Optimization method and system based on intelligent learning condition analysis and interactive teaching and medium thereof

Abstract

The invention discloses an optimization method and system based on intelligent learning condition analysis and interactive teaching and a medium thereof, and relates to the technical field of educational informationization, and the method comprises the steps: carrying out the recognition and analysis of a teaching aid data image, and generating structured homework data containing knowledge point labels; collecting classroom behavior data of students, calculating a concentration degree deviation value based on knowledge point labels, and triggering teaching strategy adjustment; constructing a student ability evaluation matrix, establishing a dynamic task grouping model based on the student ability evaluation matrix and the task attribute characteristics, and generating a task-learning progress report; collecting teacher physiological data, and combining course information and complexity indexes of the dynamic task grouping model to construct a teacher fatigue evaluation model; calling a course arrangement optimization algorithm to optimize a course arrangement scheme based on a fatigue evaluation result; acquiring student learning track data, generating a personalized learning plan by adopting machine learning, and dynamically adjusting learning parameter weights through a feedback mechanism; the interactive teaching quality and the learning efficiency of students are improved.

Description

Technical Field

[0001] The present application relates to the field of educational information technology, and in particular to a method, system and medium thereof based on intelligent learning situation analysis and interactive teaching optimization. Background Art

[0002] With the in-depth development of educational informatization, intelligent teaching platforms have shown great potential in improving teaching efficiency and personalizing teaching. However, existing technologies still have shortcomings in accurately generating student ability profiles and providing scientific feedback mechanisms.

[0003] Traditional teaching platforms rely on manual grading of homework and manual analysis of student learning data, which is not only time-consuming and labor-intensive, but also difficult to achieve real-time tracking and detailed analysis of students' learning status. As a result, there are problems with poor interactive teaching quality and it is not conducive to improving students' learning efficiency. Summary of the Invention

[0004] In order to improve the quality of interactive teaching and enhance students' learning efficiency, this application provides a method, system and medium based on intelligent learning situation analysis and interactive teaching optimization.

[0005] In the first aspect, the invention objectives of this application are achieved by adopting the following technical solutions: Based on intelligent learning situation analysis and interactive teaching optimization methods, including: Perform text recognition and question analysis on supplementary teaching material images to generate structured homework data containing question types and knowledge point labels; Collect students' classroom behavior data, calculate the concentration deviation value based on the knowledge point labels and the attention assessment model, and trigger the adjustment of teaching strategies; Based on the error rate and knowledge point mastery in the student's historical homework data, combined with the concentration deviation value, a student ability evaluation matrix is ​​constructed, and a dynamic task grouping model is established based on the student ability evaluation matrix and task attribute characteristics to generate a task-learning progress report; The teacher's physiological data is collected through biosensors, and the teacher fatigue assessment model is constructed by combining the course information with the complexity index of the dynamic task grouping model; the class scheduling optimization algorithm is called based on the fatigue assessment results to optimize the class scheduling plan; The student's learning trajectory data is obtained, a personalized learning plan is generated using machine learning, and the learning parameter weights are dynamically adjusted based on the task-learning progress report through a feedback mechanism.

[0006] By adopting the above technical solutions, this application can be applied to industry-education integration scenarios such as vocational education and training centers and higher education institutions; automatically parse teaching materials to improve the efficiency of homework grading, and at the same time, structured homework data provides standardized input for subsequent learning situation analysis; in order to optimize the quality of classroom interaction in real time, the teaching strategy is dynamically adjusted through the attention assessment model, which is conducive to improving students' concentration on the classroom; the dynamic task grouping model is conducive to improving task matching to refine the allocation of learning resources; through the teacher's physiological data and fatigue assessment model, the teacher's physical condition is effectively monitored to prevent the decline in teaching quality due to fatigue, and the class scheduling plan is dynamically optimized to improve the stability of the overall teaching system; through personalized learning plans, personalized learning plans are promoted, and students' learning patterns are mined by learning to develop learning paths that suit their individual characteristics. The personalized learning plan of this application supports dynamic adjustment of parameters such as learning content and training intensity to adapt to the learning needs of different stages and improve students' learning efficiency. Therefore, this application provides a teaching optimization solution with a higher degree of automation, more timely teaching response, stronger personalization, and more reasonable allocation of educational resources.

[0007] In a preferred embodiment of the present application, the structured homework data includes first homework data and second homework data; after performing text recognition and question analysis on the image of the supplementary teaching material, the method further includes: Obtaining learning situation analysis factors related to knowledge point mastery based on the first homework data and the second homework data, wherein the learning situation analysis factors include a first learning situation analysis factor and a second learning situation analysis factor; Obtaining a learning status influencing factor representing the student's learning status based on the learning situation analysis factor, wherein the learning status influencing factor includes a first learning status influencing factor and a second learning status influencing factor; Dividing the student groups according to the learning status influencing factors to obtain a first student group and a second student group; Obtain personalized teaching parameters based on the first student group and the second student group.

[0008] By adopting the above technical solution, the first homework data is daily practice, and the second homework data is test detection, distinguishing the two types of homework data to avoid confusing data interference in different scenarios; by extracting learning situation analysis factors, capturing hard indicators such as knowledge point mastery and error rate and soft indicators such as answering speed, comprehensively assessing students' true level; through learning status influencing factors, abstract learning behaviors are converted into quantifiable indicators (such as concentration deviation values) to facilitate tracking students' classroom learning status, and automatically divide classes and groups according to differences in learning status, avoiding one-size-fits-all teaching, allowing students of similar levels to promote each other, and generating exclusive learning rhythm suggestions (such as the frequency of redoing wrong questions), resource recommendation weights and other personalized teaching parameters for each student.

[0009] In a preferred example of the present application, the process of obtaining the first operation data and the second operation data includes: Acquiring multiple-choice question homework data and subjective question homework data according to the teaching supplementary material image; Obtaining standard answer matching results based on the multiple-choice question homework data and the subjective question homework data; The first operation data and the second operation data are obtained according to the standard answer matching result and the question type.

[0010] By adopting the above technical solutions, different scoring standards are used for objective questions (multiple-choice questions) and subjective questions (fill-in-the-blank / essays) to improve the accuracy of grading; by matching standard answers and analyzing question types, the accuracy of answer comparison is ensured, such as math questions are accurate to four decimal places.

[0011] In a preferred example of the present application, obtaining the learning status influencing factor based on the learning situation analysis factors specifically includes: Obtaining preset knowledge association factors, and filtering the first learning situation analysis factors based on the preset knowledge association factors to obtain first learning status influencing factors, wherein the first learning status influencing factors include first knowledge point mastery, a first error rate, and a first answering speed; Obtaining a first learning state influencing factor according to the first learning state influencing factor; According to the preset knowledge association factors, the second learning situation analysis factors are screened to obtain second learning status influencing factors, wherein the second learning status influencing factors include the mastery of the second knowledge point, the second error rate, the second answering speed, and the repetition rate of wrong questions in history; A second learning state influencing factor is obtained according to the second learning state influencing factor.

[0012] By adopting the above technical solutions, key learning factors are screened and irrelevant data, such as the general font size, is filtered out, while the core reflection indicators of learning ability are retained. By distinguishing two types of learning status factors, such as the first type (basic version) focuses on real-time performance (error rate), and the second type (advanced version) adds historical data (repetition rate of wrong questions), it can adapt to the needs of different teaching stages.

[0013] In a preferred embodiment of the present application, the attention assessment model is a multimodal attention assessment model; and the method further comprises: Based on the body movement characteristics and gaze focus trajectory in the classroom behavior data, a multimodal attention assessment model is constructed; continuous classroom segments are segmented and annotated based on a sliding window algorithm, and the student attention fluctuation index in each time period is calculated; Based on the covariance analysis of the student attention fluctuation index and the knowledge point difficulty coefficient matrix of the corresponding teaching materials, a teaching strategy adjustment instruction set is dynamically generated.

[0014] By employing these technical solutions, a multimodal attention assessment model monitors students' body movements (such as nodding / turning their heads) and gaze focus, more accurately detecting distraction than a single camera. Using a sliding window, segmented annotation—for example, breaking a 45-minute class into three-minute segments—can identify fluctuations in attention in real time (e.g., the class is distracted when explaining a difficult question). The teacher can then make targeted adjustments, such as slowing down the pace of challenging material or inserting interactive elements. Covariance analysis generates strategies that analyze the difficulty of knowledge points and student attention fluctuation indexes to dynamically adjust teaching strategies, such as using 3D models to aid comprehension of challenging math problems.

[0015] In a preferred example of this application, the process of constructing the student competency assessment matrix includes: Using a time series analysis model to calculate the attenuation coefficients of the first error rate and the second error rate, and combining the forgetting curve theory to correct the knowledge point mastery weights; The first level of focus deviation is graded and quantified using a fuzzy logic algorithm, establishing a three-dimensional evaluation matrix encompassing cognitive engagement, knowledge absorption rate, and error correction efficiency. Dynamically grouping the three-dimensional evaluation matrix based on the k-means++ clustering algorithm to generate student clusters with similar learning characteristics; The first learning status influencing factor and the second learning status influencing factor are associated according to the similarity of the learning trajectories of the student clusters.

[0016] By adopting the above technical solutions, students' general error rates are analyzed based on time series to accurately determine their knowledge gaps. Fuzzy logic is used to quantify concentration to avoid subjective judgment, so teachers don't have to worry about whether "a student's 10-second vacant moment counts as distraction." A three-dimensional evaluation matrix + K-means++ clustering is used to generate student clusters: accurately dividing student groups with different learning characteristics, laying the foundation for personalized teaching.

[0017] In a preferred embodiment of the present application, the method for establishing the dynamic task grouping model includes: Based on the cluster center coordinates of the student clusters, the cosine similarity matrix of the task attribute features and the student feature vectors is calculated; the improved k-medoids algorithm is used to cluster and optimize the cosine similarity matrix to generate a task grouping scheme containing multiple grouping schemes; A grouping stability prediction model is established through a reinforcement learning algorithm to output the confidence probability distribution of each grouping scheme; According to the confidence probability distribution and the class time allocation constraints in the course information, an adaptive teaching resource push strategy is generated through a multi-objective decision tree.

[0018] By adopting the above technical solutions, the cosine similarity matrix is ​​used to calculate the matching degree between tasks and students, measuring the degree of adaptation between each student and the task; the improved k-medoids algorithm is used to optimize grouping, making the grouping results more practical for teaching, and the feasibility of different grouping schemes is evaluated in advance through the grouping stability prediction model; the multi-objective decision tree generates a resource push strategy to intelligently recommend the optimal teaching resource configuration method under constraints such as time, number of people, and task complexity.

[0019] In the second aspect, the invention objective of this application is achieved by adopting the following technical solutions: The system based on intelligent learning situation analysis and interactive teaching optimization is applied to the above-mentioned method based on intelligent learning situation analysis and interactive teaching optimization, and the system includes: The teaching material digital processing module is used to perform text recognition and question analysis on teaching material images, and generate structured homework data containing question types and knowledge point labels; A classroom behavior collection and analysis module is used to collect students' classroom behavior data, calculate the concentration deviation value based on the knowledge point labels and the attention assessment model, and trigger the adjustment of teaching strategies; A student ability assessment and task grouping module is used to construct a student ability assessment matrix based on the error rate and knowledge point mastery in the student's historical homework data, combined with the concentration deviation value, establish a dynamic task grouping model based on the student ability assessment matrix and task attribute characteristics, and generate a task-learning progress report; The course scheduling optimization module is used to collect teacher physiological data through biosensors, combine course information with the complexity index of the dynamic task grouping model, and build a teacher fatigue assessment model; based on the fatigue assessment results, the course scheduling optimization algorithm is called to optimize the course scheduling plan; The personalized learning plan generation module is used to obtain student learning trajectory data, use machine learning to generate a personalized learning plan, and dynamically adjust the learning parameter weights based on the task-learning progress report through a feedback mechanism.

[0020] In a third aspect, the invention objective of this application is achieved by adopting the following technical solutions: A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned method based on intelligent learning situation analysis and interactive teaching optimization.

[0021] Fourthly, the invention objectives of this application are achieved by adopting the following technical solutions: A computer program product includes a computer program / instruction, which, when executed by a processor, implements the steps of the above-mentioned intelligent learning situation analysis and interactive teaching optimization method.

[0022] In summary, this application includes at least one of the following beneficial technical effects: 1. Dynamically optimize the course schedule to improve the stability of the overall teaching system; promote personalized learning plans through personalized learning plans, use learning to explore students' learning patterns, and develop learning paths that suit their individual characteristics. The personalized learning plan of this application supports dynamic adjustment of parameters such as learning content and training intensity to adapt to the learning needs of different stages and improve students' learning efficiency; 2. By extracting learning situation analysis factors, capturing hard indicators such as knowledge point mastery and error rate, and soft indicators such as answering speed, the true level of students can be comprehensively assessed; through learning status influencing factors, abstract learning behaviors are converted into quantifiable indicators (such as concentration deviation values) to facilitate tracking students' classroom learning status, and automatically divide classes and groups according to differences in learning status to avoid one-size-fits-all teaching. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a flow chart of an intelligent learning situation analysis and interactive teaching optimization method in one embodiment of the present application; Figure 2 This is a flowchart after step S1 in the intelligent learning situation analysis and interactive teaching optimization method in one embodiment of the present application. DETAILED DESCRIPTION

[0024] The present application is further described in detail below with reference to the accompanying drawings.

[0025] In one embodiment, if Figure 1 As shown, this application discloses a method for optimizing learning situation analysis and interactive teaching, which specifically includes the following steps: S1: Perform text recognition and question analysis on images of teaching materials to generate structured homework data containing question types and knowledge point labels.

[0026] In this embodiment, images of supplementary teaching materials are collected to obtain supplementary teaching material images; text recognition uses OCR technology to identify text, formulas, charts and other data in the supplementary teaching material images, and converts them into structured homework data in JSON format; question parsing refers to using knowledge graphs (such as GeoGebra mathematics knowledge base) to parse the knowledge points tested in the questions; structured homework data includes field question ID, question type (select / fill in the blank / calculation / subjective / objective), knowledge point label (People's Education Edition Mathematics-Function), and difficulty coefficient (level 1-5).

[0027] S2: Collect students' classroom behavior data, calculate the concentration deviation value based on knowledge point labels and attention assessment model, and trigger teaching strategy adjustments.

[0028] In this embodiment, the attention assessment model is an LSTM neural network based on multimodal data (eye movement + posture); the concentration deviation value refers to the difference between the actual concentration time and the theoretical concentration time of the course (formula: deviation value = theoretical time × 0.7 - actual concentration time).

[0029] Specifically, the eye movement heat map (such as the proportion of the blackboard area being gazed at) and posture angles in each 15-second window are recorded using TobiiPro Glasses3 (sampling rate 120Hz); the student's sitting posture is captured through OpenPose (if the head pitch angle is greater than 15° and the duration of gaze at the blackboard area within 15 seconds does not exceed 5 seconds, it is judged as distraction); when the deviation value is greater than 12 minutes, the system automatically pushes a micro-class video (such as a function graph transformation animation) and sends a reminder to the teacher's tablet (or other teaching terminals used by teachers): for example, "The average concentration of students in Class 3 has decreased, and it is recommended to increase interactive Q&A."

[0030] Furthermore, the process of constructing the student competency assessment matrix includes: S100: Use the time series analysis model to calculate the attenuation coefficients of the first error rate and the second error rate, and use the forgetting curve theory to correct the knowledge point mastery weights.

[0031] In this embodiment, the forgetting curve parameters are: based on the Ebbinghaus forgetting curve model, the knowledge retention rate attenuation coefficient λ is set to e^-kt (k is the memory decay constant). Time series analysis uses the ARIMA model to fit the error rate trend, calculate the short-term fluctuation coefficient α and the long-term trend coefficient β, and the weight correction formula is: Where E1 is the first error rate; E2 is the second error rate.

[0032] S200: The first concentration deviation value is graded and quantified using a fuzzy logic algorithm to establish a three-dimensional evaluation matrix including cognitive engagement, knowledge absorption rate, and error correction efficiency.

[0033] In this embodiment, the first concentration deviation value (A1) refers to the proportion of distracted time in classroom behavior data; the error correction efficiency (R) is the ratio of the correct rate of redoing wrong questions to the time consumed (e.g., correcting 80% of errors within 10 minutes). Concentration is graded as follows: low (A1 < 0.3), medium (0.3 ≤ A1 < 0.7), and high (A1 ≥ 0.7); correction efficiency is graded as follows: low efficiency (R < 0.5), medium (0.5 ≤ R < 0.8), and high efficiency (R ≥ 0.8).

[0034] Specifically, cognitive engagement is normalized to a value range of 0-1, knowledge absorption rate is expressed in the form of a percentage, and error correction efficiency refers to a comprehensive indicator of the speed of correcting wrong questions and the accuracy rate, expressed as a score of 0-100.

[0035] S300: Dynamically group the three-dimensional evaluation matrix based on the k-means++ clustering algorithm to generate student clusters with similar learning characteristics.

[0036] In this embodiment, the initialization optimization of the k-means++ clustering algorithm refers to selecting the point farthest from the existing cluster center as the new center. The iteration termination condition is that the cluster center movement distance is less than ε (e.g., 0.001) or the number of iterations is ≥100. For example, the k-means++ clustering algorithm generates three clusters, including cluster 1 (high-efficiency learners): high cognitive engagement (>0.8) and high absorption rate (>0.7); Cluster 2 (medium learners): medium cognitive engagement (0.4-0.6), medium absorption rate (0.4-0.6); Cluster 3 (group to be improved): low correction efficiency (<0.4).

[0037] S400: Associating the first learning state influencing factor with the second learning state influencing factor according to the similarity of the learning trajectories of the student clusters.

[0038] In this embodiment, the DTW algorithm (Dynamic Time Warping) is used to match the temporal similarity of students' historical learning paths, where the similarity calculation formula is: Among them, T1 and T2 are time series learning path data based on time series, which can represent the learning path data of students A and B respectively, and analyze the differences between students in the same task group. It is expressed as T1 = [t 11 , t 12 ,…,t 1n ], T2=[t 21 , t 22 ,…,t 2n ]. It is the sum of squares of the Euclidean distance, which calculates the sum of squares of the numerical differences of two time series at the same time point, reflecting the overall degree of difference.

[0039] S3: Based on the error rate and knowledge point mastery in students' historical homework data, combined with the concentration deviation value, a student ability assessment matrix is ​​constructed. Based on the student ability assessment matrix and task attribute characteristics, a dynamic task grouping model is established to generate a task-learning progress report.

[0040] In this embodiment, the ability assessment matrix is ​​a three-dimensional matrix (cognitive engagement × knowledge absorption rate × error correction efficiency), where cognitive engagement = concentration deviation value × number of classroom interactions; knowledge absorption rate = (current accuracy rate - accuracy rate one week ago) / knowledge point complexity coefficient; error correction efficiency = accuracy rate of redoing wrong questions / time spent on redoing; the dynamic task grouping model is a clustering model based on the improved k-medoids algorithm (with course complexity constraints added); the task-learning progress report includes course information, task completion status, recommended learning resource links, next assessment time, learning progress and other information.

[0041] In vocational education scenarios, the dynamic task grouping model prioritizes matching skill-based practical tasks (such as machine tool operation simulation). Task attribute characteristics include equipment type code (CNC-01) and matching degree of safety operation specifications.

[0042] Specifically, the mastery of knowledge points is based on the value corrected by the forgetting curve), and the k-medoids algorithm is used (the initial center selects the sample with the largest silhouette coefficient) to group tasks. The grouping examples are: Group A (characteristics: error rate <20% and high involvement), the teaching strategy is challenging exploration tasks; Group B (characteristics: error rate 20%-50%), the teaching strategy is layered consolidation exercises; Group C (characteristics: error rate >50%), the teaching strategy is filling in the gaps in basic concepts.

[0043] S4: Collect teachers' physiological data through biosensors, combine course information with the complexity index of the dynamic task grouping model, and build a teacher fatigue assessment model; based on the fatigue assessment results, call the scheduling optimization algorithm to optimize the scheduling plan.

[0044] In this embodiment, the biosensor data refers to the heart rate variability (HRV) and skin conductance (EDA) collected by the smart bracelet; the complexity index refers to the gradient value of the Pareto front in the task grouping model (reflecting the teaching load).

[0045] Specifically, the teacher fatigue assessment model is an LSTM model, the detection data is [HRV mean, EDA change rate, continuous teaching time], and the fatigue level (0-3 levels) is output. The class scheduling optimization adopts a genetic algorithm, that is, chromosome encoding is based on the course time period (08:00-12:00), teacher allocation, and classroom resources. The corresponding fitness function is 1 / (fatigue level × number of classroom conflicts), and the fatigue assessment result including the teacher fatigue level is output. An example of the optimization result is: the original three consecutive math classes on Wednesday afternoon are adjusted to math → physical education → math.

[0046] S5: Obtain student learning trajectory data, use machine learning to generate personalized learning plans, and dynamically adjust learning parameter weights based on task-learning progress reports through a feedback mechanism.

[0047] In this embodiment, learning trajectory data includes clickstream data (such as the number of times a micro-course video is paused), resource retention time, homework data, learning habits, and knowledge mastery change rate; personalized learning plans include daily tasks, course resource recommendations, time arrangements, and learning progress feedback; parameter weight adjustment refers to the Q-learning reward function in reinforcement learning (such as +10 points for completing a high-difficulty task).

[0048] Specifically, abnormal clickstream data is cleaned (for example, five consecutive clicks within five seconds are considered abnormal); a knowledge mastery time series diagram is constructed with nodes as knowledge points and edge weights as the rate of change of knowledge mastery. Transformer-XL is used to process long-term learning behaviors (for example, window length = 30 days) to track students' learning tasks (for example, if students have mastered "linear functions", the next step is to learn "quadratic functions") and match the difficulty (if students always miscalculate coefficients, the system will recommend more basic exercises, such as drawing a parabola graph first and then learning the vertex formula); and then, based on the previous analysis, a personalized learning plan is generated.

[0049] The daily tasks of the personalized learning plan include the course schedule for the day, and the dynamic adjustment of the learning parameter weights includes: error warning for answering questions, status monitoring and progress feedback; the error warning means that if a student gets three consecutive similar questions wrong, the system will automatically insert a "knowledge point make-up class"; status monitoring means that if the student is found to be always distracted during class (for example, the camera captures him turning his head to look out the window), the system will recommend highly interactive tasks (such as quick-answer games); the progress feedback part means that if the student completes all tasks ahead of time, the system will "step up" to challenge more difficult questions; if the progress falls behind, the content will be simplified or the review time will be increased.

[0050] Furthermore, when the classroom camera is blocked by more than 30%, it will automatically switch to the desktop camera; when the biosensor is disconnected, historical fatigue data interpolation and completion will be enabled.

[0051] In one embodiment, if Figure 2 As shown, the structured homework data includes first homework data and second homework data; after step S1, the method based on intelligent learning situation analysis and interactive teaching optimization further includes: S11: Obtain learning situation analysis factors related to knowledge point mastery based on the first homework data and the second homework data, wherein the learning situation analysis factors include a first learning situation analysis factor and a second learning situation analysis factor.

[0052] In this embodiment, the first homework data refers to daily classroom exercise data (such as in-class quizzes and homework), focusing on process-based learning performance; the second homework data refers to periodic examination / test data, focusing on result-based knowledge mastery assessment; the learning situation analysis factor is a set of quantitative indicators reflecting the student's learning status; the first learning situation analysis factor is a short-term learning status indicator generated based on daily homework data (such as multiple-choice questions, classroom quizzes, etc.), reflecting the student's real-time performance and potential problems in recent learning; the second learning situation analysis factor is a long-term ability assessment indicator generated based on periodic examination data (such as monthly examinations, midterm examinations, etc.), reflecting the student's knowledge system integrity and deep thinking ability.

[0053] Specifically, the knowledge point association analysis includes: for multiple-choice questions: after matching the standard answers, the knowledge point score rate of each question is calculated (such as the correct rate of set questions is 82%); for subjective questions: through OCR to recognize the handwritten steps, use the BERT model to score (such as geometry proof questions get 7 / 10 points).

[0054] The process of obtaining the first operation data and the second operation data includes: S101: Acquire multiple-choice question homework data and subjective question homework data based on the teaching supplementary material image.

[0055] In this embodiment, the multiple-choice homework data includes objective test data, including the question stem, options, and correct answer identifiers. The subjective test data includes the question stem, student handwritten answers, and non-standardized test data with grading criteria. The homework data is obtained using an image segmentation algorithm, where the image segmentation algorithm refers to an image region segmentation model based on deep learning (e.g., Mask R-CNN).

[0056] Specifically, the image preprocessing steps include: using OpenCV to grayscale and binarize the teaching aid images, eliminating interference from paper stains, and locating the question borders through the edge detection algorithm (Canny operator).

[0057] When classifying question types, a CNN classification model is used (input: question area image, output: question type label), and then feature extraction is performed. For example, the characteristics of multiple-choice questions include: rectangular option boxes, "ABCD" option labels, and the "()" symbol at the end of the question stem; the characteristics of subjective questions include: paragraph-style text, no option structure, and the presence of prompt words such as "Answer:".

[0058] S102: Obtain standard answer matching results based on the multiple-choice question homework data and the subjective question homework data.

[0059] In this embodiment, the standard answer library contains a database of correct answers and scoring rules for each question; an answer matching algorithm is used to obtain standard answer matching results, where the answer matching algorithm includes string similarity calculation (such as Levenshtein distance) and formula structure matching (such as MathML comparison).

[0060] Specifically, when extracting answers, for multiple-choice questions, the system directly reads the checkmarks in the option area (e.g., "●A" indicates option A is selected); for subjective questions, the system uses handwritten OCR to identify student answers (e.g., "y=2x+1"). The system then normalizes the mathematical formulas and text answers, and performs answer matching calculations.

[0061] S103: Obtain first homework data and second homework data according to the standard answer matching result and the question type.

[0062] In this embodiment, both the first and second assignment data are associated with corresponding data tags, which include knowledge points, difficulty levels, and question type weights. Multiple-choice question characteristics include accuracy rate, answer time, and option noise; subjective question characteristics include step completeness score, key formula coverage, and logical coherence score.

[0063] S12: Obtain a learning status influencing factor representing the student's learning status based on the learning situation analysis factors, wherein the learning status influencing factor includes a first learning status influencing factor and a second learning status influencing factor.

[0064] In this embodiment, the learning status influencing factor represents a dynamic indicator system of the student's real-time learning status; the first learning status factor is a short-term status monitoring indicator based on daily practice; and the second learning status factor is a long-term ability assessment indicator based on test data.

[0065] Specifically, based on the learning situation analysis factors, the learning status influencing factors are obtained, including: S121: Obtain preset knowledge association factors, and screen the first learning situation analysis factors based on the preset knowledge association factors to obtain the first learning status influencing factors, wherein the first learning status influencing factors include the mastery of the first knowledge point, the first error rate, and the first answering speed.

[0066] In this embodiment, the preset knowledge association factor refers to the pre-defined knowledge point association rules (such as "to master trigonometric functions, you must first understand trigonometric ratios"); the first knowledge point mastery refers to the student's real-time mastery of basic knowledge (such as the accuracy of drawing quadratic function graphs); the first error rate refers to the number of consecutive errors in similar question types in recent homework / exams; the first answering speed refers to the average time spent per unit question (such as ≤30 seconds per multiple-choice question is efficient).

[0067] Specifically, a knowledge graph based on teaching materials is established. If the knowledge graph shows that "to master knowledge point A, you need to master knowledge point B first", the learning situation data related to B will be given priority. For example, if a student has not mastered "simplifying fractions" (B), the answer data of his fractional equation questions (A) will be filtered.

[0068] S122: Obtain a first learning state influencing factor according to the first learning state influencing factor.

[0069] In this embodiment, the learning status influencing factor converts the original learning data into an algorithm output of a quantifiable evaluation indicator; the first learning status influencing factor is a set of core indicators used to quantify students' short-term learning performance and real-time learning status, which is mainly generated based on daily homework and classroom behavior data.

[0070] Specifically, the first learning status influencing factors include the first knowledge point mastery, the first error rate, the first question-answering speed, and the attention concentration. S123: Filter the second learning situation analysis factors according to the preset knowledge association factors to obtain the second learning status influencing factors, wherein the second learning status influencing factors include the second knowledge point mastery, the second error rate, the second question answering speed and the repetition rate of historical wrong questions.

[0071] In this embodiment, the second knowledge point mastery refers to the student's long-term accumulated knowledge point mastery level (such as the average score of the function chapter from the beginning of the semester to the present); the second error rate refers to the student's tendency to make repeated mistakes in high-level questions (such as consecutively losing points on the final question); the second answering speed refers to the time management ability to complete complex problems (such as the ratio of the average time spent on answering questions to the standard time); the repetition rate of historical wrong questions refers to the frequency of recurrence of wrong questions on the same knowledge point in multiple exams (such as the number of times the wrong question "vector perpendicular condition" appeared in the past three exams).

[0072] S124: Obtain a second learning state influencing factor according to the second learning state influencing factor.

[0073] In this embodiment, the second learning status influencing factor is a set of core indicators used to evaluate students' long-term learning ability and knowledge structure stability, and is mainly generated based on periodic examinations, comprehensive test questions or long-term learning trajectory data.

[0074] Specifically, the first learning status influencing factor focuses on the mastery of basic knowledge and real-time learning status; the second learning status influencing factor includes knowledge integration ability and depth of thinking.

[0075] S13: Divide the student groups according to the learning status influencing factors to obtain a first student group and a second student group.

[0076] In this embodiment, student group division is based on a machine learning algorithm to divide students into groups with different learning characteristics; the first student group refers to students with similar short-term learning status (such as a group with a recent continuous decline in concentration); the second student group refers to students with similar long-term ability characteristics (such as a group with weak function knowledge).

[0077] Specifically, the annotation features of the student group also include: 0: 'High-risk group for concentration', 1: 'Weak function knowledge group', 2: 'Stable group of computing power', 3: 'Potential Enhancement Group', 4: 'Group with excellent comprehensive abilities'.

[0078] Based on the above characteristics, you can define the annotation characteristics of the first student group and the second student group when analyzing the actual academic situation, and analyze the academic qualifications of groups with different characteristics.

[0079] S14: Obtain personalized teaching parameters according to the first student group and the second student group.

[0080] In this embodiment, the personalized teaching parameters include configuration parameters of teaching resource recommendation weights and intervention strategy priorities; the dynamic adaptation mechanism refers to real-time adjustment of parameter thresholds based on group characteristics.

[0081] Specifically, the parameter baseline settings are adjusted as follows: recommended resource weight: {video: 0.3, question bank: 0.5, gamification: 0.2}; intervention trigger threshold: {concentration <60: high priority, error rate >30%: medium priority}.

[0082] For example, we can optimize the weight of recommended learning resources based on the different group labels to which students belong, and set corresponding teaching intervention reminder rules: If it is Group 0 (students with low concentration and easy distraction): It focuses more on gamified learning content (50%), has more question bank exercises (40%), and fewer video explanations (10%); when a student's concentration is lower than 70%, the system will issue a medium-priority reminder, prompting the teacher or the system to intervene appropriately.

[0083] If it is Group 1 (students who have a poor grasp of function knowledge): The main recommendations are video explanations (60%), combined with a certain amount of question bank exercises (30%), and very little gamification content (10%); when the error rate of such students exceeds 25%, the system will issue a high-priority reminder and provide timely and focused tutoring.

[0084] In this embodiment, the attention assessment model is a multimodal attention assessment model; the method further includes: S10: Construct a multimodal attention assessment model based on the body movement characteristics and gaze focus trajectory in classroom behavior data.

[0085] In this embodiment, the multimodal attention model is a deep learning model that integrates multi-dimensional data such as vision (body movement) and vision (eye focus); body movement features refer to biological features such as head pitch angle, gesture frequency, and torso inclination; and eye focus trajectory refers to the movement path of students' eyes between teaching materials, blackboard, and teacher.

[0086] Specifically, an infrared camera was used to collect student body movement sequences (sampling rate ≥ 60 fps), including head pitch angle, gesture frequency, and torso tilt. Body movement data was de-noised using a Kalman filter, and gaze trajectory data was normalized using a Z-score (μ = 0, σ = 1). Gaze focus trajectory data was collected using an eye tracker (sampling rate ≥ 120 Hz), recording pupil position and gaze area (blackboard / teaching materials / teacher). When eye tracker data was lost for more than 10 seconds, historical attention mean interpolation was used. A Transformer multimodal fusion network was constructed, with the input layer consisting of a visual encoder (ViT) and a pose encoder (LSTM). The loss function used was a cross-entropy loss, and the optimizer was AdamW. The infrared camera used was a FLIRAX8 (resolution 640×480), the eye tracker was a Tobii Pro Glasses 3 (sampling rate 120 Hz), and the data was stored in a local MySQL database (version 8.0). "The visual encoder uses the ViT-B / 16 model (ImageNet pre-trained weights), the pose encoder LSTM hidden layer dimension is 256, and the number of Transformer multi-head attention heads is 8."

[0087] S20: Based on the sliding window algorithm, continuous classroom segments are segmented and labeled, and the student attention fluctuation index in each time period is calculated.

[0088] In this embodiment, the sliding window algorithm divides continuous class time into sub-segments of fixed length (such as a segment every 5 minutes); the attention fluctuation index is a statistical indicator used to quantify the changes in students' concentration in different time periods.

[0089] Specifically, a sliding window algorithm (window length T_window = 300s, step length T_step = 30s) is used to frame the continuous classroom behavior data; each frame of data is annotated with an attention label: 1 (Concentration): Continuous gaze at the teaching materials for ≥180s and gesture frequency ≤2 times / min; 0 (distracted): The eyes are away from the teaching materials for ≥60 seconds or the head pitch angle change rate is ≥0.5rad / s 2 .

[0090] Calculate the rate of change of the concentration labels of adjacent windows and normalize the fluctuation index.

[0091] S30: Based on the covariance analysis of the student attention fluctuation index and the knowledge point difficulty coefficient matrix of the corresponding teaching materials, a teaching strategy adjustment instruction set is dynamically generated.

[0092] Specifically, the knowledge point difficulty coefficient matrix is ​​provided with preset difficulty levels for each knowledge point (such as easy: 1, difficult: 5); covariance analysis is used to quantify the correlation between attention fluctuations and knowledge point difficulty.

[0093] Specifically, a knowledge difficulty matrix (KDM) is established, where each element Dij represents the difficulty level (1-5) of the i-th knowledge point in the j-th teaching stage; Calculate the covariance between the attention fluctuation index and the difficulty of the knowledge point: Among them, V iT is the attention fluctuation index of the i-th knowledge point in time period T, D i is the difficulty coefficient of the knowledge point in the i-th knowledge point time period T; and is the average value; n is the total number of segments into which the class time is divided. Cov>0 indicates that attention fluctuation is positively correlated with the difficulty of the knowledge point.

[0094] The rule base for teaching strategy generation includes: Covariance sign Correlation strength Types of teaching strategies >0.7 Strong positive correlation Reduce the difficulty of knowledge points and insert examples <-0.5 Strong negative correlation Add interactive Q&A to improve participation [-0.1，0.1] No significant correlation Maintain the current teaching pace It should be understood that the serial numbers of the steps in the above embodiments do not imply the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0095] In one embodiment, a system based on intelligent learning situation analysis and interactive teaching optimization is provided. The system based on intelligent learning situation analysis and interactive teaching optimization corresponds to the method based on intelligent learning situation analysis and interactive teaching optimization in the above embodiment.

[0096] The system, based on intelligent learning situation analysis and interactive teaching optimization, includes a teaching material digital processing module, a classroom behavior collection and analysis module, a student ability assessment and task grouping module, a class scheduling optimization module, and a personalized learning plan generation module. Detailed descriptions of each functional module are as follows: The teaching material digital processing module is used to perform text recognition and question analysis on teaching material images, and generate structured homework data containing question types and knowledge point labels; The classroom behavior collection and analysis module is used to collect students' classroom behavior data, calculate the concentration deviation value based on knowledge point tags and attention assessment models, and trigger teaching strategy adjustments; The student ability assessment and task grouping module is used to construct a student ability assessment matrix based on the error rate and knowledge point mastery in students' historical homework data, combined with the concentration deviation value. Based on the student ability assessment matrix and task attribute characteristics, a dynamic task grouping model is established to generate task-learning progress reports; The course scheduling optimization module is used to collect teachers' physiological data through biosensors, combine course information with the complexity index of the dynamic task grouping model, and build a teacher fatigue assessment model; based on the fatigue assessment results, the course scheduling optimization algorithm is called to optimize the course scheduling plan; the personalized learning plan generation module is used to obtain students' learning trajectory data, use machine learning to generate personalized learning plans, and dynamically adjust the learning parameter weights based on the task-learning progress report through a feedback mechanism.

[0097] For the specific limitations of the system based on intelligent learning situation analysis and interactive teaching optimization, please refer to the limitations of the method based on intelligent learning situation analysis and interactive teaching optimization in the above text, which will not be repeated here; the various modules in the above-mentioned system based on intelligent learning situation analysis and interactive teaching optimization can be implemented in whole or in part through software, hardware and their combination; the above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the corresponding operations of the above modules.

[0098] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented: S1: Perform text recognition and question analysis on images of supplementary teaching materials to generate structured homework data containing question types and knowledge point labels; S2: Collect students' classroom behavior data, calculate the concentration deviation value based on knowledge point labels and attention assessment model, and trigger teaching strategy adjustments; S3: Based on the error rate and knowledge mastery of students' historical homework data, combined with the concentration deviation value, a student ability evaluation matrix is ​​constructed. Based on the student ability evaluation matrix and task attribute characteristics, a dynamic task grouping model is established to generate a task-learning progress report. S4: Collect teacher physiological data through biosensors, combine course information with the complexity index of the dynamic task grouping model, and build a teacher fatigue assessment model; based on the fatigue assessment results, call the class scheduling optimization algorithm to optimize the class scheduling plan; S5: Obtain student learning trajectory data, use machine learning to generate personalized learning plans, and dynamically adjust learning parameter weights based on task-learning progress reports through a feedback mechanism.

[0099] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0100] In one embodiment, in particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the method for optimizing intelligent learning situation analysis and interactive teaching. In such an embodiment, the computer program can be downloaded and installed from the network through a communication module, and / or installed from a removable medium. When the computer program is executed by a central processing unit (CPU), the various functions defined in the present invention are performed.

[0101] Those skilled in the art will clearly understand that for the sake of convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0102] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, it should be understood by those skilled in the art that the technical solutions described in the aforementioned embodiments may still be modified, or some of the features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. Based on intelligent learning situation analysis and interactive teaching optimization method, it is characterized by: include: Perform text recognition and question analysis on supplementary teaching material images to generate structured homework data containing question types and knowledge point labels; Collect students' classroom behavior data, calculate the concentration deviation value based on the knowledge point labels and the attention assessment model, and trigger the adjustment of teaching strategies; Based on the error rate and knowledge point mastery in the student's historical homework data, combined with the concentration deviation value, a student ability evaluation matrix is ​​constructed, and a dynamic task grouping model is established based on the student ability evaluation matrix and task attribute characteristics to generate a task-learning progress report; The teacher's physiological data is collected through biosensors, and the teacher fatigue assessment model is constructed by combining the course information with the complexity index of the dynamic task grouping model; the class scheduling optimization algorithm is called based on the fatigue assessment results to optimize the class scheduling plan; The student's learning trajectory data is obtained, a personalized learning plan is generated using machine learning, and the learning parameter weights are dynamically adjusted based on the task-learning progress report through a feedback mechanism.

2. The method for optimizing intelligent learning situation analysis and interactive teaching according to claim 1 is characterized in that: The structured operation data includes first operation data and second operation data; After performing text recognition and title analysis on the teaching material image, the method further includes: Obtaining learning situation analysis factors related to knowledge point mastery based on the first homework data and the second homework data, wherein the learning situation analysis factors include a first learning situation analysis factor and a second learning situation analysis factor; Obtaining a learning status influencing factor representing the student's learning status based on the learning situation analysis factor, wherein the learning status influencing factor includes a first learning status influencing factor and a second learning status influencing factor; Dividing the student groups according to the learning status influencing factors to obtain a first student group and a second student group; Obtain personalized teaching parameters based on the first student group and the second student group.

3. The method for optimizing intelligent learning situation analysis and interactive teaching according to claim 2 is characterized in that: The process of acquiring the first operation data and the second operation data includes: Acquiring multiple-choice question homework data and subjective question homework data according to the teaching supplementary material image; Obtaining standard answer matching results based on the multiple-choice question homework data and the subjective question homework data; The first operation data and the second operation data are obtained according to the standard answer matching result and the question type.

4. The method for optimizing intelligent learning situation analysis and interactive teaching according to claim 2 is characterized in that: The obtaining of learning status influencing factors according to the learning situation analysis factors specifically includes: Obtaining preset knowledge association factors, and filtering the first learning situation analysis factors based on the preset knowledge association factors to obtain first learning status influencing factors, wherein the first learning status influencing factors include first knowledge point mastery, a first error rate, and a first answering speed; Obtaining a first learning state influencing factor according to the first learning state influencing factor; According to the preset knowledge association factors, the second learning situation analysis factors are screened to obtain second learning status influencing factors, wherein the second learning status influencing factors include the mastery of the second knowledge point, the second error rate, the second answering speed, and the repetition rate of wrong questions in history; A second learning state influencing factor is obtained according to the second learning state influencing factor.

5. The method for optimizing intelligent learning situation analysis and interactive teaching according to claim 4 is characterized in that: The attention assessment model is a multimodal attention assessment model; the method further includes: Constructing a multimodal attention assessment model based on the body movement characteristics and gaze focus trajectory in the classroom behavior data; Based on the sliding window algorithm, continuous classroom segments are segmented and annotated, and the student attention fluctuation index in each time period is calculated; Based on the covariance analysis of the student attention fluctuation index and the knowledge point difficulty coefficient matrix of the corresponding teaching materials, a teaching strategy adjustment instruction set is dynamically generated.

6. The method for optimizing intelligent learning situation analysis and interactive teaching according to claim 5 is characterized in that: The process of constructing the student competency assessment matrix includes: Using a time series analysis model to calculate the attenuation coefficients of the first error rate and the second error rate, and combining the forgetting curve theory to correct the knowledge point mastery weights; The first level of focus deviation is graded and quantified using a fuzzy logic algorithm, establishing a three-dimensional evaluation matrix encompassing cognitive engagement, knowledge absorption rate, and error correction efficiency. Dynamically grouping the three-dimensional evaluation matrix based on the k-means++ clustering algorithm to generate student clusters with similar learning characteristics; The first learning status influencing factor and the second learning status influencing factor are associated according to the similarity of the learning trajectories of the student clusters.

7. The method for optimizing intelligent learning situation analysis and interactive teaching according to claim 6 is characterized in that: The method for establishing the dynamic task grouping model includes: Calculating the cosine similarity matrix between the task attribute features and the student feature vectors based on the cluster center coordinates of the student clusters; The improved k-medoids algorithm is used to cluster and optimize the cosine similarity matrix to generate a task grouping scheme containing multiple grouping schemes; A grouping stability prediction model is established through a reinforcement learning algorithm to output the confidence probability distribution of each grouping scheme; According to the confidence probability distribution and the class time allocation constraints in the course information, an adaptive teaching resource push strategy is generated through a multi-objective decision tree.

8. A system based on intelligent learning situation analysis and interactive teaching optimization, characterized by: Applied to the method for optimizing intelligent learning situation analysis and interactive teaching according to any one of claims 1 to 7, the system comprises: The teaching material digital processing module is used to perform text recognition and question analysis on teaching material images, and generate structured homework data containing question types and knowledge point labels; A classroom behavior collection and analysis module is used to collect students' classroom behavior data, calculate the concentration deviation value based on the knowledge point labels and the attention assessment model, and trigger the adjustment of teaching strategies; A student ability assessment and task grouping module is used to construct a student ability assessment matrix based on the error rate and knowledge point mastery in the student's historical homework data, combined with the concentration deviation value, establish a dynamic task grouping model based on the student ability assessment matrix and task attribute characteristics, and generate a task-learning progress report; The course scheduling optimization module is used to collect teacher physiological data through biosensors, combine course information with the complexity index of the dynamic task grouping model, and build a teacher fatigue assessment model; based on the fatigue assessment results, the course scheduling optimization algorithm is called to optimize the course scheduling plan; The personalized learning plan generation module is used to obtain student learning trajectory data, use machine learning to generate a personalized learning plan, and dynamically adjust the learning parameter weights based on the task-learning progress report through a feedback mechanism.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the intelligent learning situation analysis and interactive teaching optimization method as described in any one of claims 1 to 7 are implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by the processor, the steps of the intelligent learning situation analysis and interactive teaching optimization method as described in any one of claims 1 to 7 are implemented.

Images