Personalized learning recommendation method and device based on AI workflow, and electronic equipment

Abstract

The invention relates to the technical field of computers, in particular to a personalized learning recommendation method and device based on AI workflow and electronic equipment, after a user authorizes, learning behavior data of the user on a teaching platform is collected in real time, based on the collected data, the current learning state and ability level of the user are evaluated, the learning requirement is determined, and according to the learning requirement, the learning state and ability level of the user are recommended. According to the method, the difficulty encountered when a user learns course content is predicted, a course set suitable for current learning of the user is screened out from a preset course library according to a prediction result and learning requirements, and a personalized course recommendation list is generated in combination with interest preference and real-time learning behavior data of the user. Meanwhile, by collecting and analyzing feedback suggestions of the user on the recommended course and adjusting the recommendation strategy in real time according to the feedback, immediate feedback and recommendation adjustment on new questions or demands of the user are achieved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a personalized learning recommendation method and device based on AI workflow, and an electronic device. Background Art

[0002] In the current field of educational technology, recommendation systems have become a key tool to help users discover courses of interest. Some recommendation methods rely on a relatively simple mechanism, namely, recommendation based on popularity, but it fails to take these individual differences into account, which may lead to course recommendations that are not suitable for specific learners' needs.

[0003] In order to overcome the above limitations, AI workflow technology has been introduced into the field of education. AI workflow uses AI algorithms and tools to realize functions such as automatic task allocation, intelligent data processing, and auxiliary support for decision-making, so as to improve work efficiency, reduce costs, improve quality, and enhance the competitiveness of enterprises. Existing announcement number CN118132858B discloses a personalized learning recommendation method and system based on AI. The method includes: filtering out target data from the user's historical learning data on the teaching platform according to the user's course query request, determining the first difficulty level of the user's knowledge identification of the target course from the target data, querying the second difficulty level corresponding to the first difficulty level, and filtering out the target course set from the course library according to the second difficulty level; inputting the request time of the course query request and the learning feature data extracted from the user's historical learning data into the content format analysis model, outputting the user's target recommendation format, determining the target teaching course corresponding to the target recommendation format in the target course set, and recommending courses to the user according to the heat value of each teaching course and the target teaching course. Improve the recommendation accuracy of the teaching platform for teaching courses and improve the user's learning experience and learning efficiency.

[0004] However, the above-mentioned AI-based personalized learning recommendation method and system mainly rely on the user's historical learning data on the teaching platform, and fail to make dynamic adjustments based on the user's real-time performance. Users may encounter new problems or needs during the learning process, and the existing system cannot provide instant feedback and adjust recommendations. Summary of the invention

[0005] The purpose of the present invention is to provide a personalized learning recommendation method and device, and electronic device based on AI workflow, which solves the problem that the existing AI-based personalized learning recommendation method and system mainly rely on the user's historical learning data on the teaching platform, and fail to make dynamic adjustments based on the user's real-time performance. The user may encounter new problems or needs during the learning process, and the existing system cannot provide immediate feedback and adjust the recommendations.

[0006] To achieve the above object, the present invention provides a personalized learning recommendation method based on AI workflow, comprising the following steps: After the user authorizes, the user's learning behavior data on the teaching platform is collected in real time. The learning behavior data includes learning time, learning progress, exercise completion status, test scores and the user's physiological response data, and the collected real-time learning behavior data is stored in the cloud database; Based on the collected real-time learning behavior data, the machine learning algorithm is used to evaluate the user's current learning status and ability level, obtain the evaluation results, and determine the user's learning needs based on the evaluation results; Predict the difficulty that users will encounter when learning course content based on their learning needs, and use this as the prediction result; According to the prediction results and learning needs, select the course set suitable for the user's current learning from the preset course library; In the selected course collection, a personalized course recommendation list is generated by combining the user's interest preferences and real-time learning behavior data; Collect and analyze user feedback on recommended courses in the course recommendation list, and adjust the recommendation strategy in real time based on the feedback.

[0007] Among them, after the user's authorization, the user's learning behavior data on the teaching platform is collected in real time. The learning behavior data includes learning time, learning progress, exercise completion status, test scores and the user's physiological response data, and the collected real-time learning behavior data is stored in the cloud database. The specific steps include: obtaining the user's explicit authorization through the user agreement or privacy policy on the teaching platform; recording the time when the user logs into the teaching platform, the length of study and the frequency of learning activities, tracking the user's progress in completing course chapters, tasks or projects, collecting the user's completion of exercises, homework or tests, recording the scores of various tests participated by the user, and collecting the user's physiological response data through cameras and microphones to obtain real-time learning behavior data; cleaning, organizing and standardizing the real-time learning behavior data, and storing them in the cloud database.

[0008] Among them, based on the collected real-time learning behavior data, a machine learning algorithm is used to evaluate the user's current learning status and ability level, obtain the evaluation results, and determine the user's learning needs based on the evaluation results. The specific steps include: Extracting features related to the user's learning status and ability level from the stored data as user learning status and ability level features, the user learning status and ability level features including the length of learning time, the speed of learning progress, the accuracy of exercise completion, the distribution of test scores, and the fluctuation of physiological response data; Using historical learning behavior data as a training set, the selected machine learning algorithm is trained. During the training process, the machine learning algorithm will learn how to extract useful information from the user's learning status and ability level characteristics and build a first prediction model; Applying the trained first prediction model to the real-time collected learning behavior data to evaluate the user's current learning status and ability level; Based on the evaluation results, combined with the learning resources and course objectives of the teaching platform, determine the user's current learning needs.

[0009] Among them, the difficulty that users will encounter when learning course content is predicted according to learning needs, and used as the prediction result. The specific steps include: Obtain the course content to be studied in the learning needs; Extract features closely related to the course content to be studied as course content features, which include the complexity of the course content, the relevance of knowledge points, and the distribution of difficulty; Integrate the user's learning status and ability level characteristics with the course content characteristics to form a comprehensive feature set; A difficulty prediction model is constructed and trained. During the training process, the difficulty prediction model learns how to extract useful information from a comprehensive feature set.

[0010] Among them, a difficulty prediction model is constructed and trained. During the training process, the difficulty prediction model learns how to extract useful information from the comprehensive feature set. After that, the following steps are also included: The trained difficulty prediction model is applied to the real-time collected learning behavior data and course content characteristics to predict the difficulty that users will encounter when learning the course content, including the questions they encounter or the knowledge points that need further study.

[0011] Among them, according to the prediction results and learning needs, a set of courses suitable for the user's current learning is screened out from the preset course library. The specific steps include: Conduct in-depth analysis of the course content that users are about to learn and identify key knowledge points and skills; Go through the preset course library and search for courses related to key knowledge points and skills; Use machine learning algorithms or deep learning algorithms to match and screen courses by combining historical learning data with the correlation between course knowledge points; Based on the predicted user learning needs and the difficulty they may encounter, the courses that best suit the user's current learning are selected from the matching courses and combined into a course collection.

[0012] Among them, in the selected course set, a personalized course recommendation list is generated by combining the user's interest preferences and real-time learning behavior data. The specific steps include: Assess the credibility of each course in the course collection and assign a credibility weight; Analyze the course content of each course in the course collection, identify the key knowledge points and skills, match the knowledge points and skills with the predicted questions that users will encounter when learning the course content, calculate the matching degree, and assign a correlation weight with the predicted question to each course based on the matching degree; Extract user interest tags from the user's historical learning behavior data, match each course with the student's interest tags, calculate the interest matching degree, and assign an interest preference weight to each course based on the interest matching degree; Calculate the total learning time required for users to complete the course. Based on the user's learning status and ability level, as well as the complexity of the course, the relevance of knowledge points and the distribution of difficulty, evaluate the difficulty of the user to learn each course. Combine the total time and total difficulty to assign a comprehensive difficulty weight to each course. The weight of the course's credibility, the weight of its relevance to the predicted question, the weight of its interest preference, and the weight of its comprehensive difficulty are weighted and summed to obtain the comprehensive weight of each course. According to the comprehensive weight of each course, the courses are sorted, and courses are selected according to the comprehensive weight to generate a personalized course recommendation list.

[0013] Among them, collect and analyze users' feedback on the recommended courses in the course recommendation list, and adjust the recommendation strategy in real time according to the feedback. The specific steps include: designing a user feedback mechanism in the teaching platform or learning application to collect feedback; using natural language processing technology to perform sentiment analysis on the feedback to identify users' positive or negative comments on the recommended courses, as well as their specific sentiment tendencies; counting the proportion of sentiment tendencies to explore users' deep needs for courses.

[0014] A personalized learning recommendation device based on AI workflow, comprising a data acquisition module, an evaluation module, a prediction module, a screening module, a personalized recommendation module and an optimization module, wherein the evaluation module is connected to the data acquisition module, the prediction module is connected to the evaluation module, the screening module is connected to the prediction module, the personalized recommendation module is connected to the screening module, and the optimization module is connected to the personalized recommendation module; The data collection module is used to collect the user's learning behavior data on the teaching platform in real time after the user authorizes the user. The learning behavior data includes learning time, learning progress, exercise completion status, test scores and the user's physiological response data, and store the collected real-time learning behavior data in the cloud database; The evaluation module is used to evaluate the user's current learning status and ability level based on the collected real-time learning behavior data using a machine learning algorithm, obtain an evaluation result, and determine the user's learning needs based on the evaluation result; The prediction module predicts the difficulty that the user will encounter when learning the course content according to the learning needs, and uses the prediction result as the prediction result; The screening module screens out a set of courses suitable for the user's current study from a preset course library according to the prediction results and learning needs; The personalized recommendation module generates a personalized course recommendation list in the screened course set by combining the user's interest preferences and real-time learning behavior data; The optimization module is used to collect and analyze user feedback on the recommended courses in the course recommendation list, and adjust the recommendation strategy in real time according to the feedback.

[0015] An electronic device for personalized learning recommendation based on AI workflow includes a memory, a processor, and a computer program stored in the memory and capable of running on the processor. The memory stores a computer program capable of running on the processor. When the processor loads and executes the computer program, the personalized learning recommendation method based on AI workflow is adopted.

[0016] The present invention solves the defects existing in the technical background, and the beneficial technical effects of the present invention are: The invention discloses a personalized learning recommendation method and device based on AI workflow, and an electronic device. After the user authorizes, the learning behavior data on the teaching platform is collected in real time and stored in a cloud database. Based on the collected data, the machine learning algorithm is used to evaluate the user's current learning status and ability level, determine the learning needs, and predict the difficulty that the user will encounter when learning the course content according to the learning needs. According to the prediction results and learning needs, a set of courses suitable for the user's current learning is screened out from the preset course library. In the screened course set, the user's interest preferences and real-time learning behavior data are combined to generate a personalized course recommendation list, collect and analyze the user's feedback on the recommended courses, and adjust the recommendation strategy in real time according to the feedback. According to the prediction results and learning needs, a set of courses suitable for the user's current learning is screened out from the preset course library, and the user's interest preferences and real-time learning behavior data are combined to generate a personalized course recommendation list. At the same time, by collecting and analyzing the user's feedback on the recommended courses, the recommendation strategy is adjusted in real time according to the feedback, so as to realize the instant feedback and recommendation adjustment of the user's new problems or needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art are briefly introduced below.

[0018] Figure 1 It is a flowchart of the steps of the personalized learning recommendation method based on AI workflow of the first embodiment of the present invention.

[0019] Figure 2 This is a flowchart of the steps of generating a personalized course recommendation list in the screened course set by combining the user's interest preferences and real-time learning behavior data in the first embodiment of the present invention.

[0020] Figure 3 It is a principle block diagram of a personalized learning recommendation device based on AI workflow according to the second embodiment of the present invention.

[0021] Figure 4 The data collected in the personalized learning recommendation method based on AI workflow in the first embodiment of the present invention is organized into a structured data table.

[0022] Figure 5 It is the historical learning behavior data in the personalized learning recommendation method based on AI workflow in the first embodiment of the present invention.

[0023] Figure 6 It is a data table of the comprehensive feature set in the personalized learning recommendation method based on AI workflow in the first embodiment of the present invention.

[0024] Figure 7 It is the historical data used to train the difficulty prediction model in the personalized learning recommendation method based on AI workflow in the first embodiment of the present invention.

[0025] Figure 8 It is a data table of a course library preset in the personalized learning recommendation method based on AI workflow in the first embodiment of the present invention.

[0026] Fig. 9 It is the historical learning data used to train the course matching model in the personalized learning recommendation method based on AI workflow in the first embodiment of the present invention.

[0027] Fig.10 It is a data table for calculating the matching degree of each course in the personalized learning recommendation method based on AI workflow in the first embodiment of the present invention.

[0028] Fig.11 It is a data table of a course set in the personalized learning recommendation method based on AI workflow in the first embodiment of the present invention.

[0029] Fig.12 It is a data table of credibility in the personalized learning recommendation method based on AI workflow in the first embodiment of the present invention.

[0030] Fig.13It is a data table of key knowledge points and skills of courses in the personalized learning recommendation method based on AI workflow in the first embodiment of the present invention.

[0031] Fig.14 It is a data table of the matching degree between courses and interest tags in the personalized learning recommendation method based on AI workflow in the first embodiment of the present invention.

[0032] Fig.15 This is a data table of the total learning time and difficulty of courses in the personalized learning recommendation method based on AI workflow in the first embodiment of the present invention.

[0033] Fig.16 This is a data table of the comprehensive weight of each course in the personalized learning recommendation method based on AI workflow in the first embodiment of the present invention.

[0034] In the figure: 201-data collection module, 202-evaluation module, 203-prediction module, 204-screening module, 205-personalized recommendation module, 206-optimization module. DETAILED DESCRIPTION

[0035] Embodiments of the present invention are described in detail below. Examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, but should not be construed as limiting the present invention.

[0036] The first embodiment of the present application is: See also Figure 1 and Figure 2 ,in, Figure 1 It is a flowchart of the steps of the personalized learning recommendation method based on AI workflow of the first embodiment of the present invention. Figure 2 This is a flowchart of the steps of generating a personalized course recommendation list in the screened course set by combining the user's interest preferences and real-time learning behavior data in the first embodiment of the present invention.

[0037] The present invention provides a personalized learning recommendation method based on AI workflow, comprising the following steps: S101: After the user authorizes, the user's learning behavior data on the teaching platform is collected in real time, the learning behavior data includes learning time, learning progress, exercise completion status, test scores and the user's physiological response data, and the collected real-time learning behavior data is stored in the cloud database; Specifically, when a user registers, a user agreement and privacy policy page pops up to clearly inform the user of the purpose, scope and method of use of data collection. When the user clicks the "Agree" button, it indicates that the platform is explicitly authorized to collect its learning behavior data to ensure the legality of data collection. Record the specific timestamp of each time the user logs into the teaching platform, for example, the user logs into the platform at 14:00 on January 16, 2025. Subtract the timestamps to get the total time from the user's login to the platform exit, for example, the user's learning time on the platform is 165 minutes, and count the number of times the user logs into the platform within a certain period of time, for example, the user logs into the platform twice on this day. Track the user's progress in completing course chapters, tasks or projects, and express it in percentage form, for example, the user has completed Chapter 1 (a total of 5 chapters) of the Python Basics course, and the progress is 20%. Collect the user's completion of exercises, homework or tests, record the score and total score of each question, and calculate statistical indicators such as average score and pass rate. For example, the user completes 10 exercises and gets 8 correct, with an accuracy rate of 80%; completes a small test with a total score of 100 points and a score of 85 points. Capture the user's facial expressions and eyes through the camera, and use computer vision technology to analyze the user's concentration, emotional state, etc. For example, through the facial expression recognition algorithm, determine that the user's expression is more focused in the first 30 minutes, and the expression shows confusion many times in the next 150 minutes. Record the user's voice during the learning process, analyze the speech speed, intonation and other characteristics of the voice to determine the user's learning status. For example, when the user encounters a difficult problem, the speech speed speeds up and the tone rises. Delete and correct obviously erroneous data such as abnormally long learning time and unreasonable grades, use the mean, median or mode to fill in the missing data, or use interpolation methods to fill in the missing data, and organize the collected data into a structured table form, such as Figure 4 shown.

[0038] Scale the data to a fixed range, such as [0, 1] or [-1, 1]. Common methods include Min-Max normalization and Z-Score normalization. Choose MySQL as the relational database to store the processed data in it to ensure the security and accessibility of the data and provide data support for subsequent personalized learning recommendations.

[0039] S102: Based on the collected real-time learning behavior data, use machine learning algorithms to evaluate the user's current learning status and ability level, and determine the user's learning needs based on the evaluation results. Specifically, extract the length of learning time: learning time (165 minutes), the speed of learning progress: course progress (20%), the accuracy of exercise completion: exercise question accuracy (80%), the distribution of test scores: test scores (85 points), the fluctuation of physiological response data: facial expression analysis (focus / confusion), voice analysis (faster speech speed / high tone). Assume that historical learning behavior data, such as Figure 5shown.

[0040] These historical data are used as training sets, and the random forest algorithm is selected for training. During the training process, the algorithm will learn how to extract useful information from the features and build the first prediction model. The trained first prediction model is used to evaluate the real-time learning behavior data of user 12345. According to the evaluation results (learning status: good, ability level: medium), combined with the learning resources and course objectives of the teaching platform, the current learning needs of user 12345 are determined: Course progress: The user has completed Chapter 1 of the Python Basics course, with a progress of 20%. Learning resources: The teaching platform provides the content of Chapter 2 of the Python Basics course, including video tutorials, exercises, and quizzes. Course objectives: Help users master the basic syntax and data structures of Python and be able to write simple Python programs. Based on this information, the current learning needs of user 12345 are determined: Content that needs to be further learned: Chapter 2 of the Python Basics course, especially the data structure part. Recommended exercises: Exercises related to the content of Chapter 2, focusing on the operations of lists, dictionaries, and sets. Recommended quizzes: Complete the quizzes after Chapter 2 to test the user's mastery of data structures. This provides a basis for subsequent personalized course recommendations.

[0041] S103: Predict the difficulty that the user will encounter when learning the course content based on learning needs. Specifically, according to the learning needs of user 12345, he is about to learn Chapter 2 of the Python Basic Course, which includes data structures (lists, dictionaries, and sets). For Chapter 2 of the Python Basic Course, the following features are extracted: Complexity of course content: The complexity of the data structure part is relatively high because it involves the use and operation of multiple data types. Relevance of knowledge points: The data structure part has a strong correlation with the basic grammar of Chapter 1, and users need to have a good grasp of the basic grammar. Difficulty distribution: The operation of lists is relatively simple, and the operation of dictionaries and sets is of medium difficulty. The specific values ​​are set as follows: Complexity: 0.8 (1 means very complex, 0 means very simple), Relevance: 0.7 (1 means very strong correlation, 0 means no correlation), Difficulty: 0.6 (1 means very difficult, 0 means very easy). The learning status and ability level features of user 12345 are as follows: learning status: good (0.7, 1 for excellent, 0 for poor), ability level: medium (0.5, 1 for advanced, 0 for elementary), learning time: 165 minutes (0.55 after normalization), course progress: 20% (0.20 after normalization), exercise accuracy: 80% (0.80 after normalization), test score: 85 points (0.85 after normalization). These features are integrated with the course content features to form a comprehensive feature set, such as Figure 6 shown.

[0042] Assume that the historical data used to train the difficulty prediction model is Figure 7 As shown. Using these historical data as the training set, linear regression is selected for training. During the training process, the algorithm will learn how to extract useful information from the comprehensive feature set and build a difficulty prediction model. The trained difficulty prediction model is applied to the comprehensive feature set of user 12345 to predict the difficulty he may encounter when learning Chapter 2 of the Python Basics Course: Assume that the model prediction results are as follows: Predicted difficulty: 0.55 (1 means very difficult, 0 means very easy). According to the predicted difficulty of 0.55, user 12345 may encounter moderate challenges when learning Chapter 2 of the Python Basics Course. This means that users need additional attention and resources to help them better understand and master the content of the data structure part. It not only predicts the difficulty that users may encounter when learning the course content, but also provides a basis for subsequent personalized course recommendations and resource allocation. This helps the teaching platform provide users with more accurate and targeted learning support.

[0043] S104: Based on the prediction results and learning needs, select a set of courses suitable for the user's current study from the preset course library. Specifically, user 12345 is about to study Chapter 2 of the Python Basics course, which includes data structures (lists, dictionaries, and sets). The key knowledge points and skills are as follows: Lists: create, index, slice, add, delete, sort; Dictionary: creation, key-value pair operations, traversal, update, and deletion; Sets: create, add, delete, set operations (union, intersection, difference); Traverse the preset course library and search for courses associated with key knowledge points and skills. Assume that the preset course library is Figure 8 shown.

[0044] Using machine learning or deep learning algorithms, combined with the correlation between historical learning data and course knowledge points, courses can be matched and screened. Assume that the historical learning data used to train the course matching model is Fig. 9 As shown. Using these historical data as the training set, the random forest algorithm is selected for training. During the training process, the algorithm will learn how to predict the learning effect from the user's learning status, ability level and course ID.

[0045] The learning status of user 12345 is 0.7, the ability level is 0.5, and the predicted difficulty is 0.55. Filter the most suitable course from the preset course library.

[0046] First, calculate the matching degree of each course: learning status × ability level × (1-course difficulty), such as Fig.10 shown.

[0047] According to the matching degree, filter out the top 3 courses with the highest matching degree: Python basic grammar review (course ID: 20005, matching degree: 0.245); Python list operations (course ID: 20001, matching degree: 0.21); Python dictionary operations (course ID: 20002, match: 0.175); Finally, the course set recommended for user 12345 is as follows: Python basic grammar review (Course ID: 20005); Python list operations (Course ID: 20001); Python dictionary operations (Course ID: 20002); Based on the learning status, ability level and predicted difficulty of user 12345, the following courses are recommended: Python basic grammar review: helps users consolidate basic grammar and lay a solid foundation for learning data structures.

[0048] Python list operations: Focus on list operations to help users master basic operations such as list creation, indexing, slicing, etc.

[0049] Python dictionary operations: Focus on dictionary operations to help users master basic operations such as dictionary creation and key-value pair operations.

[0050] These courses not only meet the learning needs of users, but also help users overcome the medium-difficulty challenges they may encounter in learning data structures. Through such recommendations, the teaching platform can provide users with a more accurate and targeted learning path.

[0051] S105: In the selected course set, a personalized course recommendation list is generated by combining the user's interest preferences and real-time learning behavior data; S1051: Evaluate the credibility of each course in the course set and assign a credibility weight; S1052: analyzing the course content of each course in the course set, identifying key knowledge points and skills therein, matching the knowledge points and skills with predicted questions that the user may encounter when learning the course content, calculating the matching degree, and assigning a correlation weight with the predicted question to each course according to the matching degree; S1053: extracting the user's interest tags from the user's historical learning behavior data, matching each course with the student's interest tags, calculating the interest matching degree, and assigning an interest preference weight to each course according to the interest matching degree; S1054: Calculate the total learning time required for the user to complete the course, evaluate the difficulty of the user learning each course based on the user's learning status and ability level, as well as the complexity of the course, the relevance of knowledge points and the distribution of difficulty, and assign a comprehensive difficulty weight to each course in combination with the total time and total difficulty; S1055: performing weighted summation of the credibility weight of the course, the weight of the relevance to the predicted question, the interest preference weight, and the comprehensive difficulty weight to obtain the comprehensive weight of each course; S1056: Sort the courses according to the comprehensive weight of each course, select courses according to the comprehensive weight, and generate a personalized course recommendation list.

[0052] Specifically, a collection of courses, such as Fig.11 shown.

[0053] Assess the credibility of each course based on course ratings, completion rates, user feedback, etc. Assume credibility, e.g. Fig.12 shown.

[0054] The prediction questions that user 12345 may encounter in the learning data structure section include: Advanced operations on lists (such as multidimensional lists); Complex operations on dictionaries (such as nested dictionaries); Advanced operations on sets (such as power sets); The key knowledge points and skills of each course, such as Fig.13 shown.

[0055] The weight of the relevance to the predicted question is calculated by analyzing the correlation between the course content and the predicted questions that the user may encounter.

[0056] User 12345's interest tags include: data structure, algorithm, and project practice.

[0057] The matching degree of each course with interest tags, such as Fig.14 As shown in the figure, the specific steps for calculating the interest preference weight are as follows: Collect behavioral data: record the course pages browsed by users, the time spent, the number of courses completed, etc. Calculate interest scores: assign an interest score to each course topic or category. For example, the longer a user spends on a topic and the more courses they complete, the higher the interest score for that topic. Normalization: normalize the interest score to a fixed range (such as [0, 1]) for comparison and use.

[0058] Assume that the learning status of user 12345 is 0.7 and the ability level is 0.5. The total learning time and difficulty of each course are as follows: Fig.15 shown.

[0059] The credibility weight of the course, the weight of the correlation with the predicted question, the interest preference weight and the comprehensive difficulty weight are weighted and summed to obtain the comprehensive weight of each course.

[0060] Assume that the weighting coefficients of each weight are: Credibility weight: 0.3; Weight of correlation with prediction question: 0.3; Interest preference weight: 0.2; Comprehensive difficulty weight: 0.2; Calculate the comprehensive weight of each course, such as Fig.16 shown.

[0061] Comprehensive weight calculation formula: Comprehensive weight = 0.3 × credibility weight + 0.3 × correlation weight with predicted question + 0.2 × interest preference weight + 0.2 × comprehensive difficulty weight.

[0062] Sort the courses by their comprehensive weight: Python list operations (course ID: 20001, comprehensive weight: 0.685); Python dictionary operations (course ID: 20002, comprehensive weight: 0.635); Python basic grammar review (course ID: 20005, comprehensive weight: 0.515); Generate a personalized list of course recommendations: Python list operations (Course ID: 20001); Python dictionary operations (Course ID: 20002); Python basic grammar review (Course ID: 20005); Based on the learning status, ability level, interest preference and predicted difficulty of user 12345, the following personalized course recommendation list is generated: Python List Operations: This course is highly relevant to the prediction questions that users may encounter, and users have a high interest in data structures. The course difficulty is moderate.

[0063] Python Dictionary Operations: This course is also related to the prediction questions that users may encounter. Users have a high interest in data structures and the course difficulty is moderate.

[0064] Python Basic Syntax Review: This course helps users consolidate basic grammar and lay a solid foundation for learning data structures. Although it is less relevant to prediction questions, it is helpful for the user's overall learning path.

[0065] Through such recommendations, the teaching platform can provide users with a more accurate and targeted learning path, helping users to master the Python data structure content more effectively.

[0066] S106: Collect and analyze user feedback on recommended courses, and adjust the recommendation strategy in real time based on the feedback.

[0067] Specifically, after users complete each course, a short questionnaire pops up to ask users about their satisfaction, difficulty, content relevance, etc. at each key point of the course, such as at the end of each chapter, a feedback button is provided so that users can instantly express their satisfaction with the current content. A text box is provided so that users can freely express any thoughts and suggestions about the course. Star rating: Users can give the course a star rating, with 1 star indicating very dissatisfied and 5 stars indicating very satisfied. Use natural language processing (NLP) technology to perform sentiment analysis on users' open text feedback. Use pre-trained sentiment analysis models, such as BERT, GPT, etc., to identify sentiment tendencies in text. Calculate the proportion of positive, neutral, and negative reviews for each course. Based on users' specific feedback, identify users' deep needs for course content, difficulty, exercises, etc. For courses that users report needing more exercises (such as Python list operations), add relevant exercises and project practices. For courses that users report are difficult and need more examples (such as Python dictionary operations), add detailed examples and explanations. For courses with relatively basic content (such as Python basic syntax review), provide more advanced content or skip the basic part and go directly to advanced content. Based on the above adjustment strategy, update the course content and recommendation list to ensure that users have a better experience in subsequent learning. For example: Python list operations: Add more exercises and project practices to help users consolidate what they have learned.

[0068] Python dictionary operations: Add detailed examples and explanations to help users better understand complex operations.

[0069] Python basic grammar review: Provide more advanced content or skip the basic part and go directly to advanced content to meet users' needs for more difficult content.

[0070] Continuously monitor user feedback, regularly evaluate the effectiveness of the recommendation strategy, and further optimize it based on new feedback. This can be achieved by regularly collecting user feedback, analyzing learning data, and adjusting the recommendation algorithm.

[0071] Through the above steps, the teaching platform can adjust the recommendation strategy in real time according to user feedback, provide more personalized and effective learning paths, and help users better master knowledge and skills.

[0072] By collecting real-time learning behavior data of users on the teaching platform, such as learning time, learning progress, exercise completion status, test scores, and user physiological response data, and using machine learning algorithms to evaluate users' current learning status and ability level, and determine learning needs, we can provide each user with course recommendations that accurately match their personalized needs, truly realizing personalized recommendations for each person, and avoiding the problem of inaccurate recommendations caused by traditional methods that ignore individual differences.

[0073] It realizes instant feedback and recommendation adjustments for users' new questions or needs. When the user's learning status, ability level or learning needs change, it can capture them in time and update the recommendation strategy accordingly, so that the recommended content always remains highly consistent with the user's current learning situation, ensuring that the user can obtain the most appropriate course resource support throughout the learning process, effectively solving the problem of insufficient dynamic adaptability of the existing system.

[0074] The second embodiment of the present application is: Based on the first embodiment, please refer to Figure 3 ,in, Figure 3 It is a principle block diagram of a personalized learning recommendation device based on AI workflow according to the second embodiment of the present invention.

[0075] The personalized learning recommendation device based on AI workflow in this embodiment includes a data collection module 201, an evaluation module 202, a prediction module 203, a screening module 204, a personalized recommendation module 205 and an optimization module 206.

[0076] According to this specific implementation, the evaluation module 202 is connected to the data acquisition module 201, the prediction module 203 is connected to the evaluation module 202, the screening module 204 is connected to the prediction module 203, the personalized recommendation module 205 is connected to the screening module 204, and the optimization module 206 is connected to the personalized recommendation module 205; The data collection module 201 is used to collect the user's learning behavior data on the teaching platform in real time after the user authorizes the user, and the learning behavior data includes learning time, learning progress, exercise completion status, test scores and the user's physiological response data, and store the collected real-time learning behavior data in the cloud database; The evaluation module 202 is used to evaluate the user's current learning status and ability level based on the collected real-time learning behavior data using a machine learning algorithm, and determine the user's learning needs according to the evaluation results; The prediction module 203 predicts the difficulty that the user may encounter when learning the course content according to the learning needs; The screening module 204 screens out a set of courses suitable for the user's current study from a preset course library according to the prediction results and learning needs; The personalized recommendation module 205 generates a personalized course recommendation list in the screened course set by combining the user's interest preferences and real-time learning behavior data; The optimization module 206 is used to collect and analyze user feedback on recommended courses and adjust the recommendation strategy in real time according to the feedback.

[0077] Using a personalized learning recommendation device based on an AI workflow of the present embodiment, the data acquisition module 201 cleans, organizes and standardizes the collected data, and immediately stores it in a cloud database for use by subsequent modules. The evaluation module 202 uses a machine learning algorithm to evaluate the user's current learning status and ability level based on the learning behavior data provided by the data acquisition module 201. During the evaluation process, features related to the learning status and ability level are extracted, such as the length of learning time, the speed of learning progress, the accuracy of exercise completion, the distribution of test scores, and the fluctuation of physiological response data. The historical learning behavior data is used as a training set to train the selected machine learning algorithm to construct a first prediction model. Then, the trained first prediction model is applied to the learning behavior data collected in real time to evaluate the user's current learning status and ability level, and the user's learning needs are determined based on the evaluation results. The prediction module 203 further predicts the difficulty that the user may encounter when learning the course content according to the learning needs determined by the evaluation module 202. To this end, the prediction module 203 extracts features closely related to the course content as course content features, such as the complexity of the course content, the relevance of knowledge points, and the distribution of difficulty, etc. Then, the user's learning status and ability level features are integrated with the course content features to form a comprehensive feature set. By constructing and training a difficulty prediction model, it is learned how to extract useful information from the comprehensive feature set to predict the difficulty that the user may encounter when learning the course content. The screening module 204 screens out a set of courses suitable for the user's current study from the preset course library according to the prediction results and learning needs provided by the prediction module 203. During the screening process, the course content that the user is about to learn is deeply analyzed to identify the key knowledge points and skills therein. Then, the preset course library is traversed to search for courses associated with the key knowledge points and skills. The courses are matched and screened by using machine learning or deep learning algorithms, combined with the correlation between historical learning data and course knowledge points. Finally, according to the predicted user learning needs and the difficulty that may be encountered, the courses that are most suitable for the user's current study are screened out from the matching courses and combined into a course set.The personalized recommendation module 205 generates a personalized course recommendation list in the screened course set, combining the user's interest preferences and real-time learning behavior data. The specific steps are as follows: first, evaluate the credibility of each course in the course set and assign a credibility weight. Then, analyze the course content of each course, identify the key knowledge points and skills therein, match the knowledge points and skills with the predicted questions that the user will encounter when learning the course content, calculate the matching degree, and assign a correlation weight with the predicted question to each course based on the matching degree. Next, extract the user's interest tag from the user's historical learning behavior data, match each course with the student's interest tag, calculate the interest matching degree, and assign an interest preference weight to each course based on the interest matching degree. Finally, based on the user's learning status and ability level, as well as the complexity of the course, the relevance of knowledge points and the distribution of difficulty, the difficulty of each course is evaluated, and a comprehensive difficulty weight is assigned to each course in combination with the total time and total difficulty. The course's credibility weight, the weight of the relevance to the predicted question, the interest preference weight and the comprehensive difficulty weight are weighted and summed to obtain the comprehensive weight of each course. The courses are sorted according to the comprehensive weight of each course, and courses are selected according to the comprehensive weight to generate a personalized course recommendation list. The optimization module 206 is responsible for collecting and analyzing user feedback on recommended courses, and adjusting the recommendation strategy in real time according to the feedback. A user feedback mechanism is designed in the optimization module 206, such as a comment area in a teaching platform or learning application, a satisfaction survey, etc., to collect users' positive or negative comments on recommended courses and specific emotional tendencies, and use natural language processing technology to perform sentiment analysis on feedback, count the proportion of emotional tendencies, and explore users' deep needs for courses. Based on user feedback data, the optimization module 206 can dynamically adjust the parameters and model structure of the machine learning algorithm, optimize the performance of the prediction model, difficulty prediction model and course recommendation algorithm, and at the same time, update the course content and learning resources in the course library according to changes in user needs and learning trends to ensure the timeliness and relevance of recommended courses.

[0078] Through the personalized learning recommendation device based on AI workflow in this embodiment, the teaching platform can more accurately evaluate the user's learning status and ability level, predict the difficulty that the user will encounter when learning the course content, and generate a personalized course recommendation list in combination with the user's interest preferences and real-time learning behavior data. At the same time, by continuously optimizing the recommendation strategy and learning resources, the teaching platform improves the accuracy and personalization of the recommendation of learning courses, and realizes instant feedback and recommendation adjustments to the user's new questions or needs, thereby further improving the user's learning experience and learning efficiency.

[0079] The third embodiment of the present application is: On the basis of the second embodiment, the electronic device for personalized learning recommendation based on AI workflow in this embodiment includes a memory, a processor, and a computer program stored in the memory and capable of running on the processor.

[0080] An electronic device for personalized learning recommendation based on AI workflow using this embodiment includes: one or more processors, and a memory for storing one or more computer programs; the program includes program instructions, and the processor is used to execute the program instructions stored in the memory. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is used to implement one or more instructions, specifically for loading and executing one or more instructions in a computer storage medium to implement the above method.

[0081] What is disclosed above is only one or more preferred embodiments of the present application, and cannot be used to limit the scope of rights of the present application. Ordinary technicians in this field can understand that all or part of the processes of implementing the above embodiments and equivalent changes made according to the claims of the present application are still within the scope covered by the present application.

Claims

1. A personalized learning recommendation method based on AI workflow, characterized in that: The following steps are involved: After the user authorizes, the user's learning behavior data on the teaching platform is collected in real time. The learning behavior data includes learning time, learning progress, exercise completion status, test scores and the user's physiological response data, and the collected real-time learning behavior data is stored in the cloud database; Based on the collected real-time learning behavior data, the machine learning algorithm is used to evaluate the user's current learning status and ability level, obtain the evaluation results, and determine the user's learning needs based on the evaluation results; Predict the difficulty that users will encounter when learning course content based on their learning needs, and use this as the prediction result; According to the prediction results and learning needs, select the course set suitable for the user's current learning from the preset course library; In the selected course collection, a personalized course recommendation list is generated by combining the user's interest preferences and real-time learning behavior data; Collect and analyze user feedback on recommended courses in the course recommendation list, and adjust the recommendation strategy in real time based on the feedback.

2. The personalized learning recommendation method based on AI workflow according to claim 1, characterized in that: After the user authorizes, the user's learning behavior data on the teaching platform is collected in real time. The learning behavior data includes learning time, learning progress, exercise completion status, test scores and the user's physiological response data, and the collected real-time learning behavior data is stored in the cloud database. The specific steps include: Obtain explicit authorization from users through the user agreement or privacy policy on the teaching platform; Record the time when users log into the teaching platform, the duration of study, and the frequency of learning activities, track the progress of users in completing course chapters, tasks, or projects, collect information about users completing exercises, homework, or tests, record the scores of various tests that users take, and collect users' physiological response data through cameras and microphones to obtain real-time learning behavior data; Real-time learning behavior data is cleaned, organized, and standardized, and stored in a cloud database.

3. The personalized learning recommendation method based on AI workflow according to claim 2, characterized in that: Based on the collected real-time learning behavior data, the machine learning algorithm is used to evaluate the user's current learning status and ability level, obtain the evaluation results, and determine the user's learning needs based on the evaluation results. The specific steps include: Extracting features related to the user's learning status and ability level from the stored data as user learning status and ability level features, the user learning status and ability level features including the length of learning time, the speed of learning progress, the accuracy of exercise completion, the distribution of test scores, and the fluctuation of physiological response data; Using historical learning behavior data as a training set, the selected machine learning algorithm is trained. During the training process, the machine learning algorithm will learn how to extract useful information from the user's learning status and ability level characteristics and build a first prediction model; Applying the trained first prediction model to the real-time collected learning behavior data to evaluate the user's current learning status and ability level; Based on the evaluation results, combined with the learning resources and course objectives of the teaching platform, determine the user's current learning needs.

4. The personalized learning recommendation method based on AI workflow according to claim 3, characterized in that: Predict the difficulty that users will encounter when learning course content based on learning needs, and use it as the prediction result. The specific steps include: Obtain the course content to be studied in the learning needs; Extract features closely related to the course content to be studied as course content features, which include the complexity of the course content, the relevance of knowledge points, and the distribution of difficulty; Integrate the user's learning status and ability level characteristics with the course content characteristics to form a comprehensive feature set; A difficulty prediction model is constructed and trained. During the training process, the difficulty prediction model learns how to extract useful information from a comprehensive feature set.

5. The personalized learning recommendation method based on AI workflow according to claim 4, characterized in that: Construct a difficulty prediction model and train it. During the training process, the difficulty prediction model learns how to extract useful information from the comprehensive feature set. After that, the following steps are also included: The trained difficulty prediction model is applied to the real-time collected learning behavior data and course content characteristics to predict the difficulty that users will encounter when learning the course content, including the questions they encounter or the knowledge points that need further study.

6. The personalized learning recommendation method based on AI workflow according to claim 5, characterized in that: According to the prediction results and learning needs, a set of courses suitable for the user's current learning is selected from the preset course library. The specific steps include: Conduct in-depth analysis of the course content that users are about to learn and identify key knowledge points and skills; Go through the preset course library and search for courses related to key knowledge points and skills; Use machine learning algorithms or deep learning algorithms to match and screen courses by combining historical learning data with the correlation between course knowledge points; Based on the predicted user learning needs and the difficulty they may encounter, the courses that best suit the user's current learning are selected from the matching courses and combined into a course collection.

7. The personalized learning recommendation method based on AI workflow according to claim 6, characterized in that: In the selected course collection, a personalized course recommendation list is generated by combining the user's interest preferences and real-time learning behavior data. The specific steps include: Assess the credibility of each course in the course collection and assign a credibility weight; Analyze the course content of each course in the course collection, identify the key knowledge points and skills, match the knowledge points and skills with the predicted questions that users will encounter when learning the course content, calculate the matching degree, and assign a correlation weight with the predicted question to each course based on the matching degree; Extract user interest tags from the user's historical learning behavior data, match each course with the student's interest tags, calculate the interest matching degree, and assign an interest preference weight to each course based on the interest matching degree; Calculate the total learning time required for users to complete the course. Based on the user's learning status and ability level, as well as the complexity of the course, the relevance of knowledge points and the distribution of difficulty, evaluate the difficulty of the user to learn each course. Combine the total time and total difficulty to assign a comprehensive difficulty weight to each course. The weight of the course's credibility, the weight of its relevance to the predicted question, the weight of its interest preference, and the weight of its comprehensive difficulty are weighted and summed to obtain the comprehensive weight of each course. According to the comprehensive weight of each course, the courses are sorted, and courses are selected according to the comprehensive weight to generate a personalized course recommendation list.

8. The personalized learning recommendation method based on AI workflow according to claim 7, characterized in that: Collect and analyze user feedback on the recommended courses in the course recommendation list, and adjust the recommendation strategy in real time based on the feedback. The specific steps include: Design user feedback mechanisms in teaching platforms or learning applications to collect feedback; Use natural language processing technology to conduct sentiment analysis on feedback to identify users' positive or negative comments on recommended courses, as well as their specific sentiment tendencies; Count the proportion of emotional tendencies and explore users' deep needs for courses.

9. A personalized learning recommendation device based on AI workflow, comprising a module for executing the personalized learning recommendation method based on AI workflow according to any one of claims 1 to 8, characterized in that: It includes a data collection module, an evaluation module, a prediction module, a screening module, a personalized recommendation module and an optimization module, wherein the evaluation module is connected to the data collection module, the prediction module is connected to the evaluation module, the screening module is connected to the prediction module, the personalized recommendation module is connected to the screening module, and the optimization module is connected to the personalized recommendation module; The data collection module is used to collect the user's learning behavior data on the teaching platform in real time after the user authorizes the user. The learning behavior data includes learning time, learning progress, exercise completion status, test scores and the user's physiological response data, and store the collected real-time learning behavior data in the cloud database; The evaluation module is used to evaluate the user's current learning status and ability level based on the collected real-time learning behavior data using a machine learning algorithm, obtain an evaluation result, and determine the user's learning needs based on the evaluation result; The prediction module predicts the difficulty that the user will encounter when learning the course content according to the learning needs, and uses the prediction result as the prediction result; The screening module screens out a set of courses suitable for the user's current study from a preset course library according to the prediction results and learning needs; The personalized recommendation module generates a personalized course recommendation list in the screened course set by combining the user's interest preferences and real-time learning behavior data; The optimization module is used to collect and analyze user feedback on the recommended courses in the course recommendation list, and adjust the recommendation strategy in real time according to the feedback.

10. An electronic device for personalized learning recommendation based on AI workflow, comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that: The memory stores a computer program that can be run on a processor. When the processor loads and executes the computer program, the personalized learning recommendation method based on AI workflow described in any one of claims 1 to 8 is adopted.

Images