports and decision support tools. By analyzing the learning conditions and comprehensive abilities of different student groups, it generates distribution maps of each module and summarizes the learning efficiency of the class or school; based on big data analysis, the system provides suggestions for optimizing educational resources, including adjusting curriculum investment, introducing interdisciplinary teaching practices, and personalized support measures; for regional evaluations, the system compares the effects of cultivating the comprehensive abilities of students in different schools or grades, helps formulate scientific educational resource allocation policies, and provides important support for improving educational quality, fairness, and resource optimization.

[0101] 3.3 Delivery Methods (1) Online display: Through the system port, students, teachers, administrators, parents, and third parties can view the assessment data in real time and conduct dynamic tracking. (2) Regular reports: The system supports the regular generation of reports, which are sent to users via email or their own accounts in the form of WORD and PDF. (3) Real-time reminders: Through the message push function, the system sends real-time reminders of learning progress, knowledge point mastery, and learning suggestions to students and teachers. Through the above comprehensive and professional reporting methods, the system ensures that various users can accurately obtain the information they need and apply the assessment results to the improvement of the learning and teaching process.

[0102] (IV) Introduction to Embodiments

[0103] Student Li Ming is taking a physics course, "General Physics", whose content covers knowledge points such as mechanics, thermodynamics, electromagnetism, etc. The assessment methods of this course include online exams, in-class discussions, experimental reports, project assignments, etc. The system will evaluate his learning situation and comprehensive ability in this course through the system.

[0104] 4.1 Analysis of learning foundation. Analyze Li Ming's historical exam scores and homework completion in the "General Physics" course, and calculate Li Ming's learning foundation score F i . Suppose Li Ming's scores in several exams in the mechanics part are average, but he performs well in the electromagnetism part. If Li Ming's scores in the mechanics part are poor, the system will generate a prompt, pointing out specific weak points in his mechanics thinking, and prompting him to review and strengthen these contents.

[0105] 4.2 Diagnosis of interest tendency. Analyze Li Ming's participation behaviors in the course (such as video viewing duration, classroom participation, etc.) to generate his interest vector I i . For example, if Li Ming spends a lot of time on the relevant chapters of mechanics and actively participates in course discussions, the system will calculate his interest tendency vector I i = softmax(E i ), indicating that he has a higher interest in the mechanics part. Accordingly, the system will recommend more learning resources related to mechanics to Li Ming, such as in-depth analysis videos, practice exercises, etc., to help him further improve his mastery of this part of knowledge. In addition, the system also calculates his participation degree based on Li Ming's classroom participation behaviors . Suppose Li Ming participates in class discussions more frequently and the viewing duration of teaching videos reaches 80% of the maximum duration. The system will infer that he has a relatively high overall interest in the physics course and suggest that he maintain this participation frequency, which helps to continuously improve learning effects.

[0106] 4.3 Identification of learning methods. Li Ming participates in discussions less in the course, but prefers to complete homework independently and is more inclined to watch videos alone during online learning. The system will extract his learning method feature vector f i = [Discussion i , Interaction i , Duration i . Through the decision tree model, the system classifies Li Ming as a student who prefers independent learning way i = DecisionTree(f i ). Based on this classification, the system will encourage Li Ming to participate in more course discussions and cooperative learning to improve his collaboration ability.

[0107] 4.4 Learning Style Diagnosis. When Li Ming studies, he uses video resources more (such as watching course explanation videos) and text materials less. The system constructs a learning style feature vector based on the usage frequencies of his videos, audio, text, and practical operations. Using a support vector machine, the system identifies that Li Ming belongs to a visual learner. Therefore, the system will recommend more video resources to Li Ming, reduce the push of text materials, and at the same time suggest that he appropriately increase practical operations (such as experimental courses) to improve his hands-on ability and the application ability of theoretical knowledge.

[0108] 4.5 Learning Habit Monitoring. The system performs a Fourier transform on Li Ming's learning frequency f i (t) to extract his main learning frequency and periodic characteristics: Li Ming's learning time is mainly concentrated in the evening, and his learning frequency is relatively low within a week. The system further uses a clustering algorithm to classify Li Ming's learning habits. For example, Li Ming's learning duration T is relatively concentrated, but his learning frequency is low. The system classifies him as a "high-intensity, low-frequency" learner. Based on this result, the system suggests that Li Ming appropriately increase his learning frequency in his study plan to avoid the negative impacts brought by over-relying on cramming, and suggests that he try a more balanced study arrangement. i 4.6 Learning Progress Tracking. The system tracks Li Ming's learning progress in real time. Assuming that Li Ming has completed 5 out of 8 courses, the system will calculate his learning progress

[0109] At the same time, the system also monitors Li Ming's homework submission situation. If Li Ming has submitted 3 out of 4 homework assignments in the electromagnetism part, the system will prompt his homework submission progress Further, the system generates a report showing Li Ming's learning progress and homework completion situation, and puts forward specific learning suggestions for the parts with lagging progress (such as unfinished mechanics homework).

[0110] ​4.7 Evaluation of teamwork ability. In the research of science and technology festival projects, students completed small science and technology projects in groups. The group where Li Ming was located was set as the intervention group, and the teacher took a series of intervention measures. In terms of task allocation, it was reasonably arranged according to the students' special strengths. For example, Li Ming, who had an interest and foundation in computer programming, was responsible for program writing. In terms of participation, an open and inclusive discussion atmosphere was created to encourage everyone to actively speak up and put forward ideas. Li Ming actively participated in each group discussion and contributed ideas. The teacher also developed detailed criteria to measure individual contribution. Some other groups served as the control group and organized themselves conventionally. Data such as task allocation, participation, and individual contribution in the activity were entered into the system. After the activity, the data was preprocessed. In the collaborative scoring standardization link, different standard scores were unified through a specific formula to make the scores of each team comparable; the time variable was segmented by week according to the natural cycle of the activity to obtain multi-period data. After determining the treatment group and the control group, the difference-in-differences model was used for analysis. The results showed that after the intervention, the efficiency of the group where Li Ming was located in collaborating to solve problems increased significantly. The members changed from working separately to quickly discussing solutions and finding breakthrough points. The communication was smoother, reducing misunderstandings and duplicate work, indicating that the intervention measures had a positive effect on improving teamwork ability. It was visually demonstrated by drawing a collaborative scoring trend chart and an effect comparison bar chart. The former showed that the score of the intervention group increased over time while the change of the control group was flat, and the latter highlighted the improvement amplitude of the intervention group. After robustness analyses such as parallel trends test, sensitivity analysis, and placebo test, there was no significant difference in the change of the collaborative scores of the two groups before the intervention in the parallel trends test, which met the model assumptions and confirmed the reliability of the results. Finally, an accurate intervention effect score was given, providing strong data support for the organization of subsequent similar activities and the optimization of intervention measures.

[0111] 4.8 Evaluation of communication ability. In the interdisciplinary group discussion themed "Scientific and technological applications in environmental protection and sustainable development", which involved knowledge of multiple disciplines such as physics, chemistry, biology, and geography, Li Ming participated. The system analyzed his performance from multiple aspects. In terms of term usage, it was examined whether he could accurately use the standard terms of each discipline to express his ideas. For example, when talking about the physical principles of sewage treatment, whether he could correctly say terms such as "filtration", "sedimentation", "adsorption", etc. and explain their functions. In terms of expression fluency, it was observed whether he spoke in an organized manner and could coherently elaborate on his viewpoints and ideas. For example, when talking about biological means of soil pollution control, whether he could clearly explain from the principles, implementation steps to the expected effects in sequence, and whether there were frequent pauses, repetitions, or failure to convey the intended meaning. Emotional stability was also taken into consideration, paying attention to whether he got anxious or excited due to different opinions or minor disagreements, which affected his expression.

[0112] Through the structural equation analysis of interdisciplinary communication ability, integrating these indicators, assigning corresponding weights, and calculating according to the established method, the ability score of Li Ming's interdisciplinary communication this time was obtained. With this score, the teacher, Li Ming, and other students could clearly know his performance in interdisciplinary communication, identify his strengths and weaknesses, and facilitate targeted improvement in the future.

[0113] 4.9 Evaluation of problem-solving ability. During the preparation for the mathematical modeling competition organized by the school, Li Ming encountered a practical mathematical application problem of optimizing bus route settings based on urban traffic flow data to improve traffic efficiency and reduce congestion. The system considered his problem-solving process from multiple dimensions.

[0114] In terms of the rationality of problem decomposition, Li Ming decomposed the complex problem into simple sub-problems, sorted out the traffic flow data by different time periods and road sections, and analyzed the congestion nodes on each road section, demonstrating reasonable problem decomposition ability. In terms of the quantity and effectiveness of strategies, he proposed various strategies such as increasing the number of bus trips during peak hours, adjusting the routes of some bus lines, and setting up bus lanes. He also confirmed through consulting materials and simple simulation calculations that these strategies could relieve congestion to a certain extent, indicating the effectiveness of the strategies. In terms of the optimization degree of the solution, he further optimized the strategies by considering factors such as cost, implementation difficulty, and impact on the surrounding environment, determined the appropriate road sections for setting up bus lanes, and balanced the running speed of buses and the passage of social vehicles. In terms of the diversity and innovation of the solution, in addition to the conventional adjustment of routes and increase in trips, he innovatively proposed to use the intelligent transportation system to schedule buses in real time and dynamically adjust the stopping stations and running times according to the real-time traffic flow.

[0115] After the system performs operations such as linear fusion, multi-layer linear mapping, and feature weighting on these dimensional features, it obtains the comprehensive ability score of Li Ming's problem-solving process this time. This score can not only intuitively reflect his overall problem-solving ability level, but also analyze that he has strong abilities in terms of strategy effectiveness and innovation. At the same time, it is clear that there is still room for improvement in dealing with more complex multi-factor comprehensive optimization problems, and he can improve by deeply learning relevant knowledge and methods, so as to continuously improve his own problem-solving ability.

[0116] The content not described in detail in the specification of the present invention belongs to the prior art well-known to those skilled in the art. Although the above describes the illustrative specific implementation manners of the present invention for the convenience of those skilled in the art to understand the present invention, it should be clear that the present invention is not limited to the scope of the specific implementation manners. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention defined and determined by the appended claims, these changes are obvious, and all inventions and creations using the concept of the present invention are within the scope of protection.

Claims

1. An artificial intelligence-based comprehensive student ability evaluation system, characterized in that: The system includes: a learning situation evaluation subsystem and a comprehensive ability evaluation subsystem; The learning situation evaluation subsystem includes: a learning foundation analysis and learning interest mining module, a learning method identification and learning style diagnosis module, a learning habit monitoring and learning progress tracking module, and a learning efficacy evaluation module; The comprehensive ability evaluation subsystem includes: a teamwork ability evaluation module, a communication ability evaluation module, and a problem-solving ability evaluation module; The learning foundation analysis and learning interest mining module: Analyze the learning historical data of students based on an artificial intelligence-based learning foundation evaluation model to evaluate the current knowledge reserve of students; Analyze their interest tendencies in different disciplines based on an artificial intelligence-based interest mining evaluation model, and track the participation of students in different courses and contents; The learning method identification and learning style diagnosis module: Based on an artificial intelligence-based learning method identification model, identify the learning methods of students according to their learning interaction data; Based on an artificial intelligence-based learning style diagnosis model, use support vector machines to identify the types of students' learning styles - visual, auditory, kinesthetic, and integrative, in order to provide personalized teaching resources; The learning habit monitoring and learning progress tracking module: Based on an artificial intelligence-based learning habit monitoring model, use time series analysis technology to evaluate the time arrangement and learning frequency of students, and monitor whether their learning habits are suitable for their learning goals; Based on an artificial intelligence-based learning progress tracking model, analyze the course learning situation and homework submission situation of students, generate a real-time progress report, and help teachers and students understand the completion situation of current learning and track the progress of students in each course learning; The learning efficacy evaluation module: Based on an artificial intelligence-based learning efficacy evaluation model, evaluate the degree of matching between the methods and strategies adopted by students in different learning tasks and the goals, as well as the efficiency and effect of achieving the expected learning goals; The teamwork ability evaluation module: In a teamwork environment, introduce artificial intelligence-based intervention measures to improve team performance, and use the difference-in-differences model DID to analyze the changes in the collaboration performance of different teams before and after the introduction of intervention measures to estimate the causal effect of the intervention; The communication ability evaluation module: Use the structural equation of interdisciplinary communication ability to analyze the comprehensive impact of different evaluation indicators on interdisciplinary communication ability, and improve the accuracy of evaluation through artificial intelligence technology; The problem-solving ability evaluation module: Through a multi-dimensional problem-solving ability linear fusion model, perform non-linear and linear fusion on multiple dimensional characteristics of students' problem-solving to obtain the final comprehensive score, and conduct a detailed and comprehensive evaluation of students' problem-solving ability.

2. The comprehensive ability evaluation system for students based on artificial intelligence according to claim 1, wherein: Specifically, the teamwork ability evaluation module is as follows: First, preprocess the data, including: standardization of collaboration scores, segmentation of time variables, and division of treatment groups and control groups; Standardization of collaboration scores: Standardize the collaboration scores to ensure the comparability of scores between different teams. The standardization formula is expressed as: Among them, Y' it is the collaborative score after standardization, is the average collaborative score of Team i, and σ i is the standard deviation of the collaborative score of Team i; Segmentation of time variables: Select a reasonable time window for data segmentation to obtain multi-period data before and after the intervention measures; Post t Indicates a time dummy variable; Treatment group and control group division: In the team, some teams were selected as the intervention group and the other teams as the control group; the variable Treated in the intervention group i is used to indicate whether the team has received the intervention measure; Then, use the difference-in-differences model DID to evaluate the changes in the collaboration performance of different teams before and after the intervention measures to estimate the causal effect of the intervention, which is expressed as: Y it = α + β1·Post t + β2·Treated i + δ·(Post t × Treated i ) + ∈ it Among them, Y it represents the collaborative ability score of team i at time t; α represents the intercept term in the model, that is, the benchmark value of the team collaborative ability score when all independent variables are zero; β1 represents the coefficient of the time dummy variable Post t ; β2 represents the coefficient of the treatment group dummy variable Treated i ; Post t is a time dummy variable indicating whether time t is after the implementation of the intervention measure. If it is, Post t = 1, if not, Post t = 0; Treated i is a treatment group dummy variable indicating whether team i has received the intervention measure. If it has, Treated i = 1, otherwise Treated i = 0; Post t ×Treated i is the interaction term, capturing the impact of the intervention on the collaborative ability; δ is the difference-in-differences estimator, reflecting the actual impact of the intervention measure on the collaborative ability score; ∈ it is the error term; Secondly, estimation of the intervention effect: Evaluate the effect of the intervention measure through the interaction term coefficient δ in the difference-in-differences model; this coefficient reflects the actual impact of the intervention measure on the team collaboration ability; if δ > 0, it indicates that the intervention measure has effectively improved the team collaboration ability; if δ ≤ 0, the effect of the intervention measure is not obvious or even has a negative impact; the estimation formula of the intervention effect is expressed as: Among them, represents the estimated value of the actual impact of the intervention on the collaborative ability score; it is the coefficient estimate of the interaction term Post t ×Treated i in the difference-in-differences model, evaluating the net effect of the intervention; N T and N C are the number of teams in the treatment group and the control group respectively; Y i,Post and Y i,Pre represent the collaborative ability scores after and before the intervention respectively; Then, visualization of the intervention effect: including the collaboration score trend chart and the effect comparison bar chart; Collaboration score trend chart: Plot the collaboration ability scores of the intervention group and the control group over time as a line chart to observe the trend of score changes before and after the intervention, and intuitively display the effect of the intervention; Effect comparison bar chart: Draw a bar chart showing the changes in the collaboration ability scores of the intervention group and the control group before and after the intervention, and further display the effect of the intervention by comparing the differences between groups; Furthermore, robustness analysis of the intervention effect: including parallel trends test, sensitivity analysis, placebo test, and the specific steps are as follows: Parallel trends test: The difference-in-differences model assumes that the trends of collaboration score changes in the treatment group and the control group before the intervention are the same; verify this assumption by testing whether the trends of score changes in the two groups before the intervention are consistent; if the collaboration score trends of the two groups before the intervention are significantly different, the results of the difference-in-differences model are unreliable, indicating that in this case, the conclusion about the impact of the intervention measure on the collaboration ability obtained based on this model is biased; at this time, re-screen the samples, select teams with more similar collaboration score change trends before the intervention as the treatment group and the control group, or further adjust the data to make the collaboration score change trends of the two groups tend to be consistent before the intervention, so as to ensure the reliability of the results of the difference-in-differences model; Sensitivity analysis: By adjusting different time windows or different methods of standardizing collaboration scores, observe whether the intervention effect changes significantly to verify the robustness of the model; Placebo test: In the absence of actual intervention, randomly select some teams as the virtual intervention group, run the difference-in-differences model, and check whether a significant effect appears; if the effect of the virtual intervention group is not significant, it indicates that the actual intervention effect is relatively credible; Finally, the final intervention effect score: After completing the model analysis and robustness test, obtain the final effect score of the intervention measure on the team collaboration ability; this score is used to measure the effectiveness of the intervention measure and provide decision-making support for future optimization; The formula for the final intervention effect score is: Among them, is the intervention effect of the difference-in-differences estimation, and Robustness check is the result of the robustness test. λ1 and λ2 are weight parameters used to balance the intervention effect and robustness.

3. The comprehensive ability evaluation system for students based on artificial intelligence according to claim 1, characterized in that: The specific communication and interaction ability evaluation module is as follows: First, definition of model index variables: Term coverage rate - central idea, written genre, disciplinary norms; Communication fluency - communication purpose, quality of arguments, supporting materials; Emotional stability - word choice, language organization, communication skills; Then, the structural equation analysis of cross-disciplinary communication ability includes both a measurement model and a structural model; The measurement model defines the relationship between latent variables and observed variables: Define the relationship between the observed variables term coverage rate X1, communication fluency X2, and emotional stability X3 and the latent variable η: X1 = λ1·η + ∈1 X2 = λ2·η + ∈2 X3 = λ3·η + ∈3 Among them, λ1, λ2, and λ3 are load coefficients, representing the contribution degree of each observation index to the latent variable, and ∈1, ∈2, and ∈3 represent error terms; The structural model defines the relationship between latent variables: taking the interdisciplinary communication ability as the latent variable, the structural model is expressed as: η = γ1·ζ; Among them, ζ is an external interference factor, and γ1 is the influence coefficient of the external interference factor, representing the influence of external factors on the communication ability; Finally, the final comprehensive scoring formula: According to the analysis of the interdisciplinary communication ability structural equation, the final scoring formula of the interdisciplinary communication ability is expressed as: CommunicationScore = λ1·X1 + λ2·X2 + λ3·X3 + γ1·ζ.

4. An artificial intelligence-based comprehensive student ability evaluation system according to claim 1, characterized in that: The problem-solving ability evaluation module is specifically: First, define the feature vector: The feature set of a student in solving a certain problem is X = [x1, x2, …, x n , where each x i represents a feature in a different dimension; x1: Problem decomposition ability score; x2: Strategy generation ability score; x3: Solution optimization score; x4: Diversity score; x5: Innovation score; Then, multi-layer linear mapping and feature weighting: Perform linear mapping on the features of each dimension to generate a new feature representation. The formula for the first-layer linear transformation is as follows: Among them, and are the weight matrix and bias term of the i-th feature respectively; Through the above linear mapping, the initial feature vector X is transformed into the first-layer hidden representation Multiple layers of linear transformations are introduced to capture the interactions and relationships between features. The linear transformation formula for the l-th layer is as follows: wherein, and are the weight matrix and bias term of the (l + 1)-th layer respectively; Finally, a global comprehensive representation, the global feature vector, is generated through layer-by-layer linear combination. Secondly, comprehensive score calculation: After multi-layer linear mapping, the final global feature vector is mapped to a comprehensive score S final , representing the overall problem-solving ability of the student; the formula is as follows: Among them, and b (L) are the weights and bias terms of the last layer; Finally, training and optimization of the model: Train the multi-dimensional problem-solving ability linear fusion model by minimizing the mean square error MSE loss function; When the true ability score of a student is y true , and the score predicted by the model is S final , then the loss function is: Among them, N is the number of samples, and θ contains all model parameters; continuously update the model parameters through the gradient descent method or its variants to improve the accuracy of the score.

5. The comprehensive ability evaluation system for students based on artificial intelligence according to claim 1, characterized in that: The learning foundation analysis and learning interest mining module specifically combines the historical academic performance data S of students i and the completion situation A of usual homework i through a weighted formula: F i = α·normalize(S i ) + β·normalize(A i ); where F i is the learning foundation score of the student, α and β are adjustable weights, and the normalize function is used to standardize the scores to make them comparable across different time periods and subjects; introduce a knowledge graph to analyze the knowledge coverage of students and construct an embedding vector k of the student's knowledge reserve i = GNN(KG, S i ); where GNN is a graph neural network used to capture the relationships between knowledge points to more accurately evaluate the scope of students' knowledge mastery; generate an interest vector I for each student through an interest mining model i , representing their interest tendency in different courses; each dimension of the vector represents the interest intensity of the student in a subject area, and the calculation formula is: I i = softmax(E i ); E i represents the participation degree. Through the softmax function, ensure that the values of the interest vector are in the range of [0, 1] to reflect the relative intensity of the student's interest in different subjects; calculate the participation degree according to the student's participation behavior in different courses: where T i represents the duration of participating in the class, D i represents the number of times of participating in discussions, T max and D max are the maximum values of the class participation duration and the number of discussion participations respectively, and λ1 and λ2 are weight parameters 6. The comprehensive ability evaluation system for students based on artificial intelligence according to claim 1, characterized in that: The learning method recognition and learning style diagnosis module specifically extracts the data features of students' participation in discussions, interaction with peers, and independent thinking during the learning process, and generates a learning method feature vector: f i =[Discussion i , Interaction i , Duration i , where Discussion i represents the number of times a student participates in discussions, Interaction i represents the duration of a student's interaction with peers, and Duration i is the duration of a student's independent learning; the decision tree model will automatically identify the learning method of the student according to different learning behavior characteristics: way i =DecisionTree(f i ), where way i represents the learning method of the student; a learning style feature vector is constructed based on the student's learning foundation, interests, habits, and methods: m i =[v i , a i , t i , o i , i i , where v i represents the usage frequency of videos, a i represents the usage frequency of audio, t i represents the usage frequency of text, o i represents the frequency of hands-on operations, and i i represents the frequency of integrated learning; then, the support vector machine SVM is used to diagnose the learning style of the student: style i =SVM(m i ), where style i is the learning style category of the student.

7. The comprehensive ability evaluation system for students based on artificial intelligence according to claim 1, characterized in that: The learning habit monitoring and learning progress tracking module specifically sets the learning duration of a student in a day as T i (t), and the learning frequency as f i (t). The main factors of the learning frequency are extracted through Fourier transform: Among them, is the Fourier transform operation, which is used to analyze the periodicity of the learning frequency and time arrangement; the learning habits are classified using a clustering algorithm: habit i = K-means(T i , F i ), where habit i is the learning habit of the student; the learning progress tracking model based on artificial intelligence analyzes the student's course learning situation and assignment submission situation, generates a real-time progress report, and helps teachers and students understand the completion situation of the current learning. The formula for tracking the progress of the student in each course learning is: Among them, Ccourse i is the number of completed courses, and C total is the total number of courses; the formula for the assignment submission situation is: Among them, c is a specific course, J c,total is the total number of assignments that the student needs to complete for this course, J c,submitted is the number of assignments that have been submitted for this course, and P c,homework is the assignment submission progress for this course; by analyzing the time taken by the student to complete different tasks, the constraints and bottlenecks in the learning process are identified. Specifically, the time taken by the student to complete the i-th task is T i , and by comparing it with the average completion time T avg , the speed score S speed (i) is defined as: Among them, T avg is the average completion time of this task; S speed (i)>1 indicates that the student completes the task faster than the average level, and S speed (i)<1 indicates that the student completes the task slower than the average level.

8. The comprehensive ability evaluation system for students based on artificial intelligence according to claim 1, characterized in that: The learning efficiency evaluation module is specifically: Evaluation of the matching degree between methods, strategies, and goals: Define a matching degree scoring function M score (i), which is used to measure the degree of matching between the methods and strategies adopted by students in the i-th learning task and the goals of that task; The matching degree and the difficulty-adjusted score for each task are expressed as: where M score (i) represents the matching degree of the student in the i-th task; D target (i) represents the difficulty of the task goal; The higher M adj (i) is, the more matching the student's learning methods and strategies are with the task goals; Evaluate the efficiency and effectiveness of students in achieving their goals: Define the efficiency score E for students to complete tasks efficiency (i), this efficiency score is based on the actual completion time T of the student in the task actual (i) and the expected completion time T expected (i) ratio, as well as the quality Q of the task result result (i) to measure, the efficiency score formula is expressed as: Among them, Q result (i) represents the quality of the result of the student in the i-th task, T actual (i) is the time actually taken by the student to complete the task, T expected (i) is the reasonable time estimated by the system to complete the task; the efficiency score E efficiency (i) The higher it is, the higher the efficiency of the student in achieving the learning goal; Comprehensively evaluate the overall effectiveness of students in a certain type of task: Combine the matching degree and efficiency score to define the comprehensive learning effectiveness score L efficiency (C j ) for evaluating the performance of students in task category C j : where α and β are weight parameters used to balance the proportions of the matching degree and efficiency score in the comprehensive effectiveness, and N j is the number of tasks in task category C j ; Perform bottleneck analysis using task bottleneck identification techniques; the task bottleneck identification techniques will analyze in which types of tasks students encounter bottlenecks in combination with task types; first, classify learning tasks into different task categories C through cluster analysis j , and calculate the average speed of students in each category: where N j is the number of tasks in task category C j , and S speed (i) is the speed score of the student for completing the i-th task; by comparing the speed scores of different categories, the system identifies in which types of tasks the student performs poorly, and then finds the existing task bottlenecks.

Images