An Adaptive AI Learning Path Planning System Based on Learning Profiling

Abstract

This invention relates to the field of learning path planning technology, and provides an adaptive AI learning path planning system based on learning profiles. The system includes a multi-source learning data acquisition module, a data preprocessing module, a three-dimensional cognitive diagnosis module, a multi-dimensional dynamic learning profile construction module, a subject knowledge graph module, an adaptive learning path planning module, and a dynamic path adjustment module. The multi-source learning data acquisition module collects multi-source learning data from students, including basic information data, historical learning data, initial assessment data, and real-time interaction data during the learning process. By integrating dynamic memory states based on the Ebbinghaus forgetting curve, the system can track students' memory retention in real time, ensuring the rationality of the learning sequence, avoiding knowledge gaps, improving learning effectiveness, achieving multi-objective optimization, improving the rationality of path planning, and enhancing the accuracy of cognitive diagnosis, thus providing a more accurate foundation for learning profiles and path planning.

Description

Technical Field

[0001] This invention relates to the field of learning path planning technology, specifically to an adaptive AI learning path planning system based on learning profiles. Background Technology

[0002] With the rapid development of online education and artificial intelligence technologies, personalized adaptive learning has become a research hotspot in the field of education. Its core lies in providing students with personalized learning paths and resources based on their individual differences, thus achieving individualized instruction. Existing adaptive learning systems have made some progress; for example, some systems recommend learning content to students by building learner profiles and combining them with machine learning algorithms, or adjust learning paths through reinforcement learning.

[0003] However, existing technologies still have many shortcomings. First, the learning profile is too narrow in scope and lacks integration of dynamic memory states: most existing learning profiles only focus on students' knowledge mastery and learning behavior data, without considering the forgetting patterns of students' memory, and cannot track students' memory retention of learned knowledge points in real time. This leads to an inability to reasonably arrange review nodes, making it easy for students to forget what they have learned, resulting in poor knowledge retention. Second, path planning does not fully integrate the topological constraints of the knowledge graph: most existing path planning algorithms only focus on the mastery of knowledge points, without fully considering the prerequisite dependencies between knowledge points. This easily leads to unreasonable learning order problems. For example, students may start learning subsequent complex knowledge points before mastering the prerequisite knowledge points, resulting in knowledge gaps and poor learning outcomes. Third, the dynamic adjustment capability is insufficient: most existing systems are static path planning systems, or can only perform simple path changes. They cannot replan the entire path based on real-time changes in learning progress, resulting in insufficient adaptability and an inability to cope with changes in students' learning states. Finally, the accuracy of cognitive diagnosis is insufficient: most existing cognitive diagnosis models are single-dimensional assessments, which cannot simultaneously and accurately assess students' mastery of knowledge points, proficiency in problem-solving strategies, and tendency of error causes, resulting in insufficient accuracy of learning profiles and thus affecting the accuracy of path planning. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an adaptive AI learning path planning system based on learning profiles. This system solves the problems of being unable to track students' memory retention of learned knowledge points in real time, which can easily lead to unreasonable learning order. It also addresses the issues of insufficient adaptability, inability to replan the entire learning path based on real-time changes in learning progress, and inability to simultaneously and accurately assess students' mastery of knowledge points, proficiency in problem-solving strategies, and tendency to make mistakes.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solution: an adaptive AI learning path planning system based on learning profiles, including a multi-source learning data acquisition module, a data preprocessing module, a three-dimensional cognitive diagnosis module, a multi-dimensional dynamic learning profile construction module, a subject knowledge graph module, an adaptive learning path planning module, and a dynamic path adjustment module; The multi-source learning data acquisition module is used to collect students' multi-source learning data, which includes basic information data, historical learning data, initial assessment data, and real-time interaction data during the learning process. The data preprocessing module is used to preprocess the collected multi-source learning data to obtain a standard feature dataset. The three-dimensional cognitive diagnosis module is used to evaluate students' cognitive status using an improved IRT-LSTM-BERT three-dimensional cognitive diagnosis model and output three types of evaluation results: knowledge point mastery, problem-solving strategy proficiency, and error tendency.

[0006] Preferably, the multi-dimensional dynamic learning profile construction module is used to construct a multi-dimensional dynamic learning profile based on cognitive diagnosis results and learning data. The learning profile includes knowledge mastery status, learning behavior characteristics, learning preference characteristics, and dynamic memory status. The dynamic memory status is updated in real time based on the Ebbinghaus forgetting curve to update the memory retention rate of learned knowledge points.

[0007] Preferably, the subject knowledge graph module is used to store a pre-built subject knowledge graph, which includes knowledge point nodes, prerequisite dependencies between nodes, knowledge point difficulty and association. The adaptive learning path planning module is used to plan learning paths based on a multi-dimensional dynamic learning profile and the subject knowledge graph, using an improved multi-objective locust optimization algorithm. During the planning process, it takes into account the reinforcement of weak knowledge points, prerequisite dependency constraints, review node arrangement and learning efficiency optimization, and matches learning resources according to learning preferences.

[0008] Preferably, the dynamic path adjustment module is used to update the learning profile in real time during the learning process. When the deviation between the learning status and the expected status is detected to exceed a preset threshold, path replanning is triggered to achieve closed-loop dynamic adjustment.

[0009] Preferably, the data preprocessing module uses an outlier detection algorithm based on statistical distribution to remove outliers from the original data, standardizes features of different dimensions, converts feature values ​​to the range of 0-1, and extracts features from the original data to obtain answer features, behavioral features, and preference features.

[0010] Preferably, the three-dimensional cognitive diagnosis module includes an IRT (Item Response Theory) module, an LSTM (Laser-Based Scripting) temporal module, and a BERT (Bibliographic Graph) semantic module. The IRT module is used to assess the mastery of knowledge points based on answer data. The LSTM temporal module is used to assess the proficiency of problem-solving strategies based on answer time-series data. The BERT semantic module is used to attribute errors to causes based on answer text data and obtain error tendency. The three-dimensional cognitive diagnosis module achieves parameter sharing through multi-task joint training and outputs three types of evaluation results simultaneously.

[0011] Preferably, the update formula for the dynamic memory state is: Among them, M k (t) represents the amount of memory retained at time t, M O t represents the initial amount of information to be memorized, t represents the time since the last learning session, and s represents the memory intensity parameter adjusted based on the individual student's memory parameters, dynamically adjusted according to the student's historical review performance.

[0012] Preferably, the optimization objectives of the improved multi-objective locust optimization algorithm include: minimizing the learning cost of weak knowledge points, satisfying pre-dependency constraints, maximizing memory retention rate, and minimizing total learning time.

[0013] Preferably, the dynamic path adjustment module is used to update the learning profile after each knowledge point is learned, calculate the deviation between the current cognitive state and the expected state, and trigger path replanning and regenerate the learning path when the deviation exceeds a preset threshold.

[0014] An adaptive AI learning path planning method based on learning profiles includes the following steps: S1. System initialization: Construct a knowledge graph for the target subject and input the prerequisites, difficulties, and relationships of knowledge points; S2. Collect students' initial learning data, including basic information, historical learning data and initial assessment data; S3. Preprocess the initial learning data to obtain a standard feature dataset; S4. Using the improved IRT-LSTM-BERT three-dimensional cognitive diagnostic model, the initial cognitive state of students is diagnosed to obtain the degree of mastery of knowledge points, proficiency in problem-solving strategies, and tendency of error causes. S5. Construct an initial multi-dimensional dynamic learning profile, integrating cognitive state, learning behavior, learning preferences, and initial memory state; S6. Based on the initial learning profile and knowledge graph, an improved multi-objective locust optimization algorithm is used to plan the initial learning path and match personalized learning resources. S7. Push learning paths and resources to students and collect real-time interactive data during the learning process; S8. Dynamically update the learning profile based on real-time interactive data; S9. Determine whether the learning objective has been achieved. If yes, end the process and output the learning report; otherwise, proceed to the next step. S10. Determine whether the deviation between the learning state and the expected state exceeds the threshold. If yes, return to step S6 to replan the path; otherwise, return to step S7 to continue learning.

[0015] This invention provides an adaptive AI learning path planning system based on learning profiles. It has the following beneficial effects: 1. This invention integrates dynamic memory status based on the Ebbinghaus forgetting curve, enabling real-time tracking of students' memory retention and automatically inserting review nodes in path planning, effectively improving knowledge retention and solving the problem of single-dimensional learning profiles in existing technologies.

[0016] 2. This invention combines the topological dependencies of subject knowledge graphs and adds pre-dependency constraints to path planning, ensuring the rationality of the learning sequence, avoiding the problem of knowledge gaps, and improving the effectiveness of learning.

[0017] 3. This invention uses an improved multi-objective locust optimization algorithm for path planning, which can simultaneously take into account multiple objectives such as strengthening weak knowledge points, learning efficiency, and review arrangements, thereby achieving multi-objective optimization and improving the rationality of path planning.

[0018] 4. This invention adopts an improved IRT-LSTM-BERT three-dimensional cognitive diagnostic model, which can simultaneously and accurately assess the mastery of knowledge points, proficiency in problem-solving strategies, and tendency of error causes, thereby improving the accuracy of cognitive diagnosis and providing a more accurate foundation for learning profiles and path planning. Attached Figure Description

[0019] Figure 1 This is a flowchart of the present invention; Figure 2 This is a schematic diagram of the system structure of the present invention; Figure 3 This is a schematic diagram of the structural block of the three-dimensional cognitive diagnostic model of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example: Reference Figures 1-3 This invention provides an adaptive AI learning path planning system based on student learning profiles, comprising a multi-source learning data acquisition module, a data preprocessing module, a three-dimensional cognitive diagnosis module, a multi-dimensional dynamic learning profile construction module, a subject knowledge graph module, an adaptive learning path planning module, and a dynamic path adjustment module. The multi-source learning data acquisition module collects multi-source learning data from students, including basic information such as age, grade, and subject goals; historical learning data such as historical learning records, answer records, and resource access records; initial assessment data (student entrance assessment answer data for initial cognitive diagnosis); and real-time interaction data during the learning process, such as page dwell time, answer results, learning duration, and resource interaction records. The data preprocessing module preprocesses the collected multi-source learning data to obtain a standard feature dataset. The data preprocessing module uses a statistical distribution-based outlier detection algorithm to remove outliers from the original data, standardizes features of different dimensions, converts feature values ​​to the range of 0-1, and extracts features from the original data to obtain answer features, behavioral features, and preference features.

[0022] The three-dimensional cognitive diagnosis module uses an improved IRT-LSTM-BERT three-dimensional cognitive diagnosis model to assess students' cognitive status and output three types of assessment results: knowledge point mastery, problem-solving strategy proficiency, and error tendency. The three-dimensional cognitive diagnosis module includes an IRT item response theory module, an LSTM temporal module, and a BERT semantic module. The IRT item response theory module assesses students' knowledge point mastery by modeling the relationship between students' abilities and question difficulty using students' answer data, accurately quantifying the probability of mastering each knowledge point. The LSTM temporal module processes students' answer temporal data to assess their problem-solving strategy proficiency by capturing the characteristics of students' problem-solving process through temporal data, quantifying the degree of students' mastery of problem-solving strategies. The BERT semantic module processes students' answer text and error text, performs error attribution, identifies the causes of students' errors (e.g., unclear concepts, calculation errors, or misreading the question), and assesses error tendency. Through multi-task joint training, the three modules share parameters and simultaneously output three dimensions of diagnostic results, improving accuracy compared to existing single-dimensional diagnostic models.

[0023] The multi-dimensional dynamic learning profile construction module is used to build a multi-dimensional dynamic learning profile based on cognitive diagnosis results and learning data. The learning profile includes: knowledge mastery status (storing the degree of knowledge mastery, problem-solving proficiency, and error tendency output from the three-dimensional cognitive diagnosis); learning behavior characteristics (storing students' learning time distribution, learning frequency, answering speed, and other behavioral characteristics); learning preference characteristics (storing students' resource type preferences, such as preference for video learning, text learning, or practice learning, as well as difficulty preferences); and dynamic memory status, which is updated in real time based on the Ebbinghaus forgetting curve to update the retention rate of learned knowledge points. The update formula for dynamic memory status is: Among them, M K (t) represents the amount of memory retained at time t, M O t represents the initial amount of information to be memorized, t represents the time since the last learning session, and s represents the memory intensity parameter adjusted based on the individual student's memory parameters, dynamically adjusted according to the student's historical review performance.

[0024] The subject knowledge graph module stores pre-built subject knowledge graphs, which include knowledge point nodes, prerequisite dependencies between nodes, knowledge point difficulty, and related relationships. The adaptive learning path planning module uses an improved multi-objective locust optimization algorithm to plan learning paths based on multi-dimensional dynamic learning profiles and subject knowledge graphs. The planning process considers strengthening weak knowledge points, pre-dependent constraints, review node arrangement, and learning efficiency optimization. It also matches learning resources according to learning preferences. The optimization objectives of the improved multi-objective locust optimization algorithm include: minimizing the learning cost of weak knowledge points by prioritizing their learning and ensuring the learning order conforms to the dependencies in the knowledge graph; inserting review nodes at appropriate times based on memory status to maximize memory retention; and minimizing total learning time while meeting the objectives, thereby improving learning efficiency. The improved locust algorithm weights and optimizes these objectives to search for the optimal learning path. Furthermore, it matches corresponding learning resources for each knowledge point based on student learning preferences; for example, students who prefer videos are matched with video explanations, and students who prefer exercises are matched with corresponding practice resources.

[0025] The dynamic path adjustment module updates the learning profile in real time during the learning process. When the deviation between the learning status and the expected status exceeds a preset threshold, path replanning is triggered to achieve closed-loop dynamic adjustment. After completing the learning of each knowledge point, the dynamic path adjustment module updates the learning profile, calculates the deviation between the current cognitive status and the expected status, and triggers path replanning when the deviation exceeds a preset threshold to regenerate the learning path. For example, if a student's mastery of a certain knowledge point is much higher than expected, the system will skip the subsequent exercises for that knowledge point and move directly to the next knowledge point; if a student's mastery of a certain knowledge point is lower than expected, the system will backtrack to the previous weak knowledge points for that knowledge point, adjust the subsequent learning path, strengthen the previous weak points, and then continue learning.

[0026] Reference Figure 1 This invention provides an adaptive AI learning path planning method based on learning profiles, comprising the following steps: S1. System initialization: Construct a knowledge graph for the target subject and input the prerequisites, difficulties, and relationships of knowledge points; S2. Collect students' initial learning data, including basic information, historical learning data and initial assessment data; S3. Preprocess the initial learning data to obtain a standard feature dataset; S4. Using the improved IRT-LSTM-BERT three-dimensional cognitive diagnostic model, the initial cognitive state of students is diagnosed to obtain the degree of mastery of knowledge points, proficiency in problem-solving strategies, and tendency of error causes. S5. Construct an initial multi-dimensional dynamic learning profile, integrating cognitive state, learning behavior, learning preferences, and initial memory state; S6. Based on the initial learning profile and knowledge graph, an improved multi-objective locust optimization algorithm is used to plan the initial learning path and match personalized learning resources. S7. Push learning paths and resources to students and collect real-time interactive data during the learning process; S8. Dynamically update the learning profile based on real-time interactive data; S9. Determine whether the learning objective has been achieved. If yes, end the process and output the learning report; otherwise, proceed to the next step. S10. Determine whether the deviation between the learning state and the expected state exceeds the threshold. If yes, return to step S6 to replan the path; otherwise, return to step S7 to continue learning.

[0027] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An adaptive AI learning path planning system based on learning profiles, characterized in that, It includes a multi-source learning data acquisition module, a data preprocessing module, a three-dimensional cognitive diagnosis module, a multi-dimensional dynamic learning profile construction module, a subject knowledge graph module, an adaptive learning path planning module, and a dynamic path adjustment module; The multi-source learning data acquisition module is used to collect students' multi-source learning data, which includes basic information data, historical learning data, initial assessment data, and real-time interaction data during the learning process. The data preprocessing module is used to preprocess the collected multi-source learning data to obtain a standard feature dataset. The three-dimensional cognitive diagnosis module is used to evaluate students' cognitive status using an improved IRT-LSTM-BERT three-dimensional cognitive diagnosis model and output three types of evaluation results: knowledge point mastery, problem-solving strategy proficiency, and error tendency.

2. The adaptive AI learning path planning system based on learning profiles according to claim 1, characterized in that, The multi-dimensional dynamic learning profile construction module is used to construct a multi-dimensional dynamic learning profile based on cognitive diagnosis results and learning data. The learning profile includes knowledge mastery status, learning behavior characteristics, learning preference characteristics, and dynamic memory status. The dynamic memory status is updated in real time based on the Ebbinghaus forgetting curve to update the memory retention rate of learned knowledge points.

3. The adaptive AI learning path planning system based on learning profiles according to claim 1, characterized in that, The subject knowledge graph module is used to store pre-built subject knowledge graphs, which include knowledge point nodes, prerequisite dependencies between nodes, knowledge point difficulty and association relationships. The adaptive learning path planning module is used to plan learning paths based on multi-dimensional dynamic learning profiles and subject knowledge graphs, using an improved multi-objective locust optimization algorithm. During the planning process, it takes into account strengthening weak knowledge points, prerequisite dependency constraints, review node arrangement and learning efficiency optimization, and matches learning resources according to learning preferences.

4. The adaptive AI learning path planning system based on learning profiles according to claim 1, characterized in that, The dynamic path adjustment module is used to update the learning profile in real time during the learning process. When the deviation between the learning status and the expected status is detected to exceed a preset threshold, path replanning is triggered to achieve closed-loop dynamic adjustment.

5. The adaptive AI learning path planning system based on learning profiles according to claim 1, characterized in that, The data preprocessing module uses an outlier detection algorithm based on statistical distribution to remove outliers from the original data, standardizes features of different dimensions, converts feature values ​​to the range of 0-1, and extracts features from the original data to obtain answer features, behavioral features, and preference features.

6. The adaptive AI learning path planning system based on learning profiles according to claim 1, characterized in that, The three-dimensional cognitive diagnosis module includes an IRT (Item Response Theory) module, an LSTM (Laser-Sensitive Method) temporal module, and a BERT (Bibliographic Method) semantic module. The IRT module is used to assess the mastery of knowledge points based on answer data. The LSTM temporal module is used to assess the proficiency of problem-solving strategies based on answer time-series data. The BERT semantic module is used to attribute error causes based on answer text data to obtain error tendency. The three-dimensional cognitive diagnosis module achieves parameter sharing through multi-task joint training and outputs three types of evaluation results simultaneously.

7. The adaptive AI learning path planning system based on learning profiles according to claim 2, characterized in that, The formula for updating the dynamic memory state is: Wherein, M k (t) is the memory retention amount at time t, M O is the initial memory amount, t is the time since the last learning, s is a memory strength parameter adjusted based on the individual memory parameters of the student, dynamically adjusted according to the historical review effect of the student.

8. The adaptive AI learning path planning system based on learning profiles according to claim 3, characterized in that, The optimization objectives of the improved multi-objective locust optimization algorithm include: minimizing the learning cost of weak knowledge points, satisfying pre-dependency constraints, maximizing memory retention rate, and minimizing total learning time.

9. The adaptive AI learning path planning system based on learning profiles according to claim 1, characterized in that, The dynamic path adjustment module is used to update the learning profile after each knowledge point is learned, calculate the deviation between the current cognitive state and the expected state, and trigger path replanning and regenerate the learning path when the deviation exceeds a preset threshold.

10. An adaptive AI learning path planning method based on learning profiles, characterized in that, Includes the following steps: S1. System initialization: Construct a knowledge graph for the target subject and input the prerequisites, difficulties, and relationships of knowledge points; S2. Collect students' initial learning data, including basic information, historical learning data and initial assessment data; S3. Preprocess the initial learning data to obtain a standard feature dataset; S4. Using the improved IRT-LSTM-BERT three-dimensional cognitive diagnostic model, the initial cognitive state of students is diagnosed to obtain the degree of mastery of knowledge points, proficiency in problem-solving strategies, and tendency of error causes. S5. Construct an initial multi-dimensional dynamic learning profile, integrating cognitive state, learning behavior, learning preferences, and initial memory state; S6. Based on the initial learning profile and knowledge graph, an improved multi-objective locust optimization algorithm is used to plan the initial learning path and match personalized learning resources. S7. Push learning paths and resources to students and collect real-time interactive data during the learning process; S8. Dynamically update the learning profile based on real-time interactive data; S9. Determine whether the learning objective has been achieved. If so, end the process and output the learning report. If not, proceed to the next step; S10. Determine whether the deviation between the learning state and the expected state exceeds the threshold. If yes, return to step S6 to replan the path; otherwise, return to step S7 to continue learning.

Images