A personalized learning monitoring system based on user behavior data
By integrating learning behavior data and physiological feedback through a personalized learning monitoring system, the system dynamically adjusts learning paths and tasks, solving the problems of lagging learning paths and incomplete feedback in personalized education, and realizing intelligent personalized learning support.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-20
- Publication Date
- 2026-03-27
AI Technical Summary
Current personalized education suffers from outdated learning paths, limited recommendation strategies, and inadequate feedback mechanisms, making it difficult to provide accurate and dynamic personalized learning support.
Design a personalized learning monitoring system based on user behavior data, including a data acquisition module, a learning goal setting module, a self-assessment and feedback module, an emotion monitoring module, and a dynamic task adjustment module. By integrating learning behavior data, physiological feedback, and interaction logs, the system dynamically adjusts learning tasks and paths to provide personalized learning support.
It achieves highly intelligent personalized learning support, dynamically senses the learner's cognitive and emotional state, automatically identifies state changes, reduces cognitive burden, provides accurate learning path suggestions and feedback, and forms a complete closed-loop learning process.
Smart Images
Figure CN121169645B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of personalized education, and in particular to a personalized learning monitoring system based on user behavior data. BACKGROUND
[0002] Personalized education is a teaching mode that customizes educational content, teaching methods, and assessment methods based on learners' needs, interests, abilities, and learning pace. With the development of information technology, especially the progress of big data, artificial intelligence, and educational technology, personalized education has gradually become an important direction of modern educational reform. Its core concept is "student-centered", emphasizing providing personalized learning experiences according to each student's unique characteristics to maximize learning effectiveness and promote students' overall development.
[0003] In practical applications, personalized education often relies on advanced technical means such as learning analysis, artificial intelligence, and big data analysis. By tracking students' learning behavior data, interaction logs, and physiological data, etc., it dynamically adjusts learning tasks and goals, making the learning process more efficient and in line with individual characteristics. The ultimate goal is to cultivate students' self-directed learning ability, critical thinking, and problem-solving ability, while stimulating their learning interest and intrinsic motivation. With the continuous development of educational technology, personalized education is gradually breaking through the limitations of traditional education models and providing more accurate and diverse learning support for students.
[0004] However, in the implementation process of personalized education, the existing technology generally has problems such as learning path update lag, single recommendation strategy, and imperfect feedback mechanism, making it difficult to provide precise and dynamic personalized learning support. In addition, personalized education lacks structured feedback and self-reflection guidance, making it difficult to form a complete and effective closed-loop learning process. SUMMARY
[0005] In order to solve the problem that the existing technology generally has learning path update lag, single recommendation strategy, and weak feedback mechanism, making it difficult to achieve precise and dynamic personalized learning support, the present application provides a personalized learning monitoring system based on user behavior data.
[0006] A personalized learning monitoring system based on user behavior data, the personalized learning monitoring system comprising:
[0007] A data acquisition module for acquiring learning behavior data and physiological data of a target user and integrating interaction log data in a learning platform;
[0008] A learning goal setting module for setting a personalized learning goal system including short-term and long-term goals based on learning behavior data and interaction log data;
[0009] The self-assessment and feedback module guides the target user to conduct self-assessment regularly through regular online tests, and provides targeted feedback to the target user based on the assessment results, to help the target user understand the current learning progress;
[0010] The emotion monitoring module monitors the emotional state of the target user in real time through the interaction log data and the physiological data, and identifies the current emotional fluctuation of the target user;
[0011] The dynamic task adjustment module dynamically adjusts the difficulty, content and rhythm of the learning task based on the personalized learning target system, the current learning progress and the current emotional fluctuation, to generate real-time adaptive learning task flow.
[0012] Optionally, the learning target setting module comprises:
[0013] The learning behavior and interaction data analysis module is configured to analyze the learning behavior data and the interaction log data, and extract the comprehensive learning preference, learning habit, knowledge mastery and weak link of the target user;
[0014] The learning level and ability baseline modeling is configured to determine the learning level and ability baseline of the target user based on the knowledge mastery and weak link, in combination with the learning habit and the comprehensive learning preference;
[0015] The personalized learning target system generation module is configured to set personalized short-term learning targets and long-term learning targets according to the learning level and ability baseline, and keep the association between the short-term learning targets and the long-term learning targets, to form a complete personalized learning target system.
[0016] Optionally, the expression of the learning level and ability baseline of the target user is:
[0017] ;
[0018] ;
[0019] In the formula, 、 、 and respectively represent the weights of the knowledge mastery , the weak link , the learning habit and the comprehensive learning preference ; represents the knowledge mastery; represents the weak link; represents the learning habit; represents the comprehensive learning preference; 、 and respectively represent the weights of preference consistency A, preference diversity B and preference matching degree C; A represents preference consistency; B represents preference diversity; C represents preference matching degree; represents the learning level and ability baseline of the target user.
[0020] Optionally, through regular online tests, the target user is guided to regularly self-evaluate, and based on the evaluation results, the target user is provided with targeted feedback to assist the target user in understanding the current learning progress, including:
[0021] According to the learning goal and the stage plan of the target user, the target user is regularly pushed online tests related to the current learning stage;
[0022] After the online test is completed, according to the test score, the score is substituted into the learning attraction degree update formula as a feedback signal to adjust the learning attraction degree concentration of the relevant knowledge unit on the learning path, and an optimized learning path is obtained;
[0023] Based on the knowledge field, knowledge level, test score and optimized learning path of the target user, the similarity between the target user and each reference learner is calculated, each reference learner is sorted in descending order according to the similarity score, and a parameter learner meeting the preset requirements is selected to construct a reference learning group;
[0024] According to the current learning state of the target user, combined with the learning path and test score of the reference learners in the reference learning group, the recommendation probability of the learning unit is calculated by using the recommendation strategy formula, and personalized learning suggestions are generated for the target user;
[0025] According to the personalized learning suggestions, a structured feedback report is generated for the target user, and after the target user reads the feedback report, the target user is guided to self-reflect.
[0026] Optionally, the expression of the learning attraction degree update formula is:
[0027] ;
[0028] In the formula, represents the learning attraction degree concentration on the path at S+1 moment;
[0029] represents the volatility coefficient of the learning attraction degree;
[0030] represents the learning attraction degree concentration on the path at S moment;
[0031] represents the learning rate;
[0032] This represents the change in reward along the learning path from time S to time S+1.
[0033] Optionally, based on the target user's knowledge domain, knowledge level, test scores, and optimized learning path, the similarity between the target user and each reference learner is calculated. Each reference learner is then sorted in descending order according to their similarity score. Reference learners meeting preset requirements are selected to construct a reference learning group, including:
[0034] Extract feature data for the target user and each reference learner, including knowledge domain, knowledge level, test scores, and optimized learning path, and perform normalization processing.
[0035] Based on the normalized feature data, the differences in knowledge domain, knowledge level, test scores, and learning paths between the target user and each reference learner are calculated. The differences in each dimension are weighted and fused according to preset weights to calculate the total difference index.
[0036] The total difference is mapped to a similarity score to measure the degree of similarity between the target user and each reference learner;
[0037] All reference learners are sorted in descending order of similarity score, and those who meet the preset requirements are selected to construct reference learning groups.
[0038] Optionally, based on the target user's current learning status, combined with the learning paths and test scores of reference learners in the reference learning group, the recommendation probability of the learning unit is calculated using the recommendation strategy formula, and personalized learning suggestions are generated for the target user, including:
[0039] Identify the target user's current learning state as the adaptive state, and simultaneously obtain the most recently completed learning unit L from the target user. i Learning Unit L i As the starting point of the current learning path, find all available candidate learning units L. j ;
[0040] Calculate sequentially from the current learning unit L i Jump to each candidate learning unit L j heuristic value W ij ;
[0041] Initialize the learning unit L based on the learning behavior paths and test scores of users in the reference learning group. i To candidate learning unit L j Learning attraction concentration S ij ;
[0042] Based on heuristic value W ij and learning attraction concentration Sij , the recommendation probability U of the learning unit L is calculated by using the recommendation strategy formula i jumping to the learning unit L j ij ;
[0043] Select the learning unit with the highest recommendation probability value, and generate personalized learning suggestions for each recommendation result.
[0044] Optionally, the recommendation probability U of the learning unit L i jumping to the learning unit L j ij The expression of the recommendation probability U of the learning unit L is:
[0045] ;
[0046] In the formula, represents the recommendation summary from the current learning unit L i to the candidate learning unit L j ;
[0047] represents the heuristic value of jumping from the current learning unit L i to each candidate learning unit L j ;
[0048] represents the learning attraction concentration from the current learning unit L i to the candidate learning unit L j ;
[0049] represents the learning attraction concentration from the current learning unit L i to the candidate learning unit L k ;
[0050] represents the heuristic information factor;
[0051] represents the learning attraction factor;
[0052] represents the set of all jumpable candidate learning units L i from the current learning unit L k .
[0053] Optionally, based on the personalized learning goal system, the current learning progress and the current emotional fluctuation, the difficulty, content and rhythm of the learning task are dynamically adjusted to generate a real-time adaptive learning task flow, including:
[0054] Obtain the personalized learning goal system, the current learning progress and the current emotional fluctuation;
[0055] According to the personalized learning target system, the content theme, the ability requirement and the task unit of the personalized learning target system are screened;
[0056] According to the current learning progress and the current emotional fluctuation, the knowledge mastery level, the cognitive load and the psychological fluctuation of the target user are judged, and the current bearable strength and understanding ability of the target user to the learning task are speculated;
[0057] According to the speculated result, the complexity, the content granularity and the allocation time length of the task unit are dynamically adjusted, and the learning content most suitable for the current state is located;
[0058] The learning content is arranged into a task path according to the cognitive rhythm and the logical order, a learning task flow adapted to the current state is formed, and the learning task flow is pushed to the learner.
[0059] In summary, the present application includes at least one of the following beneficial technical effects:
[0060] (1) The present application integrates learning behavior data, physiological feedback, interaction logs and stage evaluation to build a complete closed-loop process from data collection, target setting, emotion perception to task pushing. The cognitive and emotional state of the learner is dynamically perceived, and the learning content and path are adjusted accordingly, realizing highly intelligent and personalized learning support. The user does not need to actively manage the learning rhythm, and the state change is automatically recognized, the matching task is pushed, and the cognitive burden and management cost are greatly reduced.
[0061] (2) The present application guides the user to actively perform periodic self-evaluation through regular online testing, and dynamically optimizes the learning path with the test results as the feedback signal. The learning attraction degree updating mechanism is used to adjust the weight concentration of the knowledge unit, so as to reflect the user's mastery change of each knowledge point, generate more accurate learning path suggestions, and also introduce similar learner modeling, combine the knowledge field, ability level and test data of the target user, select a high-similarity reference learner group, and calculate the jump probability of the learning task through the recommendation strategy formula, select the optimal task unit, and generate personalized learning suggestion path for the user. BRIEF DESCRIPTION OF DRAWINGS
[0062] Figure 1 is the principle block diagram of the system in the present application. DETAILED DESCRIPTION
[0063] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings.
[0064] In the description of the present specification, the description referring to the terms "certain embodiments", "one embodiment", "some embodiments", "illustrative embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in connection with the described embodiments or examples are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.
[0065] The embodiments of the present application disclose a personalized learning monitoring system based on user behavior data, referring to Figure 1 The personalized learning monitoring system comprises:
[0066] A data collection module is configured to collect learning behavior data and physiological data of a target user, and integrate interactive log data in a learning platform.
[0067] It should be explained that the learning behavior data includes operation trajectory, learning path, task completion time, answering process, error rate, click behavior, page dwell time, note and discussion behavior of the target user in the platform, and is used to analyze learning habits, learning strategies, mastery and cognitive load.
[0068] The physiological data (wearable device / sensor) such as electroencephalogram (EEG), heart rate, skin electricity, eye movement trajectory, facial expression, body movement amplitude, etc. are used to assist in inferring the emotional state, concentration level, stress index and fatigue degree of the target user.
[0069] The interactive log data in the learning platform includes the interaction record between the learner and the learning terminal, such as interface click stream, video playback behavior, exercise feedback, message reply record, voice interaction content, etc.
[0070] A learning goal setting module is configured to set a personalized learning goal system including short-term and long-term goals based on the learning behavior data and the interactive log data.
[0071] Preferably, the learning goal setting module comprises:
[0072] A learning behavior and interactive data analysis module is configured to analyze the learning behavior data and the interactive log data, and extract the comprehensive learning preference, learning habits, knowledge mastery and weak links of the target user.
[0073] It should be explained that the learning behavior data and the interactive log data collected in the learning platform are analyzed, so as to extract multiple dimension features of the user in the learning process, including comprehensive learning, learning habits, knowledge mastery and weak link identification.
[0074] Among them, the comprehensive learning preference: such as content preference (video / text / subject), learning time period preference, learning form (passive acceptance / active exploration) and the like;
[0075] Learning habits: such as task completion rhythm, learning amount per unit time, review frequency, skipping content behavior and the like;
[0076] Knowledge mastery: analyze the degree of mastery through accuracy, repeated learning frequency, test performance and the like;
[0077] Weak link identification: find out the knowledge points that are repeatedly wrong, long-term unmastered or frequently skipped.
[0078] Learning level and ability baseline modeling, based on knowledge mastery and weak links, combined with learning habits and comprehensive learning preferences, determine the learning level and ability baseline of the target user.
[0079] Preferably, the expression of the learning level and ability baseline of the target user is:
[0080] ;
[0081] ;
[0082] In the formula, , , and respectively represent the weights of knowledge mastery , weak links , learning habits and comprehensive learning preferences ; represents knowledge mastery; represents weak links; represents; represents comprehensive learning preferences; , and respectively represent the weights of preference consistency A, preference diversity B and preference matching degree C; A represents preference consistency; B represents preference diversity; C represents preference matching degree; represents the learning level and ability baseline of the target user.
[0083] The personalized learning target system generation module formulates personalized short-term learning targets and long-term learning targets according to the learning level and ability baseline, and maintains the association between the short-term learning targets and the long-term learning targets, forming a complete personalized learning target system.
[0084] It needs to be explained that after the learning level and ability baseline is established, individualized short-term learning goals and long-term learning goals are set;
[0085] Short-term learning goals mainly target specific tasks at the current stage (such as mastering a certain knowledge point or improving a certain skill), and closely match the ability baseline to avoid being too difficult or too easy;
[0086] Long-term learning goals mainly face longer growth paths (such as completing a certain course or passing a certain certification);
[0087] Short-term goals should serve long-term goals, and the achievement of each short-term goal should be able to quantitatively promote the progress of long-term goals, establish a capacity development trajectory through the hierarchical relationship between goals, and support stage evaluation and dynamic adjustment.
[0088] Self-assessment and feedback module, through regular online testing, guides target users to regularly self-assess, and provides targeted feedback to target users based on evaluation results, to help target users understand current learning progress.
[0089] Specific examples are as follows:
[0090] The target user is a university mathematics learner, aiming to improve mathematics performance in 6 months, focusing on breaking through function and geometry mathematics;
[0091] (1) Learning behavior data:
[0092] Learn mathematics for 1.5 hours every day, prefer text and image materials (60%) + interactive question bank (40%); Error rate is high (65%) on function image transformation questions, but performs well on algebraic operations (correct rate 85%).
[0093] (2) Interaction log data:
[0094] Frequently consults function-related analysis, but skips geometry proof questions; Average problem solving time is long (especially for geometry questions), but algebra questions are faster.
[0095] (3) Learning habits:
[0096] Habit of learning in the morning, focus for about 40 minutes; Review frequency is low, and tends to directly do new questions rather than review wrong questions.
[0097] The analysis results are as follows:
[0098] (1) Comprehensive learning preference ( ): Prefer structured knowledge (text and image) and immediate feedback (question bank), but have low acceptance of abstract content (such as geometry proof);
[0099] (2) Knowledge mastery ( ): algebraic operation (0.85), function basis (0.5), geometric proof (0.3);
[0100] (3) Weak links (weak links): ): function image transformation (0.65), geometric proof (0.7);
[0101] (4) Learning habits (weak links): ): medium concentration (0.6), weak review behavior (0.2).
[0102] Substitute the above content into the expression of the learning level and ability baseline of the target user, and set , , and are 0.4, 0.3, 0.2 and 0.1 respectively. The learning level and ability baseline of the target user is calculated to be about 0.46, which belongs to the lower middle, and the weak links of functions and geometry should be focused on and the learning habits should be optimized.
[0103] Preferably, by regular online testing, the target user is guided to regularly self-evaluate, and based on the evaluation results, the target user is provided with targeted feedback to assist the target user to understand the specific implementation steps of the current learning progress as follows:
[0104] According to the learning goal and the stage plan of the target user, the target user is regularly pushed online tests related to the current learning stage.
[0105] After the online test is completed, according to the test results, the score is substituted into the learning attraction degree update formula as a feedback signal, and the learning attraction degree concentration of the relevant knowledge unit on the learning path is adjusted to obtain the optimized learning path.
[0106] Preferably, the expression of the learning attraction degree update formula is as follows:
[0107] ;
[0108] In the formula, represents the learning attraction degree concentration on the path at S+1 moment;
[0109] represents the volatility coefficient of the learning attraction degree;
[0110] represents the learning attraction degree concentration on the path at S moment;
[0111] represents the learning rate;
[0112] represents the reward change of the learning path from S moment to S+1 moment.
[0113] It needs to be explained that according to the learning goal and the stage plan of the target user, the online test corresponding to the current stage is regularly pushed to evaluate the learning effect; after the test is completed, the score is taken as a quantitative feedback signal, which is substituted into the learning attraction degree update formula for calculation, and the old path weight is gradually attenuated through the learning attraction degree volatility factor control, and at the same time the learning attraction degree concentration of the related knowledge unit is enhanced or weakened combined with the test reward signal, so as to dynamically adjust the priority and recommendation direction of the learning path, and the learning path can be continuously self-optimized according to the real learning performance of the user, which can not only avoid excessive dependence on mastered content, but also highlight weak links, and realize personalized guidance and continuous adaptation of the learning process.
[0114] Based on the knowledge field, knowledge level, test score and optimized learning path of the target user, the similarity between the target user and each reference learner is calculated, and each reference learner is sorted in descending order according to the similarity score value, and the parameter learner meeting the preset requirements is selected to construct the reference learning group.
[0115] Preferably, based on the knowledge field, knowledge level, test score and optimized learning path of the target user, the similarity between the target user and each reference learner is calculated, and each reference learner is sorted in descending order according to the similarity score value, and the parameter learner meeting the preset requirements is selected to construct the reference learning group. The specific implementation steps are as follows:
[0116] The feature data of the target user and each reference learner including the knowledge field, the knowledge level, the test score and the optimized learning path are extracted and normalized;
[0117] Based on the normalized feature data, the knowledge field difference degree, the knowledge level difference degree, the test score difference degree and the learning path difference degree between the target user and each reference learner are calculated, and each dimension difference degree is weighted and fused according to the preset weight to calculate the total difference degree index;
[0118] The total difference degree is mapped into a similarity score value to measure the similarity between the target user and each reference learner;
[0119] All reference learners are sorted in descending order according to the similarity score value, and the parameter learner meeting the preset requirements is selected to construct the reference learning group.
[0120] It needs to be explained that by extracting the characteristic data of the target user and other reference learners in multiple dimensions such as knowledge field, knowledge level, test score and optimized learning path, and performing unified normalization processing, the difference between each dimension is calculated and weighted and fused into a total difference index; then it is mapped into a similarity score, all learners are sorted according to the similarity, and finally a group of reference learners most similar to the target user is selected to construct a personalized reference learning group, which provides support for path recommendation, strategy transfer and learning resource adaptation.
[0121] According to the current learning state of the target user, the learning path and test score of the reference learners in the reference learning group are combined, the recommendation probability of the learning unit is calculated by using the recommendation strategy formula, and personalized learning suggestions are generated for the target user.
[0122] Preferably, according to the current learning state of the target user, the learning path and test score of the reference learners in the reference learning group are combined, the recommendation probability of the learning unit is calculated by using the recommendation strategy formula, and personalized learning suggestions are generated for the target user. The specific implementation steps are as follows:
[0123] The current learning state of the target user is identified as an adaptive state, and the last learning unit L i completed by the target user is obtained. i As the starting point of the current learning path, all candidate learning units L j available for selection are found.
[0124] The heuristic value W i of each candidate learning unit L j jumping from the current learning unit L ij is calculated in turn.
[0125] The learning attraction concentration S i from the current learning unit L j to the candidate learning unit L ij is initialized according to the learning behavior path and test score of the users in the reference learning group.
[0126] Based on the heuristic value W ij and the learning attraction concentration S ij , the recommendation probability U i of the learning unit L j jumping to the learning unit L ij is calculated by using the recommendation strategy formula.
[0127] The learning unit with the highest recommendation probability value is selected, and personalized learning suggestions are generated for each recommendation result.
[0128] Preferably, the recommendation probability of the learning unit L iJump to learning unit L j Recommendation probability U ij The expression is:
[0129] ;
[0130] In the formula, Indicates starting from the current learning unit L i To candidate learning unit L j Recommendation summary;
[0131] Indicates starting from the current learning unit L i Jump to each candidate learning unit L j heuristic value;
[0132] Indicates starting from the current learning unit L i To candidate learning unit L j The concentration of learning attraction;
[0133] Indicates starting from the current learning unit L i To candidate learning unit L k The concentration of learning attraction;
[0134] Indicates heuristic information factor;
[0135] This represents the learning attraction factor;
[0136] Indicates starting from the current learning unit L i Starting from the point, navigate to all available candidate learning units L. k A set of.
[0137] Based on personalized learning suggestions, structured feedback reports are generated for target users, and after the target users read the feedback reports, they are guided to conduct self-reflection.
[0138] It should be explained that by analyzing the target user's current learning behavior and test scores, their learning state is identified, which is usually divided into adaptive state, transitional state, and consolidation state. Here, the target user is in the adaptive state, which means that the user is learning new or relatively unfamiliar knowledge points, and a suitable learning path needs to be recommended based on the performance of the reference learning group.
[0139] Extract the most recently completed learning unit (such as a chapter, module, or knowledge point) from the target user's learning record. i This serves as a recommended starting point; by analyzing the knowledge graph or course structure, it automatically lists resources related to the current learning unit L. i Related candidate learning units Lj These units can be with L j At the same level, or subsequent knowledge points, or review units;
[0140] For each candidate learning unit L j , the heuristic value W ij is calculated by heuristic rules (such as knowledge difficulty, learning duration, historical learner performance, etc.); ij Generally, the relevance of the candidate learning unit L j to the current learning goal, the content difficulty, and the matching degree of the learner's current ability are comprehensively calculated);
[0141] The heuristic value represents the expected effectiveness from the learning unit L i to L j , usually related to factors such as similarity of learning content and difficulty adaptation; combined with user data (learning path and test scores) in the reference learning group, the learning attraction concentration S ij is initialized;
[0142] The size of the learning attraction concentration reflects the historical success degree of this path among similar learners; if many reference learners perform well on similar learning paths, the learning attraction concentration of this path is higher;
[0143] The recommendation probability formula combines both, adjusts the weights m and n for flexible adjustment, and provides personalized recommendations according to different user needs and learning stages;
[0144] Once the recommended path is selected, personalized learning suggestions are provided for the target user:
[0145] Specifically, the next learning unit is recommended; learning methods (such as videos, practice questions, reading materials, etc.) are recommended; according to the learning progress, suitable learning time and rhythm are recommended.
[0146] The target user background is as follows:
[0147] Identity: university mathematics learner;
[0148] Learning goal: improve mathematics performance within 6 months, focus on breaking through "function" and "geometry" modules;
[0149] The ability baseline score is about 0.46, at a slightly below average level;
[0150] Weaknesses: function image transformation (error rate 65%), geometry proof (error rate 70%);
[0151] Habit preference: text + question bank, morning focus, weak review behavior;
[0152] The online test is designed as follows:
[0153] Test frequency: bi-weekly test (12 times in total); the 4th week, 8th week, and 12th week are set as key test nodes, and the learning path is updated synchronously;
[0154] First test task (2nd week):
[0155] Test content: 4 questions on function image transformation, 2 questions on function property application, 2 questions on simple geometric reasoning, and 2 questions on algebraic basis detection;
[0156] Test platform: text and image questions with interactive answer boxes; instant feedback after submission + answer analysis allows for one-time pause during the test;
[0157] Example test scores (hypothetical results) are shown in the following table:
[0158] Table 1: Test Score Table
[0159] Knowledge points Score rate Error type explanation Function image transformation 25% Coordinate transformation understanding deviation, unable to draw image Function property application 50% Will suit formula but will not reverse image Simple geometric reasoning 0% Directly skip, no answer Algebraic basis detection 90% High accuracy, fast calculation
[0160] Use the score as the feedback signal ΔR, and substitute the test results into the learning attraction degree update formula:
[0161] For the "image transformation" path L 图 : ΔR = -0.4 → learning attraction degree decreases → prefer to add basic exercise units in subsequent recommendations;
[0162] For the "geometric proof" path L 几 : ΔR = -0.7 → learning attraction degree evaporates quickly → reduce the recommendation of deep geometric content;
[0163] For the "algebra" path L 代 : ΔR = +0.8 → learning attraction degree increases → recommend more challenging algebraic content;
[0164] According to the test results and the updated learning attraction degree concentration, use the recommendation formula, combine the knowledge graph and historical learning data, calculate the recommendation probability, and output the recommended path.
[0165] Table 2: Recommended Path Table
[0166] Candidate learning unit Heuristic value W ij ]] Learning attractiveness concentration S ij ]]> Recommended probability U ij ]] Suggestion explanation Image transformation basic review 0.7 0.5 0.34 Start with "translation" and "symmetry" from image examples Function symmetry and parity 0.6 0.4 0.25 Text and image explanation + quick detection question Algebraic expression transformation advanced 0.8 0.7 0.41 Add comprehensive solution questions, challenge medium difficulty
[0167] According to the above learning suggestions, generate a structured feedback report for the target user, and after the target user reads the feedback report, guide the target user to reflect on themselves.
[0168] The emotion monitoring module monitors the target user's emotional state in real time through interactive log data and physiological data, and identifies the target user's current emotional fluctuations.
[0169] It needs to be explained that by combining interaction log data and physiological data, the emotional state of the target user is tracked in real time, and the emotional fluctuation is identified; Specifically, the interaction log data can provide information about the user's behavior in the interactive process, such as learning duration, frequently queried knowledge points, task completion time, etc., so as to infer the user's emotional tendency (for example, anxiety, confusion or joy); Physiological data includes heart rate, skin resistance, eye movement and other indicators. These physiological signals can directly reflect the physiological state and emotional response of the user; By fusing and analyzing these two kinds of data, the emotional fluctuation is identified, which helps to adjust the learning suggestion or provide emotional regulation support in time.
[0170] The dynamic task adjustment module dynamically adjusts the difficulty, content and rhythm of the learning task based on the personalized learning target system, the current learning progress and the current emotional fluctuation, and generates a real-time adaptive learning task flow.
[0171] Preferably, dynamically adjusting the difficulty, content and rhythm of the learning task based on the personalized learning target system, the current learning progress and the current emotional fluctuation, and generating a real-time adaptive learning task flow includes:
[0172] Obtain the personalized learning target system, the current learning progress and the current emotional fluctuation;
[0173] According to the personalized learning target system, the content theme, the ability requirement and the task unit of the personalized learning target body are screened;
[0174] According to the current learning progress and the current emotional fluctuation, the knowledge mastery level, the cognitive load and the psychological fluctuation of the target user are judged, and the current learning task carrying strength and understanding ability of the target user are inferred;
[0175] According to the inference result, the complexity, content granularity and allocation time of the task unit are dynamically adjusted, and the learning content most suitable for the current state is located;
[0176] The learning content is arranged into a task path according to the cognitive rhythm and logical order, a learning task flow suitable for the current state is formed, and the learning task flow is pushed to the learner.
[0177] It needs to be explained that three kinds of core input information are obtained
[0178] (1) Personalized learning target system: including the overall goal of the learner (such as "mastering function and geometry in six months"), ability dimension (logical reasoning, spatial perception, etc.) and priority improvement content;
[0179] (2) Current learning progress: judge the knowledge mastery degree and progress completion condition through learning record, test result, etc.;
[0180] (3) Current emotional fluctuation: in combination with the results of the emotion monitoring module, the current emotional state of the learner (such as anxiety, fatigue, calmness, concentration, etc.) is identified.
[0181] According to the theme dimension and ability requirement in the learning goal, the suitable task unit (for example: image transformation, geometric construction, etc.) is matched in the knowledge graph or course library;
[0182] Determine the learner's current: knowledge mastery, cognitive load and psychological fluctuation;
[0183] Whether the level of knowledge mastery is in the understanding, application or transfer stage; whether the task is too complex or repetitive in cognitive load, causing psychological burden; whether the emotion is stable, whether it is appropriate to enter new content, or whether it needs to relax and practice to relieve stress in psychological fluctuation.
[0184] According to the above analysis, adjust the parameters of the task unit (complexity, content granularity, allocation time length and feedback mechanism):
[0185] Among them, the complexity is to simplify or deepen the difficulty of the question; the content granularity is to choose micro task or comprehensive exercise; the allocation time length is to adjust the learning rhythm according to the concentration; the feedback mechanism is to adapt to the immediate prompt or delayed evaluation;
[0186] Arrange the screened learning tasks according to the cognitive order and logical rhythm (such as basic questions first and then advanced questions), and "difficulty interlacing" according to the emotional fluctuation (such as arranging light exercises in a state of fatigue); generate learning task flow adapted to the current state, and push to the learning interface.
[0187] It should be noted that the calculation formula and each parameter participating in the operation in the present application are pre-processed by dimensionless processing, and the process of dimensionless processing is known in the industry, which is not described here.
[0188] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application, and those skilled in the art can make changes, modifications, replacements and modifications to the above embodiments within the scope of the present application.
Claims
1. A personalized learning monitoring system based on user behavior data, characterized in that, This personalized learning monitoring system includes: The data acquisition module is used to collect learning behavior data and physiological data of target users, and integrate the interaction log data in the learning platform; The learning goal setting module, based on the learning behavior data and the interaction log data, sets up a personalized learning goal system that includes short-term and long-term goals; The self-assessment and feedback module guides target users to conduct regular self-assessments through periodic online tests and provides targeted feedback based on the assessment results to help target users understand their current learning progress. The emotion monitoring module monitors the target user's emotional state in real time through the interaction log data and the physiological data, and identifies the target user's current emotional fluctuations. The dynamic task adjustment module dynamically adjusts the difficulty, content, and pace of learning tasks based on the personalized learning goal system, the current learning progress, and the current emotional fluctuations, generating a real-time adapted learning task flow. The method involves guiding target users to conduct regular self-assessments through periodic online tests, and providing targeted feedback based on the assessment results to help target users understand their current learning progress. Based on the learning goals and phased plans of the target users, regularly push online tests related to the current learning stage to the target users; After the online test is completed, the test results are used as feedback signals and substituted into the learning attraction update formula to adjust the learning attraction concentration of relevant knowledge units on the learning path, thereby obtaining the optimized learning path. Based on the target user's knowledge domain, knowledge level, test scores, and optimized learning path, the similarity between the target user and each reference learner is calculated. Each reference learner is sorted in descending order according to the similarity score, and reference learners that meet the preset requirements are selected to construct a reference learning group. Based on the target user's current learning status, combined with the learning paths and test scores of reference learners in the reference learning group, the recommendation probability of the learning unit is calculated using the recommendation strategy formula, and personalized learning suggestions are generated for the target user. Based on personalized learning suggestions, generate structured feedback reports for target users, and guide them to reflect on their own learning after reading the reports. The expression for the learning attraction update formula is: ； In the formula, express Learning attraction concentration along the path at time +1; The volatility coefficient, representing the attractiveness of learning; express The concentration of learning attraction along the time path; Indicates the learning rate; Indicates from Time's up The reward change of the learning path at time +1; Calculate the recommendation probability U using the recommendation strategy formula ij The expression is: ； In the formula, Indicates starting from the current learning unit L i To candidate learning unit L j The probability of recommendation; Indicates starting from the current learning unit L i Jump to each candidate learning unit L j heuristic value; Indicates starting from the current learning unit L i To candidate learning unit L j The concentration of learning attraction; Indicates starting from the current learning unit L i To candidate learning unit L k The concentration of learning attraction; Indicates heuristic information factor; This represents the learning attraction factor; Indicates starting from the current learning unit L i Starting from the point, navigate to all available candidate learning units L. k A set of.
2. The personalized learning monitoring system based on user behavior data according to claim 1, characterized in that, The learning objective setting module includes: The learning behavior and interaction data analysis module is used to analyze learning behavior data and interaction log data to extract the target user's comprehensive learning preferences, learning habits, knowledge mastery and weaknesses. Learning level and ability baseline modeling, based on knowledge mastery and weaknesses, combined with learning habits and comprehensive learning preferences, determines the learning level and ability baseline of the target users; The personalized learning goal system generation module formulates personalized short-term and long-term learning goals based on the learning level and ability baseline, and maintains the correlation between short-term and long-term learning goals to form a complete personalized learning goal system.
3. The personalized learning monitoring system based on user behavior data according to claim 2, characterized in that, The expression for the learning level and ability baseline of the target user is: ； ； In the formula, , , and They represent knowledge mastery respectively. Weak links Study habits and comprehensive learning preferences weight , and These represent the weights of preference consistency (A), preference diversity (C), and preference matching degree (E), respectively. This represents the baseline of the target user's learning level and ability.
4. The personalized learning monitoring system based on user behavior data according to claim 1, characterized in that, The process involves calculating the similarity between the target user and each reference learner based on the target user's knowledge domain, knowledge level, test scores, and optimized learning path. Each reference learner is then sorted in descending order according to their similarity score. Reference learners meeting preset requirements are selected to construct a reference learning group, which includes: Extract feature data for the target user and each reference learner, including knowledge domain, knowledge level, test scores, and optimized learning path, and perform normalization processing. Based on the normalized feature data, the differences in knowledge domain, knowledge level, test scores, and learning paths between the target user and each reference learner are calculated. The differences in each dimension are weighted and fused according to preset weights to calculate the total difference index. The total difference is mapped to a similarity score to measure the degree of similarity between the target user and each reference learner; All reference learners are sorted in descending order of similarity score, and reference learners who meet the preset requirements are selected to construct reference learning groups.
5. The personalized learning monitoring system based on user behavior data according to claim 1, characterized in that, The step of calculating the recommendation probability of learning units based on the target user's current learning status, combined with the learning paths and test scores of reference learners in the reference learning group, and generating personalized learning suggestions for the target user using the recommendation strategy formula includes: Identify the target user's current learning state as the adaptive state, and simultaneously obtain the most recently completed learning unit L from the target user. i Learning Unit L i As the starting point of the current learning path, find all available candidate learning units L. j ; Calculate sequentially from the current learning unit L i Jump to each candidate learning unit L j heuristic value W ij ; Initialize the learning unit L based on the learning paths and test scores of users in the reference learning group. i To candidate learning unit L j Learning attraction concentration S ij ; Based on heuristic W ij and learning attraction concentration S ij The learning unit L is calculated using the recommendation strategy formula. i Jump to learning unit L j Recommendation probability U ij ; The learning unit with the highest recommendation probability is selected, and personalized learning suggestions are generated for each recommendation result.
6. The personalized learning monitoring system based on user behavior data according to claim 1, characterized in that, The process of dynamically adjusting the difficulty, content, and pace of learning tasks based on the personalized learning goal system, the current learning progress, and the current emotional fluctuations to generate a real-time adapted learning task flow includes: Obtain a personalized learning goal system, current learning progress, and current emotional fluctuations; Based on the personalized learning objectives system, select task units whose content themes and ability requirements conform to the personalized learning objectives system; Based on the current learning progress and current emotional fluctuations, assess the target user's knowledge level, cognitive load, and psychological fluctuations, and infer the target user's current capacity to handle and comprehension of the learning task. Based on the predicted results, the complexity, content granularity, and time allocation of task units are dynamically adjusted to identify the learning content most suitable for the current state. The learning content is arranged into task paths according to cognitive rhythm and logical order, forming a learning task flow that adapts to the current state, and then pushed to the learner.
Citation Information
Patent Citations
Intelligent adaptive educational training system based on big data
CN120852104A