AI-driven learning intelligent evaluation and diagnosis method

Abstract

The invention relates to the technical field of artificial intelligence, in particular to an AI-driven learning intelligent evaluation and diagnosis method, which comprises the following steps: acquiring learning duration, answer accuracy and repeated learning times in a learning record database, extracting learning behavior parameters, calculating knowledge point mastering state indexes, and generating the knowledge point mastering state indexes. According to the method, learning record data is deeply analyzed, cognitive load and error rate are comprehensively considered, knowledge point mastering state is accurately evaluated, interdisciplinary knowledge mapping is enhanced, structure matching and migration efficiency are enhanced, understanding of path continuity is enhanced through dynamic analysis of learning behaviors, learning rhythm is optimized by monitoring learning track change, and learning efficiency is improved. And by utilizing a forgetting curve and an error rate trend, the prediction precision of knowledge mastering and keeping is improved, so that personalized learning recommendation is more accurate, and the application effect of an education technology is effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an AI-driven learning intelligent assessment and diagnosis method. Background Art

[0002] The field of artificial intelligence technology includes core technologies such as machine learning, natural language processing, computer vision, and intelligent decision-making. Its main goal is to process complex data and perform intelligent tasks by simulating human intelligence. The development of artificial intelligence technology covers methods such as deep learning, neural networks, and reinforcement learning. These methods are widely used in many fields such as image recognition, speech recognition, autonomous driving, and intelligent recommendations. In the field of education, artificial intelligence is used in intelligent assessment, personalized learning recommendations, and learning behavior analysis to improve teaching efficiency and learning experience.

[0003] Among them, the AI-driven learning intelligent assessment and diagnosis method refers to the use of artificial intelligence technology to assess and diagnose the learner's knowledge mastery, and provide targeted learning suggestions based on the analysis results. This method covers knowledge point extraction, learning behavior analysis, answer pattern recognition, etc. Specifically, it realizes the modeling of the learner's knowledge structure by constructing a knowledge graph, and combines deep learning technology for intelligent evaluation. The sequence modeling method is used to analyze the learner's answer data, the pattern recognition technology is used to determine the learner's weak links, and the decision tree method is used to match the appropriate learning content, thereby completing the intelligent assessment and diagnosis of learning progress and knowledge mastery.

[0004] In the process of learning assessment, the existing technology relies on knowledge structure modeling based on knowledge graphs and combines deep learning technology for intelligent evaluation. However, the method of extracting learning features of knowledge points is relatively simple, mainly relying on answering patterns and learning records, which makes it difficult to fully reflect the behavioral differences of learners on different knowledge points. Knowledge transfer analysis mainly focuses on the static association of knowledge points, lacks in-depth analysis of interdisciplinary knowledge structures, resulting in the lack of logical rationality of the recommended learning paths. At the level of learning behavior analysis, the use of interactive behavior data is relatively limited, and the dynamic changes in learning methods cannot be accurately captured, making it difficult to effectively evaluate the coherence of learning paths. Learning progress monitoring relies on sequence modeling, fails to combine the learning trajectory change rate with the knowledge point switching attenuation rate, and is difficult to accurately identify learning blockages, resulting in limited timeliness of personalized learning recommendations. Mastery trend prediction mainly relies on static matching of forgetting curves, fails to combine individualized knowledge retention characteristics, and results in insufficient accuracy in predicting long-term mastery status. These deficiencies affect the accuracy of personalized learning recommendations, the effectiveness of learning progress monitoring, and the rationality of knowledge transfer adaptation. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and propose an AI-driven learning intelligent evaluation and diagnosis method.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: an AI-driven learning intelligent evaluation and diagnosis method, comprising the following steps: S1: Obtain the learning time, answer accuracy, and number of repeated learning in the learning record database, extract learning behavior parameters, calculate the knowledge point mastery status index, and generate the knowledge point mastery status index; S2: Based on the knowledge point mastering status index, call the interdisciplinary knowledge mapping model to calculate the knowledge structure fitting degree, screen the interdisciplinary knowledge points, and generate the interdisciplinary knowledge fitness value; S3: Based on the knowledge point mastering state index and the interdisciplinary knowledge adaptation value, obtain learning interaction data, calculate the learning mode characteristic value, call the learning behavior pattern library to compare the operation behavior, analyze the learning mode category, judge the behavior pattern coherence according to the learning path feature model, and generate the learning behavior dynamic pattern parameters; S4: calling the learning behavior dynamic mode parameters, calculating the learning time gradient, trajectory change rate, switching attenuation rate, analyzing the knowledge point learning trajectory, identifying the learning stagnation point, and generating the learning path blocking factor; S5: Based on the learning path retardation factor, calculate the learning progress change trend, determine the long-term mastery status of the knowledge points, and generate a knowledge mastery trend attenuation index.

[0007] As a further solution of the present invention, the knowledge point mastery status index includes the proportion of learning time, mastery fluctuation value, cognitive difficulty index, and mastery stability; the interdisciplinary knowledge adaptation value includes knowledge structure fitting, concept vector projection relationship, migration knowledge point set, dependency weight, and matching standard; the learning behavior dynamic pattern parameters include learning method characteristic values, learning method categories, and behavior consistency parameters; the learning path blocking factors include learning time gradient, learning trajectory change rate, switching attenuation rate, knowledge absorption rate, and blocking knowledge points; the knowledge mastery trend attenuation index includes knowledge retention rate, forgetting rate, error rate change rate, and long-term mastery status.

[0008] As a further solution of the present invention, the specific steps of obtaining the learning time, answering accuracy, and number of repeated learning in the learning record database, extracting the learning behavior parameters, and calculating the knowledge point mastering state index are as follows: S101: Obtaining the learning time, answer accuracy, and repeated learning times of knowledge points in the learning record database, calculating the learning time proportion of knowledge points, extracting knowledge points with large differences in proportion, and obtaining differentiated knowledge point time distribution values; S102: calling the time distribution value of the differentiated knowledge points and the answer accuracy change trend data, calculating the fluctuation range index of the answer accuracy, calculating the mastery fluctuation characteristic value for the knowledge points with a large fluctuation range, matching it with the learning burden parameter library, calculating the cognitive burden change index, screening the knowledge points with significant mastery fluctuation, analyzing the mastery stability in combination with their change trend, and obtaining the knowledge point mastery fluctuation value; S103: calling the knowledge point mastery fluctuation value, calculating the knowledge point answer error rate distribution index, combining the knowledge point cognitive difficulty index and the mastery fluctuation value, and generating the knowledge point mastery status index.

[0009] As a further solution of the present invention, the calculation formula of the knowledge point answer error rate distribution index is specifically: ; in, Represents the distribution index of the error rate of answering questions at knowledge points, Representative The error rate of this knowledge point in the test, Represents the average error rate of this knowledge point in all tests. Represents the number of tests for this knowledge point. Representative The learners’ error deviation in answering questions on this knowledge point is represents the total number of all learners, Represents the cognitive difficulty index of this knowledge point, Represents the weight coefficient of this knowledge point in the question, Represents the mastery fluctuation value of this knowledge point, Represents the baseline value of the mastery stability of this knowledge point in the learning group.

[0010] As a further solution of the present invention, based on the knowledge point mastering status index, the specific steps of calling the interdisciplinary knowledge mapping model to calculate the knowledge structure fitting degree, screening the interdisciplinary knowledge points, and generating the interdisciplinary knowledge fitness value are as follows: S201: Based on the knowledge point mastering status index, obtain interdisciplinary knowledge mapping data, calculate the structural matching degree between the knowledge point and the cross-domain knowledge point, select the knowledge points whose matching degree meets the standard, and obtain the interdisciplinary knowledge matching degree value; S202: calling the interdisciplinary knowledge matching value, analyzing the spatial projection relationship of the knowledge point concept vectors, calculating the distance matrix between the concept vectors, extracting the knowledge point set with close spatial projection relationship, screening the interdisciplinary knowledge point set that meets the knowledge transfer condition, calling the problem-solving path data to compare the logical association of the knowledge points, calculating the weight matrix of the knowledge point dependency relationship, screening the knowledge point set with stable dependency relationship, and obtaining the transferable knowledge point set; S203: calling the transferable knowledge point set, screening available knowledge points according to the transfer knowledge point matching standard, calculating the matching index, and generating an interdisciplinary knowledge adaptation value.

[0011] As a further solution of the present invention, based on the knowledge point mastering state index and the interdisciplinary knowledge adaptation value, the specific steps of acquiring learning interaction data, calculating the learning mode characteristic value, calling the learning behavior pattern library to compare the operation behavior, parsing the learning mode category, judging the behavior pattern coherence according to the learning path feature model, and generating the learning behavior dynamic pattern parameters are as follows: S301: Based on the knowledge point mastering state index and the interdisciplinary knowledge adaptability value, the mouse dwelling time, page scrolling amplitude, knowledge point switching frequency, and problem-solving time distribution in the learning process interaction data are obtained, and the learning method characteristic value is calculated to obtain the knowledge point learning method characteristic value; S302: calling the characteristic value of the learning mode of the knowledge point, comparing the learner's operation behavior data, calculating the weight value of the differentiated operation behavior, classifying the learning mode categories, analyzing the correlation between the categories, screening the categories with high correlation of the operation behavior in the learning path, calculating the coherence index of the learning behavior pattern, screening the learning path with stable behavior pattern changes, and obtaining the learning behavior pattern correlation value; S303: calling the learning behavior pattern association value, calculating the learning path pattern matching degree, screening the behavior pattern of the path consistency standard, and generating the learning behavior dynamic pattern parameters.

[0012] As a further solution of the present invention, the learning path pattern matching degree calculation formula is specifically: ; Calculate the matching degree of learning path pattern, screen the behavior pattern of path coherence standard, and generate the dynamic pattern parameters of learning behavior; in, Represents the learning path pattern matching degree, Representative The learning behavior pattern association value of each knowledge point, Representative The weight of each knowledge point in the learning path, Represents the total number of knowledge points in the learning path.

[0013] As a further solution of the present invention, the specific steps of calling the learning behavior dynamic mode parameters, calculating the learning time gradient, trajectory change rate, switching attenuation rate, analyzing the knowledge point learning trajectory, identifying the learning stagnation point, and generating the learning path blocking factor are as follows: S401: calling the learning behavior dynamic mode parameters, parsing the knowledge point learning interval in the learning trajectory database, calculating the learning time gradient of adjacent knowledge points, and obtaining the knowledge point learning trajectory change rate; S402: calling the learning trajectory change rate of the knowledge point, screening out abnormal points in the learning progress, analyzing the knowledge point switching attenuation rate of the abnormal points, obtaining the knowledge points whose attenuation rates exceed the normal range, calculating the knowledge absorption rate of the abnormal points, screening out the knowledge points whose absorption rates are lower than the set threshold, screening out the blocked knowledge points according to the learning stagnation judgment standard, and obtaining the knowledge point blocking characteristic value; S403: calling the knowledge point blocking characteristic value, calculating the learning path blocking index, screening the knowledge point area with greater blocking influence, and generating the learning path blocking factor.

[0014] As a further solution of the present invention, the calculation formula of the switching attenuation rate of the knowledge point is specifically: ; Calculate the knowledge absorption rate of abnormal points, filter out knowledge points whose absorption rate is lower than the set threshold, filter out blocked knowledge points according to the learning stagnation judgment standard, and obtain the knowledge point blocking characteristic value; in, Represents the switching decay rate of knowledge points, Represents at a point in time The amount of knowledge point activity, Represents at a point in time The amount of knowledge point activity, Represents the total number of time points.

[0015] As a further solution of the present invention, based on the learning path retardation factor, the specific steps of calculating the learning progress change trend, determining the long-term mastery status of the knowledge point, and generating the knowledge mastery trend attenuation index are as follows: S501: Based on the learning path blocking factor, analyzing the time series of the knowledge point learning trend data, calculating the change trend index of the learning progress over time, and obtaining the learning trend change rate; S502: calling the learning trend change rate, comparing the knowledge forgetting curve database, calculating the knowledge point forgetting rate, screening differentiated knowledge points, calculating the learner's knowledge retention rate, combining the error rate growth model, calculating the knowledge point error rate change rate, screening the knowledge points with fast error rate growth, judging the long-term mastery state according to the mastery stability standard, and obtaining the knowledge point mastery attenuation value; S503: calling the knowledge point mastery decay value, calculating the overall knowledge mastery trend, screening the knowledge points that are unstable in long-term mastery, and generating a knowledge mastery trend decay index.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, by deeply analyzing the learning record data, comprehensively considering the cognitive load and error rate, accurately evaluating the mastery status of knowledge points, and strengthening the structural matching and transfer efficiency through cross-disciplinary knowledge mapping, the dynamic analysis of learning behavior enhances the understanding of path coherence, and by monitoring the changes in learning trajectories, optimizing the learning rhythm, and utilizing the forgetting curve and error rate trend, the prediction accuracy of knowledge mastery and retention is improved, making personalized learning recommendations more accurate, and effectively improving the application effect of educational technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0018] Figure 1 It is a schematic diagram of the steps of the present invention; Figure 2 is a flow chart of the steps of S1 of the present invention; Figure 3 is a flow chart of the steps of S2 of the present invention; Figure 4 is a flow chart of the steps of S3 of the present invention; Figure 5 is a flow chart of the steps of S4 of the present invention; Figure 6 This is a flow chart of the steps of S5 of the present invention. DETAILED DESCRIPTION

[0019] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0020] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "example" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "example" is intended to present the concept in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either of the two.

[0021] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same. "of", "corresponding, relevant" and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same.

[0022] In the embodiments of the present invention, sometimes the subscripts such as W 1 It may be written in non-subscript form such as W1. When the difference is not emphasized, the meaning is the same.

[0023] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0024] See also Figure 1 , the AI-driven learning intelligent assessment and diagnosis method includes the following steps: S1: Obtain the learning time, correct answer rate, and repeated learning times of knowledge points in the learning record database, extract the learning behavior parameters of knowledge points, calculate the learning time ratio and compare the time distribution of differentiated knowledge points, calculate the knowledge point mastery fluctuation value based on the correct answer rate change curve, call the cognitive load parameter library to analyze the learning burden change, calculate the knowledge point cognitive difficulty index based on the error rate distribution model, summarize the knowledge point mastery status in combination with the knowledge point stability calculation model, and generate the knowledge point mastery status index; S2: Based on the knowledge point mastery status index, the interdisciplinary knowledge mapping model is called to calculate the knowledge structure fit between the knowledge point and the cross-domain knowledge point, the spatial projection relationship of the knowledge point concept vector is analyzed, and the interdisciplinary knowledge point set that meets the migration conditions is extracted. The problem-solving path model is called to compare the logical relevance of the knowledge points, and the dependency weight of the knowledge points is calculated. The available knowledge points are screened according to the migration knowledge point matching standard to generate the interdisciplinary knowledge adaptation value; S3: Based on the knowledge point mastery status index and the interdisciplinary knowledge adaptation value, the mouse dwell time, page scrolling amplitude, knowledge point switching frequency, and problem-solving time distribution in the learning process interaction data are obtained, and the characteristic values ​​of the knowledge point learning methods are calculated. The learning behavior pattern library is called to compare the learner's operation behavior, and the learning method category is analyzed. According to the learning path feature model, the coherence of the behavior pattern between knowledge points is judged, and the learning behavior dynamic pattern parameters are generated; S4: Call the learning behavior dynamic mode parameters, parse the knowledge point learning interval in the learning trajectory database, calculate the learning time gradient of adjacent knowledge points, obtain the learning trajectory change rate of knowledge points, screen the abnormal points of learning progress and calculate the knowledge point switching attenuation rate, call the cognitive load dynamic adjustment library to calculate the knowledge absorption rate of abnormal points, screen the blocking knowledge points according to the learning stagnation judgment criteria, and generate the learning path blocking factor; S5: Based on the learning path blocking factor, analyze the time series in the knowledge point learning trend data, calculate the changing trend of learning progress over time, call the knowledge forgetting curve database to compare the knowledge point forgetting rate, calculate the learner's knowledge retention rate in differentiated knowledge points, and calculate the error rate change rate of knowledge points in combination with the error rate growth model. According to the mastery stability standard, determine the long-term mastery status of the knowledge points and generate the knowledge mastery trend attenuation index.

[0025] The knowledge point mastery status index includes the proportion of learning time, mastery fluctuation value, cognitive difficulty index, and mastery stability. The interdisciplinary knowledge adaptation value includes the knowledge structure fitting degree, concept vector projection relationship, migration knowledge point set, dependency weight, and matching standard. The learning behavior dynamic pattern parameters include learning method characteristic value, learning method category, and behavior consistency parameter. The learning path blocking factors include learning time gradient, learning trajectory change rate, switching attenuation rate, knowledge absorption rate, and blocking knowledge points. The knowledge mastery trend attenuation index includes knowledge retention rate, forgetting rate, error rate change rate, and long-term mastery status.

[0026] See also Figure 2 , the specific steps of S1 are: S101: Obtaining the learning time, answer accuracy, and repeated learning times of knowledge points in the learning record database, calculating the learning time proportion of knowledge points, extracting knowledge points with large differences in proportion, and obtaining differentiated knowledge point time distribution values; First, retrieve the total learning time of each knowledge point, the student's answer record and the number of times each knowledge point is studied from the database of the learning management system. These data are automatically recorded and updated based on the students' daily learning activities. For example, when a student is learning algebra knowledge points, the system will record the start and end time of his or her learning, the number of times he or she answered questions, and whether they were correct or not. In this way, the system can obtain the student's total learning time, number of answers and correctness rate on algebra knowledge points. For all students' data sets, we calculate the average learning time and average correctness rate for each knowledge point. In order to ensure the accuracy and reliability of the data, it is necessary to eliminate outliers, such as records with abnormally short or abnormally long learning time. Next, To normalize the learning time of each knowledge point and calculate its proportion in the total learning time. During the normalization process, first find the sum of the learning time of all knowledge points, and then divide the learning time of each knowledge point by the sum to get the learning time proportion of each knowledge point. Through this data, we can observe whether the time allocation of different knowledge points in the learning process is balanced, and further screen out knowledge points with large differences in proportions. These knowledge points indicate that students may have differences in learning difficulty or learning interest in these aspects. By comparing the differences in the proportions of each knowledge point, we can select the knowledge point with the largest standard deviation as the group with the largest difference in proportions, so that we can get differentiated knowledge point time distribution values.

[0027] S102: calling the time distribution value of differentiated knowledge points and the change trend data of the answer accuracy, calculating the fluctuation range index of the answer accuracy, calculating the mastery fluctuation characteristic value for the knowledge points with a large fluctuation range, matching it with the learning burden parameter library, calculating the cognitive burden change index, screening the knowledge points with significant mastery fluctuation, analyzing the mastery stability in combination with their change trend, and obtaining the mastery fluctuation value of the knowledge points; First, use the differentiated knowledge point time distribution value obtained in the previous step and combine it with the answer accuracy change data of each knowledge point. For example, if the learning time proportion of a knowledge point is extremely high but the accuracy changes greatly, it means that this knowledge point may be both important and difficult to master. For each knowledge point, we calculate the fluctuation range of its answer accuracy. The fluctuation range is obtained by calculating the difference between the maximum and minimum values ​​of the answer accuracy. This step is to find out the knowledge points that are important in time distribution but have large fluctuations in answer accuracy, and mark these knowledge points as knowledge points with large fluctuations. Next, further calculate the mastery fluctuation characteristic value of these knowledge points. This calculation includes statistics on the standard deviation of the answer accuracy. The larger the standard deviation, the more likely the student is to master it. The greater the fluctuation in the degree of mastery of this knowledge point, the greater the fluctuation in the degree of mastery of the knowledge point. Then these characteristic values ​​are matched with the learning burden parameter library. The learning burden parameter library may include the difficulty of the knowledge point, historical mastery data, etc. During the matching process, the cognitive burden change index is calculated by finding records in the parameter library that are close to the characteristic values. This index can be used to evaluate the changes in the psychological and cognitive burden that students bear during the learning process. After screening out the knowledge points with significant mastery fluctuations, the trend of changes in the accuracy of answering questions for these knowledge points is analyzed. This analysis calculates the trend line of the knowledge point answer data, such as using the least squares method to fit the trend line of the change in the accuracy of answering questions, so as to obtain the knowledge point mastery fluctuation value. This value can reflect the stability of students' mastery of knowledge points.

[0028] S103: calling the knowledge point mastery fluctuation value, calculating the knowledge point answer error rate distribution index, combining the knowledge point cognitive difficulty index and the mastery fluctuation value, and generating the knowledge point mastery status index; The calculation formula for the knowledge point answer error rate distribution index is as follows: ; in, Represents the distribution index of the error rate of answering questions at knowledge points, Representative The error rate of this knowledge point in the test, Represents the average error rate of this knowledge point in all tests. Represents the number of tests for this knowledge point. Representative The learners’ error deviation in answering questions on this knowledge point is represents the total number of all learners, Represents the cognitive difficulty index of this knowledge point, Represents the weight coefficient of this knowledge point in the question, Represents the mastery fluctuation value of this knowledge point, Represents the mastery stability benchmark value of this knowledge point in the learning group; The goal of the formula calculation is to evaluate the distribution index of the error rate of knowledge point answers ,This indicator combines three main parts: the average error deviation, the standard deviation of the ,error deviation of knowledge points, and a weighting factor that takes into account the ,fluctuation of knowledge point mastery and the cognitive difficulty index.

[0029] Average Error Deviation: The first component calculates the average error deviation of the knowledge point in all tests. The error rate of each test is obtained through actual test data. , and calculate the average error rate For example, if the error rates of a knowledge point in four tests are 0.2, 0.3, 0.25, and 0.4 respectively, then and . We further calculate the deviation for each test and take the average: ; Error Standard Deviation: Part 2 uses the square root sum divided by the total number of learners To calculate the standard deviation. Assume that for four learners, the error deviation of answering questions on the same knowledge point is They are 0.1, 0.05, 0.15, and 0.1 respectively. , calculated as follows: ; Weighting factors: The last part combines cognitive difficulty , assumed to be 0.6 (usually determined by educational experts), and the weight coefficient , assuming it is 0.5 (based on the importance of the knowledge point in the course). Mastering Fluctuations Assuming 0.25, grasp the stability benchmark value Assuming it is 0.2, the calculation is as follows: ; Combine the results of these three parts: ; The results show that the knowledge point answer error rate distribution index The value is 0.4545, which combines the average deviation of the error rate, the discrete degree of the error deviation, and the influence of mastery fluctuation and cognitive difficulty. When generating the knowledge point mastery status index, this indicator helps to measure the learner's overall mastery of the knowledge point and reflect the stability of their mastery.

[0030] See also Figure 3 , the specific steps of S2 are: S201: Based on the knowledge point mastering status index, obtain interdisciplinary knowledge mapping data, calculate the structural matching degree between the knowledge point and the cross-domain knowledge point, select the knowledge points that meet the matching standards, and obtain the interdisciplinary knowledge matching value; First, the knowledge point mastery status index is called. This index is calculated based on parameters such as the student's correct answer rate, mastery fluctuation value, and knowledge point cognitive difficulty index. According to the numerical value of the mastery status index, each knowledge point is preliminarily classified to screen out knowledge points with a relatively stable mastery status index. Then, the interdisciplinary knowledge mapping data is extracted. This data comes from the association matrix between knowledge points. The matrix is ​​established based on the association data between different subjects. For example, between mathematics and physics, the trigonometric function knowledge point is associated with the law of optical refraction, so the knowledge mapping data of the two are included in the matrix. When calculating the structural matching degree between knowledge points and cross-domain knowledge points, first extract the set of interdisciplinary knowledge points corresponding to each knowledge point, and calculate the similarity of their structural relationships. In the specific calculation process, obtain the connection relationship of the current knowledge point within the subject, count the dependence strength between it and other knowledge points, and obtain the connection relationship of interdisciplinary knowledge points at the same time. Use vector operations to calculate the structural similarity between the two. Assume that the knowledge connection strength vector of a certain knowledge point A within the subject is , the connection strength vector of interdisciplinary knowledge point B is The matching degree calculation formula is: , the matching degree between knowledge points A and B is obtained after calculation. Finally, the matching degree of all knowledge points is screened and the matching degree threshold is set. ,when , it is considered that the knowledge point meets the matching criteria, and the selected matching knowledge points form the interdisciplinary knowledge matching value.

[0031] S202: calling the interdisciplinary knowledge matching value, analyzing the spatial projection relationship of the knowledge point concept vectors, calculating the distance matrix between the concept vectors, extracting the knowledge point set with close spatial projection relationship, screening the interdisciplinary knowledge point set that meets the knowledge transfer conditions, calling the problem-solving path data to compare the logical association of the knowledge points, calculating the weight matrix of the knowledge point dependency relationship, screening the knowledge point set with stable dependency relationship, and obtaining the transferable knowledge point set; First, we extract a set of knowledge points with a high matching degree from the interdisciplinary knowledge matching value. For these knowledge points, we construct a concept vector matrix of the knowledge points based on the vector representation of the knowledge point concepts. For example, the vector of the knowledge point "Newton's Second Law" may be composed of concepts such as mass, acceleration, and force, while the vector of "Kinetic Energy Theorem" may be composed of concepts such as work, energy, and speed. When analyzing the spatial projection relationship of the knowledge point concept vectors, we calculate the projection of the knowledge point vectors in the multidimensional concept space, obtain the projection similarity between the knowledge points, and calculate the distance matrix between the concept vectors. The calculation formula is: ,in , The concept vectors of two knowledge points are extracted, and a set of knowledge points with a small distance is formed to form a set of knowledge points with close spatial projection relationship. On this basis, the set of interdisciplinary knowledge points that meet the conditions of knowledge transfer is screened. The screening process takes into account the projection proximity threshold of the knowledge points. ,when When the two knowledge points are considered to have the possibility of migration, the problem-solving path data is called, the sequence of knowledge points involved in the student's problem-solving process is analyzed, the logical association of the knowledge points is calculated, and the weight matrix of the knowledge point dependency relationship is established. The elements of this matrix are Representative knowledge points To knowledge point The dependence strength is calculated by counting the knowledge points in the historical answer data. After the appearance, knowledge points The probability of being answered correctly, finally screening the set of knowledge points with stable dependency relationships, and setting the dependency stability threshold ,when When thinking about knowledge points Depends on knowledge points The relationship between them is stable, and the set of knowledge points screened out is the set of transferable knowledge points.

[0032] S203: calling the set of transferable knowledge points, screening available knowledge points according to the transfer knowledge point matching standard, calculating the matching index, and generating the interdisciplinary knowledge adaptation value; First, extract the list of knowledge points in the set of transferable knowledge points, and calculate the migration matching index for each knowledge point. This index is based on the weighted calculation of the structural matching, spatial projection similarity, and logical dependency stability between knowledge points. Assume that the knowledge points And knowledge points The matching degree is , the calculation formula is ,in , , is the weight coefficient. When determining the weight coefficient, it is adjusted based on historical data. For example, for the knowledge point migration task, a higher value of logical dependency weight may be given and a matching threshold is set. ,when , the knowledge point is considered suitable for migration, and finally the matching index of all knowledge points is calculated to form the interdisciplinary knowledge adaptation value.

[0033] See also Figure 4 , the specific steps of S3 are: S301: Based on the knowledge point mastering state index and the interdisciplinary knowledge adaptation value, the mouse dwelling time, page scrolling amplitude, knowledge point switching frequency, and problem-solving time distribution in the learning process interaction data are obtained, and the learning method characteristic value is calculated to obtain the knowledge point learning method characteristic value; First, the knowledge point mastering status index and the interdisciplinary knowledge adaptation value are extracted, and the interaction data of students in the learning process are obtained from the learning record system. The mouse dwell time is calculated by counting the mouse hovering time of students on each knowledge point page. For example, if a student stays on the mathematics knowledge point "Calculus" page for 120 seconds and on the "Statistics" page for 60 seconds, the recorded mouse dwell time is 120 seconds and 60 seconds respectively. The page scrolling amplitude is calculated by analyzing the user's scrolling distance on the learning page. Assuming that the total length of the page is 1000 pixels and the student scrolls 800 pixels on the page, the page scrolling amplitude is calculated as , the knowledge point switching frequency refers to the number of times students switch between different knowledge points in a unit time. The unit time is set to 10 minutes. If a student switches knowledge points 5 times in 10 minutes, the switching frequency is times / minute. Problem-solving time distribution refers to the time it takes students to solve problems at different knowledge points. It is calculated by counting the average problem-solving time for each knowledge point. For example, the average problem-solving time for the knowledge point "Geometry" is 45 seconds, while the average problem-solving time for the knowledge point "Algebra" is 30 seconds. After these data are obtained, the learning interaction data of each student is normalized to eliminate data scale differences. Then, the learning method characteristic value is calculated. This characteristic value integrates the standardized values ​​of the mouse dwell time, page scrolling amplitude, knowledge point switching frequency, and problem-solving time distribution, and performs weighted calculation. For example, the weight of the mouse dwell time is set to 0.4, the weight of the page scrolling amplitude is set to 0.2, the weight of the knowledge point switching frequency is set to 0.3, and the weight of the problem-solving time distribution is set to 0.1. Then the learning method characteristic value is calculated as follows: ; in, Represents the normalized value of the dwell time (MouseTime), Represents the normalized value of the scroll amplitude (ScrollPercentage), Represents the normalized value of the knowledge point switching frequency (SwitchCount), Represents the standardized value of the problem-solving time distribution (Duration of Tasks). After the calculation is completed, the characteristic value of the knowledge point learning method is obtained.

[0034] S302: calling the characteristic value of the learning method of the knowledge point, comparing the learner's operation behavior data, calculating the weight value of the differentiated operation behavior, classifying the learning method categories, analyzing the correlation between the categories, screening the categories with high correlation of operation behaviors in the learning path, calculating the coherence index of the learning behavior pattern, screening the learning path with stable behavior pattern changes, and obtaining the learning behavior pattern correlation value; First, we extract the learning behavior data of different students from the learning style characteristic value data set, compare the learning style characteristic values ​​of students at different knowledge points, calculate the difference between various operation behaviors at different knowledge points, set the standard deviation of different behaviors as an indicator to measure the degree of differentiation, and use the correlation between the learning style characteristic value and various operation behaviors when calculating the weight value of the operation behavior. (ObservedBehavior) is learning eigenvalues The weight value on (LearningStyleFeatureValue) is calculated as ; in, Representative The learning style characteristic value of each student, Represents its operational behavior data, and They are the means of the characteristic values ​​and the operation behavior data respectively. After calculating the weight values, students are classified according to the characteristics of the operation behaviors, and the learning methods are divided into high-interaction categories (long mouse dwell time, large page scrolling amplitude, high knowledge point switching frequency), low-interaction categories (short dwell time, small scrolling amplitude, low switching frequency), and balanced categories (all parameters are evenly distributed). When analyzing the correlation between categories, the transfer probability between categories is calculated, and the probability of students in a certain category transferring to another category is counted. For example, the probability of high-interaction students transferring to low-interaction classes after a period of time is 30%, while the probability of low-interaction students transferring to high-interaction classes is 15%. When screening categories with high correlation in operation behaviors in the learning path, set the correlation threshold. (CategoryBehaviorThreshold), when the transition probability of a category is greater than , the category is considered to have a strong correlation with the operation behavior, and then the consistency index of the learning behavior pattern is calculated. This index is based on whether the student maintains consistent operation behavior during the learning process. Assuming that a student belongs to the high interaction category in 4 of the 5 learning processes, the consistency index is calculated as ; in, stands for Learning Behavior Coherence, Represents the number of high-interaction categories (High-InteractionCount), Represents the total number of times (TotalCount), setting the continuity threshold ,when , the student’s learning path is considered stable, and finally the learning path with stable behavior pattern changes is screened to obtain the learning behavior pattern association value.

[0035] S303: calling the learning behavior pattern association value, calculating the learning path pattern matching degree, screening the behavior pattern of the path coherence standard, and generating the learning behavior dynamic pattern parameters; The specific calculation formula for learning path pattern matching is: ; Calculate the matching degree of learning path pattern, screen the behavior pattern of path coherence standard, and generate the dynamic pattern parameters of learning behavior; in, Represents the learning path pattern matching degree, Representative The learning behavior pattern association value of each knowledge point, Representative The weight of each knowledge point in the learning path, Represents the total number of knowledge points in the learning path; Learning path pattern matching It is calculated by weighted average method, which takes into account the importance of each knowledge point in the overall learning path and the association value of its behavior pattern. For example, if a learning path contains three knowledge points, their pattern association value is They are 0.8, 0.6, and 0.9 respectively, and the corresponding weights They are 0.5, 0.3, and 0.2 respectively. These weights represent the importance or frequency of each knowledge point in the learning process, and may be set based on previous learning activity data or the advice of education experts.

[0036] The calculation formula is as follows: ; This result shows that the average pattern matching degree of the entire learning path is 0.76, which means that the learner's behavior on this path matches the preset pattern to a high degree, indicating that the design of the learning path is consistent with the learner's actual behavior. Therefore, the pattern matching degree can be used to further generate dynamic pattern parameters of learning behavior. These parameters may include customized learning strategies for each knowledge point to optimize learning effects and improve learning continuity.

[0037] See also Figure 5 , the specific steps of S4 are: S401: calling the learning behavior dynamic mode parameters, parsing the knowledge point learning interval in the learning trajectory database, calculating the learning time gradient of adjacent knowledge points, and obtaining the learning trajectory change rate of the knowledge points; First, the learning path data is extracted from the learning behavior dynamic pattern parameters, and the learner's learning time record at each knowledge point is obtained. The knowledge point learning interval is calculated as the time difference between the entry time and the exit time of the knowledge point. For example, if a student enters the "Newton's Second Law" knowledge point at 09:00 and leaves at 09:20, the learning interval is 20 minutes. After extracting the learning interval data of all knowledge points in turn, the learning time gradient of adjacent knowledge points is calculated. The gradient represents the change in the student's learning time at different knowledge points. When calculating specifically, the learning time of the current knowledge point is taken as the difference between the learning time of the previous knowledge point and the learning time of the previous knowledge point. For example, if the student's learning time on knowledge point A is 25 minutes, and the learning time on knowledge point B drops to 10 minutes, it means that the learning time gradient has changed significantly. The time gradient changes between all adjacent knowledge points are counted, and the changing trend in the overall learning path is extracted. If a student's learning time gradient shows an overall downward trend, it may mean that his or her interest in learning subsequent knowledge points has declined or is difficult to understand. If the learning time gradient fluctuates greatly, it means that his or her learning status at different knowledge points is unstable. After all calculations are completed, the change rate of the knowledge point learning trajectory is obtained.

[0038] S402: calling the learning trajectory change rate of knowledge points, screening out abnormal points in learning progress, analyzing the knowledge point switching attenuation rate of the abnormal points, obtaining knowledge points whose attenuation rates exceed the normal range, calculating the knowledge absorption rate of the abnormal points, screening out knowledge points whose absorption rates are lower than the set threshold, screening out the blocked knowledge points according to the learning stagnation judgment standard, and obtaining the knowledge point blocking characteristic value; The calculation formula for the switching attenuation rate of knowledge points is as follows: ; Calculate the knowledge absorption rate of abnormal points, filter out knowledge points whose absorption rate is lower than the set threshold, filter out blocked knowledge points according to the learning stagnation judgment standard, and obtain the knowledge point blocking characteristic value; in, Represents the switching decay rate of knowledge points, Represents at a point in time The amount of knowledge point activity, Represents at a point in time The amount of knowledge point activity, Represents the total number of time points; formula Designed to calculate the accelerated decay of knowledge point activity, it measures the cube of the change in activity between each time point and the previous time point and normalizes it relative to the square root of the previous activity. This calculation not only measures the magnitude of the change, but also emphasizes the acceleration of the change, making small activity changes more significant in low cardinality situations, and is suitable for capturing sudden changes in learning behavior.

[0039] Given activity data: The total time points are 200, 180, 160, and 140. .

[0040] The formula is as follows: 1. First, calculate the cube of the change in activity at each time point relative to the square root of the previous activity:

[0041] ; Then, calculate the average of these changes: ; Therefore, the switching decay rate of knowledge points , indicating that when the activity of a knowledge point changes between consecutive time points, the magnitude of the change shows a certain accelerated decay after cubic processing. This decay rate can be used to identify potential learning anomalies, especially when learning activities suddenly decrease. This calculation method with increased complexity helps to analyze the dynamic changes of learning behavior more finely, thereby providing more accurate data support for educational intervention.

[0042] S403: calling the knowledge point blocking characteristic value, calculating the learning path blocking index, screening the knowledge point area with greater blocking influence, and generating the learning path blocking factor; First, extract the set of knowledge points with higher blocking characteristic values ​​in the learning path, and calculate the blocking influence of the overall learning path based on the weight of the blocking characteristic values. For example, if there are multiple blocking knowledge points in a student's learning path, and the blocking characteristic values ​​of these knowledge points are high, the overall blocking index of the path is also high. When screening the knowledge point areas with greater blocking influence, count the distribution of blocking knowledge points in the entire learning path. If the proportion of blocking knowledge points in a certain knowledge unit (such as calculus) exceeds 50%, it is considered that the blocking in this area is more serious. Further analyze the learning time, correct answer rate and interactive behavior of the knowledge points in this area to determine the main influencing factors of the blocking. For example, if it is found that the learning time of the knowledge points in this area is generally short and the correct answer rate is extremely low, it may indicate that the knowledge difficulty in this area is too high or the students are not interested in this type of knowledge. Finally, based on these analyses, generate the learning path blocking factor, which can be used to measure the impact of the blocking phenomenon in the learning process and provide a basis for subsequent learning path optimization.

[0043] See also Figure 6 , the specific steps of S5 are: S501: Based on the learning path blocking factor, analyze the time series of the knowledge point learning trend data, calculate the learning progress change trend index over time, and obtain the learning trend change rate; First, extract the knowledge point set in the high-blocking area from the learning path blocking factor, and obtain the learning data of the knowledge point at different time stages. The time series of learning trend data consists of variables such as learning time, answering accuracy, and number of interactions at each time point. For example, a student studies the "Calculus" knowledge point for 60 minutes in the first week, which is reduced to 30 minutes in the second week, and only 10 minutes in the third week. The learning trend of this knowledge point is in a downward state. When calculating the trend indicator of learning progress over time, obtain the learning data of the knowledge point at each time point, and calculate the learning time, accuracy, interaction behavior and other variables. The amount changes over time. For example, in different weeks, if the study time is reduced by 50% but the accuracy does not change, it means that the student may have mastered the knowledge point. If the study time is reduced but the accuracy also decreases, it indicates that the knowledge point may not be fully mastered. The learning trend change rate is further extracted. The change rate reflects the dynamic characteristics of the learning progress. For example, if the study time of a knowledge point is reduced by 20% per week, its trend change rate is -20%. When the trend change rate drops by more than the set threshold, it indicates that there may be problems with the learning progress of the knowledge point. After all calculations are completed, the learning trend change rate is obtained.

[0044] S502: calling the learning trend change rate, comparing the knowledge forgetting curve database, calculating the knowledge point forgetting rate, screening the differentiated knowledge points, calculating the learner's knowledge retention rate, combining the error rate growth model, calculating the knowledge point error rate change rate, screening the knowledge points with fast error rate growth, judging the long-term mastery state according to the mastery stability standard, and obtaining the knowledge point mastery decay value; First, the learning trend change rate data is extracted, and the standard forgetting rate data is obtained from the knowledge forgetting curve database. The database stores the forgetting curve models of different types of knowledge points. For example, for knowledge points such as "physical formulas", the standard forgetting rate may be faster, while the forgetting rate of knowledge points such as "mathematical theorems" may be relatively slow. By comparing the learner's learning trend change rate with the forgetting curve, the forgetting rate of its knowledge point is calculated. For example, if the learning trend change rate of a knowledge point decreases by 50%, and the standard forgetting rate of the knowledge point is 30%, it means that the actual forgetting rate of the knowledge point is higher than the standard value. When screening differentiated knowledge points, select knowledge points whose forgetting rate is significantly higher or lower than the standard value. For example, if the forgetting rate of a knowledge point is 40% higher than the standard value, it means that the knowledge point is easy to be forgotten. When calculating the learner's knowledge retention rate, obtain the correct rate of the answer at different time points, and calculate the degree of decline in the correct rate over time. For example, if a student's initial correct rate on knowledge point A is 90%, and it drops to 50% after two weeks, then its retention rate is ,Further combined with the error rate growth model, the error rate change rate of knowledge points is calculated. This rate measures the error rate growth of knowledge points over time. For example, if the error rate of a knowledge point increases by 10% per week, it indicates that the mastery of the knowledge point is rapidly deteriorating. When screening knowledge points with fast error rate growth, a set of knowledge points with a higher error rate change rate is extracted, and the long-term mastery status is determined based on the mastery stability standard. For example, if the knowledge retention rate of a knowledge point is lower than 60% and the error rate growth rate exceeds 15%, it is considered that the long-term mastery status of the knowledge point is unstable. Finally, the mastery attenuation value of all knowledge points is calculated.

[0045] S503: calling the knowledge point mastery decay value, calculating the overall knowledge mastery trend, screening the knowledge points that are unstable in long-term mastery, and generating the knowledge mastery trend decay index; First, extract the learners' mastery decay value data on all knowledge points, and count the knowledge point decay trends in the learning path. For example, if the mastery decay value of the knowledge points in a certain learning path decreases by 30% on average, it indicates that the overall knowledge mastery trend is declining. When screening knowledge points with unstable long-term mastery, extract the knowledge points with large fluctuations in decay value changes, and analyze their learning time, accuracy, forgetting rate and other variables. For example, if the learning time of a certain knowledge point is long, but the mastery decay value continues to decrease, it may indicate that the knowledge point is difficult to maintain for a long time. Further obtain the changes in the overall mastery trend. If more than 50% of the knowledge points in the learning path show a mastery decay trend, it is considered that the knowledge mastery of the path is long-term unstable. Finally, calculate the overall knowledge mastery trend of all learners and generate a knowledge mastery trend decay index.

[0046] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. AI-driven learning intelligent assessment and diagnosis method, characterized by: The following steps are involved: S1: Obtain the learning time, answer accuracy, and number of repeated learning in the learning record database, extract learning behavior parameters, calculate the knowledge point mastery status index, and generate the knowledge point mastery status index; S2: Based on the knowledge point mastering status index, call the interdisciplinary knowledge mapping model to calculate the knowledge structure fitting degree, screen the interdisciplinary knowledge points, and generate the interdisciplinary knowledge fitness value; S3: Based on the knowledge point mastering state index and the interdisciplinary knowledge adaptation value, obtain learning interaction data, calculate the learning mode characteristic value, call the learning behavior pattern library to compare the operation behavior, analyze the learning mode category, judge the behavior pattern coherence according to the learning path feature model, and generate the learning behavior dynamic pattern parameters; S4: calling the learning behavior dynamic mode parameters, calculating the learning time gradient, trajectory change rate, switching attenuation rate, analyzing the knowledge point learning trajectory, identifying the learning stagnation point, and generating the learning path blocking factor; S5: Based on the learning path retardation factor, calculate the learning progress change trend, determine the long-term mastery status of the knowledge points, and generate a knowledge mastery trend attenuation index.

2. The AI-driven learning intelligent assessment and diagnosis method according to claim 1, characterized in that: The knowledge point mastery status index includes the proportion of learning time, mastery fluctuation value, cognitive difficulty index, and mastery stability. The interdisciplinary knowledge adaptation value includes knowledge structure fitting, concept vector projection relationship, migration knowledge point set, dependency weight, and matching standard. The learning behavior dynamic mode parameters include learning method characteristic values, learning method categories, and behavior consistency parameters. The learning path blocking factors include learning time gradient, learning trajectory change rate, switching attenuation rate, knowledge absorption rate, and blocked knowledge points. The knowledge mastery trend attenuation index includes knowledge retention rate, forgetting rate, error rate change rate, and long-term mastery status.

3. The AI-driven learning intelligent assessment and diagnosis method according to claim 1 is characterized in that: Obtain the learning time, answer accuracy, and number of repeated learning times in the learning record database, extract learning behavior parameters, and calculate the knowledge point mastery status index. The specific steps for generating the knowledge point mastery status index are: S101: Obtaining the learning time, answer accuracy, and repeated learning times of knowledge points in the learning record database, calculating the learning time proportion of knowledge points, extracting knowledge points with large differences in proportion, and obtaining differentiated knowledge point time distribution values; S102: calling the time distribution value of the differentiated knowledge points and the answer accuracy change trend data, calculating the fluctuation range index of the answer accuracy, calculating the mastery fluctuation characteristic value for the knowledge points with a large fluctuation range, matching it with the learning burden parameter library, calculating the cognitive burden change index, screening the knowledge points with significant mastery fluctuation, analyzing the mastery stability in combination with their change trend, and obtaining the knowledge point mastery fluctuation value; S103: calling the knowledge point mastery fluctuation value, calculating the knowledge point answer error rate distribution index, combining the knowledge point cognitive difficulty index and the mastery fluctuation value, and generating the knowledge point mastery status index.

4. The AI-driven learning intelligent assessment and diagnosis method according to claim 3 is characterized in that: The calculation formula for the knowledge point answer error rate distribution index is specifically: ； in, Represents the distribution index of the error rate of answering questions at knowledge points, Representative The error rate of this knowledge point in the test, Represents the average error rate of this knowledge point in all tests. Represents the number of tests for this knowledge point. Representative The learners’ error deviation in answering questions on this knowledge point is represents the total number of all learners, Represents the cognitive difficulty index of this knowledge point, Represents the weight coefficient of this knowledge point in the question, Represents the mastery fluctuation value of this knowledge point, Represents the baseline value of the mastery stability of this knowledge point in the learning group.

5. The AI-driven learning intelligent assessment and diagnosis method according to claim 1, characterized in that: Based on the knowledge point mastering status index, the specific steps of calling the interdisciplinary knowledge mapping model to calculate the knowledge structure fitting degree, screening the interdisciplinary knowledge points, and generating the interdisciplinary knowledge fitness value are as follows: S201: Based on the knowledge point mastering status index, obtain interdisciplinary knowledge mapping data, calculate the structural matching degree between the knowledge point and the cross-domain knowledge point, select the knowledge points whose matching degree meets the standard, and obtain the interdisciplinary knowledge matching degree value; S202: calling the interdisciplinary knowledge matching value, analyzing the spatial projection relationship of the knowledge point concept vectors, calculating the distance matrix between the concept vectors, extracting the knowledge point set with close spatial projection relationship, screening the interdisciplinary knowledge point set that meets the knowledge transfer condition, calling the problem-solving path data to compare the logical association of the knowledge points, calculating the weight matrix of the knowledge point dependency relationship, screening the knowledge point set with stable dependency relationship, and obtaining the transferable knowledge point set; S203: calling the transferable knowledge point set, screening available knowledge points according to the transfer knowledge point matching standard, calculating the matching index, and generating an interdisciplinary knowledge adaptation value.

6. The AI-driven learning intelligent assessment and diagnosis method according to claim 1, characterized in that: Based on the knowledge point mastering state index and the interdisciplinary knowledge adaptation value, the specific steps of acquiring learning interaction data, calculating the learning mode characteristic value, calling the learning behavior pattern library to compare the operation behavior, analyzing the learning mode category, judging the behavior pattern coherence according to the learning path characteristic model, and generating the learning behavior dynamic pattern parameters are as follows: S301: Based on the knowledge point mastering state index and the interdisciplinary knowledge adaptability value, the mouse dwelling time, page scrolling amplitude, knowledge point switching frequency, and problem-solving time distribution in the learning process interaction data are obtained, and the learning method characteristic value is calculated to obtain the knowledge point learning method characteristic value; S302: calling the characteristic value of the learning mode of the knowledge point, comparing the learner's operation behavior data, calculating the weight value of the differentiated operation behavior, classifying the learning mode categories, analyzing the correlation between the categories, screening the categories with high correlation of the operation behavior in the learning path, calculating the coherence index of the learning behavior pattern, screening the learning path with stable behavior pattern changes, and obtaining the learning behavior pattern correlation value; S303: calling the learning behavior pattern association value, calculating the learning path pattern matching degree, screening the behavior pattern of the path consistency standard, and generating the learning behavior dynamic pattern parameters.

7. The AI-driven learning intelligent assessment and diagnosis method according to claim 6, characterized in that: The learning path pattern matching degree calculation formula is specifically as follows: ； Calculate the matching degree of learning path pattern, screen the behavior pattern of path coherence standard, and generate the dynamic pattern parameters of learning behavior; in, Represents the learning path pattern matching degree, Representative The learning behavior pattern association value of each knowledge point, Representative The weight of each knowledge point in the learning path, Represents the total number of knowledge points in the learning path.

8. The AI-driven learning intelligent assessment and diagnosis method according to claim 1, characterized in that: The specific steps of calling the learning behavior dynamic mode parameters, calculating the learning time gradient, trajectory change rate, switching attenuation rate, analyzing the knowledge point learning trajectory, identifying the learning stagnation point, and generating the learning path blocking factor are as follows: S401: calling the learning behavior dynamic mode parameters, parsing the knowledge point learning interval in the learning trajectory database, calculating the learning time gradient of adjacent knowledge points, and obtaining the knowledge point learning trajectory change rate; S402: calling the learning trajectory change rate of the knowledge point, screening out abnormal points in the learning progress, analyzing the knowledge point switching attenuation rate of the abnormal points, obtaining the knowledge points whose attenuation rates exceed the normal range, calculating the knowledge absorption rate of the abnormal points, screening out the knowledge points whose absorption rates are lower than the set threshold, screening out the blocked knowledge points according to the learning stagnation judgment standard, and obtaining the knowledge point blocking characteristic value; S403: calling the knowledge point blocking characteristic value, calculating the learning path blocking index, screening the knowledge point area with greater blocking influence, and generating the learning path blocking factor.

9. The AI-driven learning intelligent assessment and diagnosis method according to claim 8, characterized in that: The calculation formula for the switching attenuation rate of the knowledge point is specifically: ； Calculate the knowledge absorption rate of abnormal points, filter out knowledge points whose absorption rate is lower than the set threshold, filter out blocked knowledge points according to the learning stagnation judgment standard, and obtain the knowledge point blocking characteristic value; in, Represents the switching decay rate of knowledge points, Represents at a point in time The amount of knowledge point activity, Represents at a point in time The amount of knowledge point activity, Represents the total number of time points.

10. The AI-driven learning intelligent assessment and diagnosis method according to claim 1, characterized in that: Based on the learning path retardation factor, the specific steps of calculating the learning progress change trend, determining the long-term mastery status of the knowledge points, and generating the knowledge mastery trend attenuation index are as follows: S501: Based on the learning path blocking factor, analyzing the time series of the knowledge point learning trend data, calculating the change trend index of the learning progress over time, and obtaining the learning trend change rate; S502: calling the learning trend change rate, comparing the knowledge forgetting curve database, calculating the knowledge point forgetting rate, screening differentiated knowledge points, calculating the learner's knowledge retention rate, combining the error rate growth model, calculating the knowledge point error rate change rate, screening the knowledge points with fast error rate growth, judging the long-term mastery state according to the mastery stability standard, and obtaining the knowledge point mastery attenuation value; S503: calling the knowledge point mastery decay value, calculating the overall knowledge mastery trend, screening the knowledge points that are unstable in long-term mastery, and generating a knowledge mastery trend decay index.

Images