Student learning behavior prediction method based on artificial intelligence

Abstract

The invention belongs to the technical field of artificial intelligence education, and particularly relates to a student learning behavior prediction method based on artificial intelligence, and the method comprises the steps: obtaining student learning data and teacher teaching task data; constructing a student agent according to the student learning data, and constructing a teacher agent according to the teacher teaching task data; a dynamic teaching strategy generated by the teacher agent is input into the student agent, and the student agent predicts possible learning behavior change of the student based on a learned behavior mode in combination with a strategy-behavior causal model; the prediction result is linked with a teaching management mechanism, a dynamic learning file is generated, the teacher intelligent agent matches personalized learning resources through an intelligent recommendation algorithm according to the dynamic learning file, and the student intelligent agent receives feedback of students on the learning resources and performs targeted intervention actions in combination with feedback information. Therefore, the problems of insufficient causal reasoning ability, weak calibration ability, insufficient data fusion ability and the like in the prior art are solved.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence education technology, and specifically relates to a student learning behavior prediction method based on artificial intelligence. Background Art

[0002] With the development of smart education and the surge in demand for personalized teaching, the market urgently needs a high-precision, dynamically adaptable method for predicting student learning behavior. Current AI-based learning behavior analysis technology mainly relies on machine learning models, which achieve behavioral modeling by mining the statistical laws of historical learning data (such as grades and homework duration) to meet the needs of scenarios such as precise teaching and intervention strategy optimization.

[0003] However, traditional methods generally lack causal reasoning capabilities, capturing only data correlations but failing to parse the causal mechanisms between changes in teaching strategies and learning behaviors. This results in a lack of interpretability in prediction results. Furthermore, they suffer from weak dynamic calibration capabilities, relying on fixed model parameters and struggling to adapt to individual student differences and real-time adjustments to teaching strategies, requiring frequent manual model updates. Furthermore, they lack data fusion capabilities, focusing solely on structured data while ignoring the value of unstructured data such as text feedback and video behavior, resulting in a one-sided portrayal of behavioral characteristics. As the digitalization of education drives the demand for dynamic prediction and precise intervention, the limitations of traditional technologies in causal modeling, individual adaptability, and data utilization are becoming increasingly prominent, making them unable to meet the real-time analysis needs of complex teaching scenarios. Summary of the Invention

[0004] This application provides an artificial intelligence-based student learning behavior prediction method to solve the problems of insufficient causal reasoning ability, weak calibration ability, and insufficient data fusion ability in the existing technology.

[0005] The first aspect of the present application provides an artificial intelligence-based student learning behavior prediction method, comprising the following steps: acquiring student learning data and teacher teaching task data; constructing a student intelligent body based on the student learning data; inputting the dynamic teaching strategy generated by the teacher intelligent body into the student intelligent body, wherein the student intelligent body predicts possible changes in students' learning behaviors based on the learned behavior patterns and the strategy-behavior causal model, wherein the learning behaviors include changes in the quality of homework completion, fluctuations in class participation, and adjustments to the progress of knowledge mastery; linking the prediction results with the teaching management mechanism to generate a dynamic learning profile, wherein the teacher intelligent body matches personalized learning resources based on the dynamic learning profile through an intelligent recommendation algorithm, and the student intelligent body receives students' feedback on the learning resources and performs targeted intervention actions based on the feedback information.

[0006] Preferably, the constructing of a student intelligent body based on the student learning data includes: constructing a dynamic cognitive diagnosis algorithm; decoupling the learning data based on the dynamic cognitive diagnosis algorithm, and capturing the correlation between learning behavior and knowledge points through a dynamic knowledge graph; through a forgetting curve model, combined with the correlation, constructing a hierarchical student intelligent body that includes dynamic evolution of learning style, real-time monitoring of fatigue status, and intelligent positioning of knowledge blind spots.

[0007] Preferably, the dynamic cognitive diagnosis algorithm formula is: ,

[0008] Where P is the probability symbol; is the performance of student i on a certain learning task or test item j; is the attribute parameter of student i; is the comprehensive parameter related to test item j; , is the exponential parameter associated with student i; Indicates k traversal set All elements in is the degree of mastery of knowledge point k by student i or related parameters.

[0009] Preferably, the constructing of a student intelligent agent based on the student learning data includes: obtaining teaching data, wherein the teaching data includes explanation methods, homework assignments, and interactive forms; personalizing the teaching data and student characteristic data, and designing an action set centered on improving student learning outcomes; training an intelligent agent based on the action set in combination with historical teaching data, and through a proximal strategy optimization algorithm, enabling the intelligent agent to continuously trial and error in a simulated teaching environment, thereby constructing a teacher intelligent agent that can dynamically adapt to strategies.

[0010] Preferably, the proximal strategy optimization algorithm formula is: in, is the policy-dependent network parameter CLIP loss function; To find the expectation for time step t; is the importance sampling ratio; is the advantage function; is the clipping function; is a hyperparameter.

[0011] Preferably, the prediction results are linked with the teaching management mechanism to generate a dynamic learning archive, including: constructing a neural tensor network; reconstructing the knowledge graph of the learning behavior prediction results based on the neural tensor network, and constructing an intelligent archive model that includes ability shortcoming warning indicators, learning path optimization suggestions and cognitive development milestones; based on the intelligent archive model, combined with the microservice architecture, a dynamic learning archive is generated, and at the same time, the learning archive data and the teaching management platform data are collaboratively and hierarchically managed according to the corresponding permissions.

[0012] Preferably, the student intelligent agent receives students' feedback on learning resources and performs targeted intervention actions based on the feedback information, including: constructing an emotional computing model; parsing the implicit needs in the feedback data through the emotional computing model, and dynamically adjusting the intervention strategy using the Pareto algorithm; based on the adjusted intervention strategy, generating a set of intervention actions including intelligent reminder triggering rules, intelligent reconstruction of learning paths, and dynamic balance of cognitive load for different students and learning scenarios.

[0013] Preferably, the student agent predicts possible changes in students' learning behavior based on the learned behavior patterns in combination with the strategy-behavior causal model, including: obtaining historical teaching strategy data and student learning behavior data; based on the historical teaching strategy data and student learning behavior data, using a causal inference algorithm to analyze the causal relationship between teaching strategies and student learning behaviors, and establishing a strategy-behavior causal model; based on the strategy-behavior causal model, combined with the counterfactual reasoning mechanism, by simulating the intervention effect of a specific teaching strategy, predicting possible changes in students' learning behavior, and comparing the deviation between the predicted results and the actual observed data.

[0014] The second embodiment of the present application provides a student learning behavior prediction system based on artificial intelligence, including: an acquisition module for acquiring student learning data and teacher teaching task data; a construction module for constructing a student intelligent body based on the student learning data; a prediction module for inputting the dynamic teaching strategy generated by the teacher intelligent body into the student intelligent body, and the student intelligent body predicts the possible changes in students' learning behavior based on the learned behavior pattern and the strategy-behavior causal model, wherein the learning behavior includes changes in the quality of homework completion, fluctuations in class participation, and adjustments to the progress of knowledge mastery. A generation module is used to link the prediction results with the teaching management mechanism to generate a dynamic learning profile, and the teacher intelligent body matches personalized learning resources through an intelligent recommendation algorithm based on the dynamic learning profile. The student intelligent body receives students' feedback on learning resources and performs targeted intervention actions based on the feedback information.

[0015] The third aspect of the present application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and runnable on the processor. The processor executes the program to implement the student learning behavior prediction method based on artificial intelligence as described in the above embodiment.

[0016] Therefore, the present application includes the following beneficial effects: the embodiment of the present application realizes the multi-dimensional accurate prediction of learning behaviors such as fluctuations in the quality of student homework completion, changes in classroom participation, and speed of knowledge mastery by constructing a collaborative interaction mechanism between student agents and teacher agents based on a strategy-behavior causal model, breaking through the limitation of traditional methods that only rely on data correlation, revealing the internal mechanism of teaching strategy and behavior changes through causal reasoning, and improving the interpretability of prediction results; deeply linking the prediction results with the teaching management mechanism to generate dynamic learning archives, mapping students' knowledge gaps and behavior trends in real time, supporting the teacher agent to dynamically match personalized learning resources based on the intelligent recommendation algorithm, and solving the inefficiency problem of the traditional "one-size-fits-all" resource push model; the student agent captures resource usage feedback to form a closed-loop intervention mechanism, automatically adjusts the teaching strategy parameters, and realizes the upgrade from "static prediction" to "dynamic optimization", effectively responding to individual differences of students and changes in teaching scenarios, improving the timeliness and accuracy of learning behavior prediction, and improving teaching efficiency and student learning outcomes through real-time calibration of teaching strategies and precise delivery of resources. Thus, the problems of insufficient causal reasoning ability, weak calibration ability, and insufficient data fusion ability in the existing technology are solved.

[0017] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which: Figure 1 A flowchart of a method for predicting student learning behavior based on artificial intelligence according to an embodiment of the present application; Figure 2 This is an example diagram of a student agent system for an online education platform provided according to one embodiment of the present application; Figure 3 This is an example diagram of an English vocabulary learning platform provided according to one embodiment of the present application; Figure 4 An example diagram of a physical mechanics unit testing system provided according to one embodiment of the present application; Figure 5 This is an example diagram of a teaching system for a mechanical and electrical professional course in a vocational school provided according to one embodiment of the present application; Figure 6An example diagram of an intelligent education system provided according to one embodiment of the present application; Figure 7 This is an example diagram of a K12 online learning platform provided according to one embodiment of the present application; Figure 8 This is an example diagram of an online programming learning platform provided according to one embodiment of the present application; Figure 9 A flowchart of a method for predicting student learning behavior based on artificial intelligence according to one embodiment of the present application; Figure 10 This is a schematic diagram of the structure of a student learning behavior prediction system based on artificial intelligence provided according to an embodiment of the present application; Figure 11 A schematic diagram of the structure of an electronic device provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0019] The following describes in detail embodiments of the present application. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0020] The following describes an artificial intelligence-based student learning behavior prediction method according to an embodiment of the present application with reference to the accompanying drawings. In response to the problem of weak calibration ability mentioned in the above background technology, the present application provides an artificial intelligence-based student learning behavior prediction method. In this method, by constructing a collaborative interaction mechanism between student intelligent agents and teacher intelligent agents, a strategy-behavior causal model is used to achieve multi-dimensional and accurate prediction of learning behaviors such as fluctuations in the quality of student homework completion, changes in classroom participation, and the speed of knowledge mastery. This method breaks through the limitation of traditional methods that only rely on data correlation, reveals the inherent mechanism of teaching strategy and behavior changes through causal reasoning, and improves the interpretability of prediction results; deeply links the prediction results with the teaching management mechanism to generate dynamic learning archives, maps students' knowledge gaps and behavior trends in real time, and supports teacher intelligent agents to dynamically match personalized learning resources based on intelligent recommendation algorithms to solve the inefficiency of the traditional "one-size-fits-all" resource push model; the student intelligent agent captures resource usage feedback to form a closed-loop intervention mechanism, automatically adjusts teaching strategy parameters, and realizes the upgrade from "static prediction" to "dynamic optimization", effectively responding to individual differences among students and changes in teaching scenarios, improving the timeliness and accuracy of learning behavior prediction, and improving teaching efficiency and student learning outcomes through real-time calibration of teaching strategies and precise delivery of resources. This solves the problems of insufficient causal reasoning ability, weak calibration ability and insufficient data fusion ability in the existing technology.

[0021] Specifically, Figure 1A flowchart of a student learning behavior prediction method based on artificial intelligence provided in an embodiment of the present application.

[0022] like Figure 1 As shown, the student learning behavior prediction method based on artificial intelligence includes the following steps: In step S101, student learning data and teacher teaching task data are obtained.

[0023] It can be understood that the embodiment of the present application integrates the two-way data of students' learning trajectories and teaching plans to construct a data set covering learning behavior characteristics and teaching strategy parameters, providing key data for intelligent agent modeling and behavior prediction, and deeply characterizing the correlation between students' learning behavior and teaching strategies.

[0024] In step S102, a student agent is constructed based on the student learning data.

[0025] Among them, the student agent refers to the artificial intelligence model entity built based on student learning data (such as homework completion quality, class participation, etc.).

[0026] It can be understood that the embodiments of the present application, through deep modeling of student learning data, dynamically simulate the learning behavior patterns of individual students, analyze the correlation mechanism between teaching strategies and learning outcomes, predict learning behaviors such as changes in the quality of homework completion and fluctuations in classroom participation, and transform the student learning process into a computable and simulatable digital entity, providing real-time dynamic quantitative data for the formulation of personalized teaching strategies and optimization of learning intervention measures, thereby improving the intelligence and accuracy of teaching management.

[0027] For example, Figure 2 As shown in the figure, in the online education platform, when using the student agent, we first collect data such as student A's homework accuracy, frequency of classroom questions, and time spent on online tests to build his or her exclusive agent model; when the teacher agent inputs the teaching strategy of "increasing the difficulty gradient of function chapter homework", the student agent predicts that student A's homework completion error rate may increase from 15% to 25% and class participation will decrease based on historical behavior patterns and strategy-behavior causal models; based on this, the system generates a dynamic learning archive containing hints of weak knowledge points, and the teacher agent further pushes special function graph analysis exercises to student A through an intelligent recommendation algorithm; the student agent receives its feedback data on resources in real time (such as a 30% increase in exercise completion efficiency), dynamically adjusts the prediction model and triggers targeted intervention, ultimately helping student A achieve the teaching goal by mastering the knowledge of this chapter 2 weeks earlier.

[0028] In an embodiment of the present application, a student intelligent body is constructed based on student learning data, including: constructing a dynamic cognitive diagnosis algorithm; decoupling the learning data based on the dynamic cognitive diagnosis algorithm, and capturing the correlation between learning behavior and knowledge points through a dynamic knowledge graph; through a forgetting curve model, combined with the correlation relationship, constructing a hierarchical student intelligent body that includes dynamic evolution of learning style, real-time monitoring of fatigue status, and intelligent positioning of knowledge blind spots.

[0029] Among them, the forgetting curve model is a theoretical model that describes the decay pattern of human memory retention over time, which first decays quickly and then decays slowly.

[0030] It can be understood that the embodiments of the present application embed the correlation between learning behavior and knowledge points into the memory decay function, quantitatively simulate the dynamic evolution process of students' learning style, monitor the memory retention rate in real time under fatigue state, and intelligently locate the easily forgotten knowledge points in knowledge blind spots. It integrates the memory dynamics in the time dimension into the student intelligent body modeling, predicts changes in learning behavior, generates personalized review paths that fit the laws of memory, reduces the loss of learning efficiency caused by forgetting, and improves the stability of knowledge mastery and the timeliness of teaching intervention.

[0031] For example, Figure 3 As shown in the figure, in an English word learning platform, when using the forgetting curve model, the system first analyzes student B's word test data through the student agent and constructs the memory decay curve of the student for words of different word frequencies (high frequency / low frequency); when it detects that student B's memory retention of the low-frequency word "ambivalent" drops to 20% on the third day after the first learning (lower than the average retention of high-frequency words of 50%), combined with the association relationship of the "emotional expression" knowledge point to which the word belongs in the dynamic knowledge graph, it intelligently locates the word as a knowledge blind spot that is easy to forget; the system automatically triggers a tiered intervention strategy: on the fourth day (near the forgetting peak), it pushes intensive review resources including word meaning analysis and example application, and adjusts the subsequent learning interval of similar low-frequency words (shortened from 7 days to 5 days); by continuously tracking the changes in memory retention, student B's long-term memory retention rate of low-frequency words increases from 35% to 65%, and the number of knowledge blind spots decreases by 40%.

[0032] In the embodiment of the present application, the dynamic cognitive diagnosis algorithm formula is: ,

[0033] Where P is the probability symbol; is the performance of student i on a certain learning task or test item j; is the attribute parameter of student i; is the comprehensive parameter related to test item j; , is the exponential parameter associated with student i; Indicates k traversal set All elements in is the degree of mastery of knowledge point k by student i or related parameters.

[0034] It can be understood that the embodiments of the present application quantitatively analyze the performance data of students in learning tasks or tests, analyze the interactive relationship between individual student attribute parameters and comprehensive test item parameters, and combine the dynamic mapping of index parameters and the degree of mastery of knowledge points to perform real-time diagnosis of students' knowledge structure, ability level and cognitive weaknesses, convert complex learning behaviors into computable probability distributions, provide fine-grained cognitive state characterization for student intelligent bodies, improve the adaptability of teaching strategies, and intelligently generate personalized learning paths.

[0035] For example, Figure 4 As shown in the figure, in the physical mechanics unit test, when using the dynamic cognitive diagnosis algorithm, the system collected student C's answer data on test items such as buoyancy and pressure, and calculated the interaction probability of its attribute parameters (such as the logical reasoning ability value of 0.72) and the comprehensive parameters of the test items (such as the difficulty coefficient of buoyancy application questions of 0.85) through the algorithm formula; analysis found that student C's performance probability in the "combined calculation of buoyancy and pressure" test item was only 0.41. Combining the index parameter (cognitive load index 1.2) and the degree of mastery of knowledge points (buoyancy formula application proficiency 0.58), it was diagnosed that he had cognitive weaknesses in the comprehensive application of cross-knowledge points; based on this, the system generated targeted training plans, pushed special exercises including step-by-step analysis and mind maps, and adjusted the depth and frequency of cross-knowledge point explanations in subsequent courses; after 2 weeks of intervention, student C's performance probability in similar test items increased to 0.79, and the cognitive load index dropped to 0.95.

[0036] In an embodiment of the present application, a teacher's teaching task data is used to construct a teacher's intelligent agent, including: obtaining teaching data, wherein the teaching data includes explanation methods, homework assignments, and interactive forms; personalizing the teaching data and student characteristic data, and designing an action set centered on improving student learning outcomes; based on the action set, the intelligent agent is trained in combination with historical teaching data, and through a proximal strategy optimization algorithm, the intelligent agent is continuously trialed and errored in a simulated teaching environment, thereby constructing a teacher's intelligent agent that can dynamically adapt to strategies.

[0037] Among them, the teacher agent refers to an artificial intelligence model constructed based on teaching data (such as explanation methods, homework assignments, interaction forms, etc.) and student characteristic data.

[0038] It can be understood that the embodiment of the present application integrates teaching data such as explanation methods and homework assignments with student characteristic data to design an action set centered on learning effects, and uses a proximal strategy optimization algorithm to continuously iterate in a simulation environment to dynamically generate personalized teaching strategies that adapt to students' individual needs. Through a data-driven strategy generation mechanism, it accurately matches students' learning status and cognitive characteristics, thereby improving the real-time and effectiveness of teaching interventions.

[0039] For example, Figure 5 As shown in the figure, in the teaching of mechanical and electrical professional courses in vocational colleges, when using teacher agents, the system first collects the teacher's explanation method (such as case teaching accounts for 60% and theoretical teaching accounts for 40%), homework assignments (including circuit diagram drawing, troubleshooting, etc.) and student characteristic data (such as the class average practical operation ability value of 0.75 and theoretical understanding ability value of 0.62); based on this data, the teacher agent designs an action set, increases the case teaching ratio to 80%, and adds a three-level homework gradient of "basic circuit simulation → fault scenario simulation → group collaborative troubleshooting"; through trial and error iteration in a simulation environment through the proximal strategy optimization algorithm, when it is detected that the student's mastery rate of the knowledge point of "motor control circuit fault analysis" is less than 45%, the agent automatically triggers the reinforcement strategy of "dynamic disassembly of circuit diagram + real-time operation guidance", and at the same time adjusts the classroom interaction form to a "teacher demonstration-student practice-AI real-time error correction" cycle mode; after a four-week teaching cycle, the students' average theoretical assessment score increased from 68 points to 82 points, and the practical operation pass rate increased from 70% to 91%.

[0040] In the embodiment of the present application, the proximal strategy optimization algorithm formula is: in, is the policy-dependent network parameter To find the expectation for time step t; is the importance sampling ratio; is the advantage function; is the clipping function; is a hyperparameter.

[0041] It can be understood that the embodiment of the present application limits the step size of updating the old and new strategies, reduces the drastic fluctuations in the strategy parameters, and enables the intelligent agent to stably update the strategy in the continuous trial and error of the simulated teaching environment. With a smaller sample size and computing cost, it balances strategy exploration and utilization, enables the teacher intelligent agent to quickly learn the optimal strategy to adapt to different student characteristics and teaching scenarios, and improves the efficiency and accuracy of dynamic optimization of teaching strategies.

[0042] For example, Figure 6As shown in Figure 1, a proximal strategy optimization algorithm is used in an intelligent education system to improve teaching effectiveness. This system, centered around a teacher agent, constructs a teaching strategy for a junior high school mathematics course. Initial state information is collected, including student learning data such as past math scores, classroom interaction activity, and homework completion time. In addition, data on the teacher's teaching methods, assignment assignments, and classroom interaction frequency are collected as initial state information. Within a simulated teaching environment, the teacher agent selects actions from a set of actions, such as adjusting the pace of instruction, altering the difficulty distribution of assignments, and changing the interaction format. For example, the teacher agent can shift from a full-class presentation of new knowledge to a first-half lecture with group discussions and a second-half Q&A session for consolidation. The system can also increase the proportion of difficult homework assignments from 20% to 30%, supplemented by detailed problem-solving guidance. After each action is executed, the system assigns rewards or penalties based on the student's subsequent classroom performance, homework completion quality, and test scores. For example, a significant improvement in the student's accuracy rate in assignments is rewarded, while a decrease in class participation is rewarded with negative feedback. The teacher agent uses a proximal policy optimization algorithm to continuously adjust the probability of strategy selection based on these reward feedbacks, limiting the update range between new and old strategies to avoid significant fluctuations in teaching effectiveness due to sudden changes in strategy. After multiple rounds of iteration, the system found the most suitable teaching strategy for the class. On the final exam, the class's average math score increased by 8 points compared to the previous semester, and the excellent rate increased from 25% to 35%.

[0043] In step S103, the dynamic teaching strategy generated by the teacher agent is input into the student agent. Based on the learned behavior patterns and the strategy-behavior causal model, the student agent predicts possible changes in students' learning behaviors, where learning behaviors include changes in the quality of homework completion, fluctuations in class participation, and adjustments to knowledge mastery progress.

[0044] Among them, the strategy-behavior causal model is an analytical framework used to reveal the causal relationship between strategy implementation and behavioral outcomes.

[0045] It can be understood that the embodiments of the present application systematically deconstruct the association mechanism between teaching strategies and learning behaviors, locate the key nodes and impact paths of strategy effects, and provide scientific causal reasoning logic for student intelligent agents to predict changes in learning behaviors. This enables the system to predict potential changes in dimensions such as homework completion quality, classroom participation, and knowledge mastery progress based on the dynamic teaching strategies input by the teacher intelligent agent, combined with historical behavior pattern data, thereby improving the explanatory power of the causal relationship between teaching strategies and student behavioral responses, reducing the one-sidedness of reliance on correlation analysis, and promoting teaching intervention from experience-driven to data-causal dual-wheel drive, thereby guiding and regulating changes in student learning behaviors.

[0046] For example, in a smart classroom experiment, the teacher agent planned to introduce a "real-time error attribution feedback strategy." The strategy-behavior causal model first conducted a causal analysis of historical data. Using propensity score matching, it isolated the differential impacts of "precise error analysis" and "regular homework feedback" on learning behavior. It found that when feedback included "knowledge point tracing + error cause visualization," the key causal path for improving student homework completion quality by 30% was achieved through "reducing ineffective repetitive practice (reducing homework time by 25%) → enhancing the efficiency of identifying weak points (increasing targeted correction by 40%)." Counterfactual reasoning was further used to simulate the effects of the strategy intervention: if a class's current homework error rate was 20%, after implementing this strategy, the model predicted that it would drop below 12% within three weeks, and that the frequency of proactive questioning in class would increase by 15%-20% due to a deeper understanding of the homework. In actual applications, the system adjusted the original daily feedback to "reinforcement of high-frequency wrong questions every other day + weekly summary of low-frequency wrong questions" based on the "feedback frequency-behavioral response" causal curve output by the model, which accelerated students' knowledge acquisition progress by 18% compared with the traditional feedback model, verifying the effectiveness of the model in optimizing strategy details by revealing causal mechanisms.

[0047] In an embodiment of the present application, the student agent predicts possible changes in students' learning behavior based on the learned behavior patterns in combination with the strategy-behavior causal model, including: obtaining historical teaching strategy data and student learning behavior data; based on the historical teaching strategy data and student learning behavior data, using a causal inference algorithm to analyze the causal relationship between teaching strategies and student learning behaviors, and establish a strategy-behavior causal model; based on the strategy-behavior causal model, combined with the counterfactual reasoning mechanism, by simulating the intervention effect of a specific teaching strategy, predicting possible changes in students' learning behavior, and comparing the deviation between the predicted results and the actual observed data.

[0048] Among them, the causal inference algorithm is a statistical analysis method that uses counterfactual analysis, propensity score matching and other techniques to identify causal relationships between variables from data, eliminate interference from confounding factors, reveal causal mechanisms and support decision optimization.

[0049] It can be understood that the embodiments of the present application deeply analyze historical teaching strategies and learning behavior data, disassemble the true causal relationship between the two, eliminate the interference of confounding factors such as individual differences among students and family environment, and construct a clear and explainable causal analysis framework to explore internal causal paths such as "personalized feedback strategies to improve knowledge mastery efficiency" for student intelligent bodies. At the same time, it relies on counterfactual reasoning to simulate the intervention effects of different teaching strategies, predict changes in learning behaviors such as homework completion quality and classroom participation, reduce the pseudo-causal traps that are prone to traditional correlation analysis, and dynamically optimize teaching strategies based on quantitative causal effect evaluation to improve the accuracy of learning behavior prediction.

[0050] For example, in a teaching experiment on an online education platform, the teacher agent planned to implement a "two-stage error reinforcement strategy" (i.e., pre-class error analysis + post-class variation training) for students with weak math foundations. The causal inference algorithm first used propensity score matching to identify a control group with similar learning profiles (receiving only standard error explanations) from historical data, eliminating confounding factors such as students' initial grades and study time. The algorithm then used counterfactual analysis to simulate the effectiveness of the strategy intervention. After controlling for variables such as "home tutoring frequency" and "independent practice time," the algorithm identified a causal path from "pre-class error analysis" to "improving classroom relevance" (preemptively exposing knowledge gaps), reducing homework error rates by 22%. Meanwhile, "post-class variation training" increased unit test accuracy by 18% by "strengthening transferability across similar question types." In practice, the system dynamically adjusted the "variation training frequency" in the strategy from three times per week to two times per week based on the causal effect value output by the algorithm. This reduced the learning burden by 15% while maintaining the same level of knowledge acquisition, validating the core value of the causal inference algorithm by extracting true causal relationships and precisely optimizing teaching strategies.

[0051] In step S104, the prediction results are linked to the teaching management mechanism to generate a dynamic learning profile. The teacher agent matches personalized learning resources based on the dynamic learning profile through an intelligent recommendation algorithm. The student agent receives students' feedback on the learning resources and performs targeted intervention actions based on the feedback information.

[0052] Among them, the intelligent recommendation algorithm is based on user behavior data and item attribute characteristics. With the help of collaborative filtering, deep learning, reinforcement learning and other technologies, it explores the relationship between user preferences and items, and actively pushes content, products or services that meet their needs to users, thereby improving user experience and business efficiency. It is an intelligent data processing method.

[0053] It can be understood that the embodiments of the present application use the behavioral data generated by students during the learning process, such as homework completion status, classroom interaction records, resource browsing time, etc., combined with the type, difficulty, knowledge point coverage and other attribute characteristics of learning resources, and comprehensively use collaborative filtering to discover the common preferences of similar learning groups. With the help of deep learning, the complex learning behavior patterns and resource demand relationships are analyzed. Through reinforcement learning, the recommendation strategy is dynamically optimized based on student feedback, and the student's dynamic learning profile is converted into an accurate personalized learning resource recommendation plan. On the one hand, it improves the efficiency and adaptability of students in obtaining learning resources, stimulates students' interest in learning, and prevents the frustration of learning enthusiasm due to resource mismatch; on the other hand, it provides a more scientific basis for teaching intervention for the teacher intelligent body, timely adjusts teaching strategies, optimizes teaching management processes, and promotes the rational allocation and efficient use of educational resources.

[0054] For example, Figure 7As shown in the figure, on a K12 online learning platform, the system detected that Student B's recent English reading comprehension accuracy rate remained consistently below 70%, and that she spent relatively short time in the vocabulary learning module. An intelligent recommendation algorithm then drew on behavioral data such as Student B's historical reading comprehension records, vocabulary test scores, and resource preferences. This data was then combined with the difficulty level, subject matter, and other attributes of resources available on the platform, including English reading skills courses, high-frequency vocabulary flashcards, and graded reading texts. First, using collaborative filtering technology, the system identified a user group with a similar English learning level and significant improvements in reading comprehension. Based on this group's resource usage habits, the system recommended a series of courses on "Three-Step English Reading Comprehension Problem Solving" and current affairs reading materials of the same difficulty level to Student B. Then, using reinforcement learning, the system dynamically adjusted the recommendation weights based on Student B's post-course test results and his progress in completing the reading materials. The system then recommended "Specialized Training on Parsing Difficult Sentences" and "High-Frequency Reading Vocabulary Development Pack," helping Student B address his shortcomings and significantly improve his English reading scores.

[0055] In an embodiment of the present application, the prediction results are linked to the teaching management mechanism to generate a dynamic learning archive, including: constructing a neural tensor network; reconstructing the knowledge graph of the learning behavior prediction results based on the neural tensor network, and constructing an intelligent archive model that includes ability shortcoming warning indicators, learning path optimization suggestions and cognitive development milestones; based on the intelligent archive model, combined with the microservice architecture, a dynamic learning archive is generated, and at the same time, the learning archive data and the teaching management platform data are collaboratively and hierarchically managed according to the corresponding permissions.

[0056] Among them, the neural tensor network is a model that deeply combines neural networks with tensor operations.

[0057] It can be understood that the embodiments of the present application, by integrating the nonlinear representation ability of neural networks and the multi-dimensional information processing ability of tensor operations, perform structured analysis and deep encoding of the complex relationships in the learning behavior prediction results. In the process of knowledge graph reconstruction, it captures the potential correlation between factors such as students' ability shortcomings and learning path deviations, and constructs a high-dimensional, dynamic and semantically understanding intelligent archive model, so that the ability shortcoming warning indicators are more predictive, the learning path optimization suggestions are more in line with the laws of cognitive development, and the cognitive development milestones are more accurately traceable. The level of structuring and intelligence of learning archive data is improved, and learning behavior data is deeply mined and value extracted. Through the microservice architecture, dynamic archives can be efficiently updated and flexibly called. At the same time, in collaborative hierarchical management, with the help of the multi-dimensional data processing characteristics of tensors, the security and accuracy of data interaction at different authority levels are enhanced.

[0058] For example, in intelligent learning systems at universities, neural tensor networks are being used to analyze the programming ability graph of computer science students. The system inputs multidimensional behavioral data on students' performance in data structures, algorithm design, and program debugging (e.g., number of code submissions, debugging time, and error type distribution). Through multi-layer nonlinear transformations applied by the tensor network, it captures the high-order tensor correlation between "recursive algorithm comprehension bias" and "dynamic programming application barriers," discovering a third-order interaction between the two in the dimension of abstract thinking. Based on this analysis, the system reconstructs a programming ability knowledge graph, not only accurately identifying students' core weakness of "weak state transition equation modeling" but also automatically generating personalized training paths that include visual tutorials on recursive algorithms and a library of classic dynamic programming examples. This has increased students' efficiency in solving similar programming problems by 40%.

[0059] In an embodiment of the present application, the student intelligent agent receives students' feedback on learning resources and performs targeted intervention actions based on the feedback information, including: constructing an emotional computing model; parsing the implicit needs in the feedback data through the emotional computing model, and dynamically adjusting the intervention strategy using the Pareto algorithm; based on the adjusted intervention strategy, generating a set of intervention actions for different students and learning scenarios, including intelligent reminder triggering rules, intelligent reconstruction of learning paths, and dynamic balance of cognitive load.

[0060] Among them, the emotional computing model is an intelligent model that analyzes multimodal data such as text, voice, and images, and uses machine learning, deep learning and other algorithms to achieve emotion recognition, emotional tendency analysis, emotional state prediction or emotion generation.

[0061] It can be understood that the embodiment of the present application analyzes the implicit emotional states (such as frustration, confusion, and interest fluctuations) and deep needs (such as cognitive load overload warning and personalized guidance demands) in student feedback by integrating data such as text semantic analysis, voice emotion recognition, and facial expression micro-movement capture, providing decision-making data of the emotional dimension for intelligent intervention strategies, and mining implicit associations such as "abnormal fluctuations in answering time → germination of anxiety" and "jumping of video playback progress → uncomfortable content difficulty" based on machine learning algorithms, so that the student intelligent body can perceive the emotional turning points in the learning process. On the one hand, through quantitative evaluation of emotional states (such as generating a real-time emotional stress index), the triggering timing and intensity of the intervention strategy are dynamically adjusted. On the other hand, based on emotional tendency analysis, the learning path reconstruction logic is optimized. For students with low interest, gamified learning resources are recommended to enhance participation motivation.

[0062] For example, Figure 8As shown in the example, on an online programming learning platform, an affective computing model analyzes students' multimodal feedback data in real time. It uses camera feeds to detect micro-expressions like frowning and frequent eye rubbing to gauge learning stress. Combined with the urgency of voice communication and text feedback such as "completely lost" after repeated code debugging failures, the model determines that the student is experiencing intense frustration. Based on this information, the model immediately triggers an intervention strategy, prioritizing basic grammar consolidation lessons. It also arranges for virtual teaching assistants to provide gentle, step-by-step guidance and reduce the difficulty of tasks in subsequent learning paths. Ultimately, this helps students regain confidence and complete programming projects, validating the significant value of affective computing models in accurately identifying emotions and optimizing the learning experience.

[0063] According to the student learning behavior prediction method based on artificial intelligence proposed in the embodiment of the present application, by constructing a collaborative interaction mechanism between student intelligent agents and teacher intelligent agents, a multi-dimensional accurate prediction of learning behaviors such as fluctuations in the quality of student homework completion, changes in classroom participation, and speed of knowledge mastery is achieved based on a strategy-behavior causal model, breaking through the limitation of traditional methods that only rely on data correlation, revealing the inherent mechanism of teaching strategy and behavior changes through causal reasoning, and improving the interpretability of prediction results; deeply linking the prediction results with the teaching management mechanism to generate dynamic learning archives, mapping student knowledge gaps and behavior trends in real time, supporting teacher intelligent agents to dynamically match personalized learning resources based on intelligent recommendation algorithms, and solving the inefficiency problem of the traditional "one-size-fits-all" resource push model; student intelligent agents capture resource usage feedback, form a closed-loop intervention mechanism, automatically adjust teaching strategy parameters, and achieve an upgrade from "static prediction" to "dynamic optimization", effectively responding to individual differences among students and changes in teaching scenarios, improving the timeliness and accuracy of learning behavior prediction, and improving teaching efficiency and student learning outcomes through real-time calibration of teaching strategies and precise delivery of resources. Thus, the problems of insufficient causal reasoning ability, weak calibration ability, and insufficient data fusion ability in the existing technology are solved.

[0064] The following will describe the student learning behavior prediction method based on artificial intelligence through a specific embodiment. Figure 9 As shown, including: On an online education platform with over 100,000 registered students, the system collects learning data from multiple channels: over 200,000 pieces of structured data are captured daily from the backend, covering details of course viewing, assignments, and tests; approximately 500,000 text feedback messages are collected monthly from chat windows and discussion forums; and 10,000 hours of multimodal data are collected weekly using cameras and microphones. After collection, a cleaning algorithm removes 15% of duplicate data and over 30% of missing data. The text is processed to reduce the average word length from 18.5 to 12.3. 130-dimensional facial features and 80-dimensional acoustic features are extracted from images and speech, respectively, resulting in an analytical dataset containing 20 million records and over 300 dimensions.

[0065] Three models were constructed based on preprocessed data. The causal inference model used a propensity score matching algorithm to reduce the standardized difference between different teaching strategy groups from 0.25 to 0.08. Combined with counterfactual analysis, it found causal relationships, such as a 22% improvement in homework quality (a 15-point increase in scores) with tiered assignments. The sentiment computing model, using CNN and RNN architectures and trained on datasets such as FER2013, achieved accuracy rates of 88%, 85%, and 92% for image, speech, and text sentiment analysis, respectively, with an overall accuracy of 89% after feature fusion. The neural tensor network model mined complex relationships, such as the third-order tensor correlation (coefficient 0.65) between mathematics and physics, reconstructed a knowledge graph, and achieved an accuracy rate of 82% for predicting skill gaps. After 100 training cycles and parameter adjustments using the Adam optimizer, the causal inference model achieved a mean squared error of 5.2 for homework quality predictions, an F1 score of 0.87 for the sentiment computing model, and an 85% accuracy rate for learning path recommendations for the neural tensor network model.

[0066] While students are learning, the system collects behavioral, feedback, and multimodal data 10 times per second, feeding it into the model via a Kafka queue. A policy-behavior causal model predicts that high-intensity homework will cause a class's homework quality to decline by 18% next week (mean squared error 3.8). The sentiment computing model, with a confidence level of 0.85, determined that 23% of students in a live class were anxious about the pace of the class. Based on sentiment analysis, the Pareto algorithm adjusts strategies, such as reducing task difficulty to alleviate anxiety by 65% and recommending gamified resources to increase click-through rates by 40%. This generates intervention actions, including intelligent reminders (78% break execution rate) and path reconstruction (25% improvement in knowledge acquisition efficiency). The neural tensor network updates 200 pieces of knowledge graph data weekly. Under the microservices architecture, the teacher receives 5,000 daily accesses and the student receives 8,000. Tiered management ensures zero permission violations for six months.

[0067] Monthly comparisons of predictions with actual data revealed that the causal inference model achieved 88% accuracy, 85% recall, an F1-value of 0.86, and a mean squared error of 4.5 in predicting operational quality; the affective computing model achieved 90% accuracy, 87% recall, and an F1-value of 0.88 in sentiment recognition; and the neural tensor network model achieved 83% accuracy, 80% recall, and an F1-value of 0.81 in capacity warning. To address discrepancies, the causal model was supplemented with training on similar strategy data, the affective model incorporated Transformer technology to optimize long text analysis, and the neural tensor network employed distributed computing to improve efficiency. After a year of optimization, the overall performance of each model improved by an average of 12%.

[0068] In summary, the present invention collects massive learning data through multiple channels, acquires more than 200,000 pieces of structured data daily, collects about 500,000 text feedbacks monthly, and collects 10,000 hours of multimodal data weekly. After deep processing, it forms a data set of 20 million records and 300+ dimensions. The three types of models work together. The causal inference model mines the causal relationship between teaching strategies and learning behaviors, which improves the quality of homework by 22% under the tiered homework strategy; the emotional computing model integrates multimodal data, with an overall accuracy rate of 89%, accurately identifying students' emotions; the neural tensor network analyzes complex relationships, and the accuracy rate of ability shortcoming warning reaches 82%. During the learning process, the system collects data in real time 10 times per second to drive dynamic prediction and intelligent intervention. For example, reducing the difficulty of tasks can reduce students' anxiety by 65%, recommending personalized resources can increase click-through rate by 40%, and improve knowledge mastery efficiency by 25%. At the same time, the neural tensor network updates 200 pieces of knowledge graph data every week to realize dynamic management of learning archives and ensure that there are no data permission violations for half a year. Through monthly evaluation and optimization, the comprehensive performance of each model has increased by an average of 12% per year, ultimately achieving an accuracy rate of over 88% in learning behavior prediction, optimizing teaching strategies, upgrading the learning experience, and pushing online education towards new heights of intelligence and precision, effectively improving the quality of education and teaching and student learning outcomes.

[0069] Next, a device for constructing a turbidity testing system based on a photoelectric detector according to an embodiment of the present application will be described with reference to the accompanying drawings.

[0070] Figure 10 It is a block diagram of the student learning behavior prediction system based on artificial intelligence in an embodiment of the present application.

[0071] like Figure 10 As shown, the student learning behavior prediction system 10 based on artificial intelligence includes: an acquisition module 100, a construction module 200, a prediction module 300, and a generation module 400.

[0072] Among them, the acquisition module 100 is used to obtain student learning data and teacher teaching task data; the construction module 200 is used to construct a student agent based on student learning data; the prediction module 300 is used to input the dynamic teaching strategy generated by the teacher agent into the student agent. The student agent, based on the learned behavior patterns and combined with the strategy-behavior causal model, predicts possible changes in students' learning behavior. Learning behavior includes changes in the quality of homework completion, fluctuations in classroom participation, and adjustments to knowledge mastery progress. The generation module 400 is used to link the prediction results with the teaching management mechanism to generate a dynamic learning profile. The teacher agent matches personalized learning resources based on the dynamic learning profile through an intelligent recommendation algorithm. The student agent receives students' feedback on the learning resources and performs targeted intervention actions based on the feedback information.

[0073] It should be noted that the above explanation of the embodiment of the student learning behavior prediction method based on artificial intelligence is also applicable to the student learning behavior prediction system based on artificial intelligence in this embodiment, and will not be repeated here.

[0074] According to the student learning behavior prediction system based on artificial intelligence proposed in the embodiment of the present application, by constructing a collaborative interaction mechanism between student intelligent agents and teacher intelligent agents, a multi-dimensional accurate prediction of learning behaviors such as fluctuations in the quality of student homework completion, changes in classroom participation, and speed of knowledge mastery is achieved based on a strategy-behavior causal model, breaking through the limitation of traditional methods that only rely on data correlation, revealing the inherent mechanism of teaching strategy and behavior changes through causal reasoning, and improving the interpretability of prediction results; deeply linking the prediction results with the teaching management mechanism to generate dynamic learning archives, mapping student knowledge gaps and behavior trends in real time, supporting teacher intelligent agents to dynamically match personalized learning resources based on intelligent recommendation algorithms, and solving the inefficiency problem of the traditional "one-size-fits-all" resource push model; student intelligent agents capture resource usage feedback, form a closed-loop intervention mechanism, automatically adjust teaching strategy parameters, and achieve an upgrade from "static prediction" to "dynamic optimization", effectively responding to individual differences among students and changes in teaching scenarios, improving the timeliness and accuracy of learning behavior prediction, and improving teaching efficiency and student learning outcomes through real-time calibration of teaching strategies and precise delivery of resources. Thus, the problems of insufficient causal reasoning ability, weak calibration ability, and insufficient data fusion ability in the existing technology are solved.

[0075] Figure 11 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device may include: A memory 1101 , a processor 1102 , and a computer program stored in the memory 1101 and executable on the processor 1102 .

[0076] When the processor 1102 executes the program, the student learning behavior prediction method based on artificial intelligence provided in the above embodiment is implemented.

[0077] Furthermore, the electronic device further includes: The communication interface 1103 is used for communication between the memory 1101 and the processor 1102 .

[0078] The memory 1101 is used to store computer programs that can be run on the processor 1102 .

[0079] The memory 1101 may include a high-speed RAM (Random Access Memory) memory, and may also include a non-volatile memory, such as at least one disk memory.

[0080] If the memory 1101, processor 1102, and communication interface 1103 are implemented independently, the communication interface 1103, memory 1101, and processor 1102 can be connected to each other via a bus and communicate with each other. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 11 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0081] Optionally, in a specific implementation, if the memory 1101, the processor 1102 and the communication interface 1103 are integrated on a chip, the memory 1101, the processor 1102 and the communication interface 1103 can communicate with each other through an internal interface.

[0082] The processor 1102 may be a CPU (Central Processing Unit), or an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present application.

[0083] In the description of this specification, reference to the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification, as well as features of different embodiments or examples, unless they are mutually inconsistent.

[0084] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features being referred to. Thus, a feature specified as "first" or "second" may explicitly or implicitly include at least one such feature. Throughout the description of this application, "plurality" means at least two, for example, two, three, etc., unless otherwise specifically defined.

[0085] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present application belong.

[0086] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having logic gate circuits for implementing logical functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0087] Those skilled in the art will appreciate that all or part of the steps in the method for implementing the above-mentioned embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0088] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A student learning behavior prediction method based on artificial intelligence, characterized in that: include: Obtain student learning data and teacher teaching task data; Constructing a student agent based on the student learning data and a teacher agent based on the teacher teaching task data; The dynamic teaching strategy generated by the teacher agent is input into the student agent. The student agent, based on the learned behavior patterns and in combination with the strategy-behavior causal model, predicts possible changes in the student's learning behavior, where the learning behavior includes changes in the quality of homework completion, fluctuations in class participation, and adjustments in knowledge mastery progress. The prediction results are linked to the teaching management mechanism to generate a dynamic learning profile. The teacher agent matches personalized learning resources based on the dynamic learning profile through an intelligent recommendation algorithm. The student agent receives students' feedback on the learning resources and performs targeted intervention actions based on the feedback information.

2. The method for predicting student learning behavior based on artificial intelligence according to claim 1, characterized in that: The step of constructing a student agent based on the student learning data includes: Construct dynamic cognitive diagnostic algorithms; Decoupling learning data based on the dynamic cognitive diagnosis algorithm, and capturing the relationship between learning behavior and knowledge points through a dynamic knowledge graph; By combining the forgetting curve model with the above-mentioned association relationship, a hierarchical student intelligent agent is constructed, which includes dynamic evolution of learning style, real-time monitoring of fatigue status and intelligent positioning of knowledge blind spots.

3. The method for predicting student learning behavior based on artificial intelligence according to claim 2, characterized in that: The dynamic cognitive diagnosis algorithm formula: ， ； Among them, P is the probability symbol; is the performance of student i on a certain learning task or test item j; is the attribute parameter of student i; is the comprehensive parameter related to test item j; , is the exponential parameter associated with student i; Indicates k traversal set All elements in is the degree of mastery of knowledge point k by student i or related parameters.

4. The method for predicting student learning behavior based on artificial intelligence according to claim 1, characterized in that: The teacher teaching task data is used to construct a teacher agent, including: Acquiring teaching data, wherein the teaching data includes teaching methods, homework assignments, and interactive forms; Performing personalized adjustments to the teaching data and student characteristic data to design an action set centered on improving student learning outcomes; Based on the action set, the intelligent agent is trained in combination with historical teaching data. Through the proximal strategy optimization algorithm, the intelligent agent is enabled to continuously trial and error in a simulated teaching environment, and a teacher intelligent agent that can dynamically adapt to strategies is constructed.

5. The method for predicting student learning behavior based on artificial intelligence according to claim 4 is characterized in that: The proximal strategy optimization algorithm formula: in, is the policy-dependent network parameter CLIP loss function; To find the expectation for time step t; is the importance sampling ratio; is the advantage function; is the clipping function; is a hyperparameter.

6. The method for predicting student learning behavior based on artificial intelligence according to claim 1, characterized in that: The prediction results are linked with the teaching management mechanism to generate a dynamic learning archive, including: Construct neural tensor networks; Reconstructing a knowledge graph of learning behavior prediction results based on the neural tensor network to build an intelligent profile model that includes ability shortcoming warning indicators, learning path optimization suggestions, and cognitive development milestones; Based on the intelligent archive model and combined with the microservice architecture, dynamic learning archives are generated. At the same time, the learning archive data and the teaching management platform data are collaboratively managed in a hierarchical manner according to the corresponding permissions.

7. The method for predicting student learning behavior based on artificial intelligence according to claim 1, characterized in that: The student agent receives student feedback on learning resources and performs targeted intervention actions based on the feedback information, including: Build an affective computing model; The emotional computing model is used to analyze the implicit needs in the feedback data, and the Pareto algorithm is used to dynamically adjust the intervention strategy; Based on the adjusted intervention strategy, a set of intervention actions including intelligent reminder triggering rules, intelligent reconstruction of learning paths, and dynamic balance of cognitive load is generated for different students and learning scenarios.

8. The method for predicting student learning behavior based on artificial intelligence according to claim 1, characterized in that: The student agent predicts possible changes in students' learning behavior based on the learned behavior patterns and the strategy-behavior causal model, including: Obtain data on history teaching strategies and student learning behavior; Based on the historical teaching strategy data and student learning behavior data, using a causal inference algorithm to analyze the causal relationship between teaching strategies and student learning behaviors, and establish a strategy-behavior causal model; Based on the strategy-behavior causal model, combined with the counterfactual reasoning mechanism, by simulating the intervention effect of a specific teaching strategy, the possible changes in students' learning behavior are predicted, and the deviations between the predicted results and the actual observed data are compared.

9. A student learning behavior prediction system based on artificial intelligence, characterized by: include: The acquisition module is used to obtain student learning data and teacher teaching task data; a construction module, configured to construct a student agent based on the student learning data; A prediction module is configured to input the dynamic teaching strategy generated by the teacher agent into the student agent. The student agent, based on the learned behavior patterns and in combination with the strategy-behavior causal model, predicts possible changes in the student's learning behavior, where the learning behavior includes changes in the quality of homework completion, fluctuations in class participation, and adjustments in knowledge mastery progress. A generation module is used to link the prediction results with the teaching management mechanism to generate a dynamic learning profile. The teacher intelligent agent matches personalized learning resources based on the dynamic learning profile through an intelligent recommendation algorithm. The student intelligent agent receives students' feedback on the learning resources and performs targeted intervention actions based on the feedback information.

10. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method for predicting student learning behavior based on artificial intelligence as described in claims 1-8.

Images