High data intelligent tutoring method and system based on personalized learning path

Abstract

The invention discloses a high data intelligent tutoring method and system based on a personalized learning path, and the method comprises the following steps: S1, enabling a system to receive personal data transmitted by a user, and carrying out the multi-dimensional diagnosis and data analysis; s2, dynamically screening data resources according to user requirements, and designing a personalized learning path; s3, based on database retrieval, outputting a question answering step, and tracking the learning progress in real time; and S4, timely reminding a lagged user or a user with a poor effect, and providing personalized improvement suggestions. According to the method, data resources are dynamically screened and a personalized learning path is designed according to personal data and learning requirements of the user, so that the pertinence and effectiveness of learning are improved, secondly, feature extraction and prediction are performed by using the CNN model, the learning style and preference of the user can be accurately captured, and the learning efficiency is improved. And more accurate learning resource recommendation and answering steps are provided for the user.

Description

Technical Field

[0001] The present invention relates to the field of intelligent teaching technology, and in particular to a high-level intelligent tutoring method based on a personalized learning path. Background Art

[0002] The purpose of advanced mathematics tutoring is to help students deeply understand and master the core concepts and problem-solving methods of advanced mathematics, laying a solid foundation for their subsequent academic research and practical application. The tutoring content is extensive, from reviewing the basic knowledge of elementary mathematics to in-depth analysis of the core content of advanced mathematics such as series and limits, calculus, spatial analytic geometry and linear algebra, series and ordinary differential equations, ensuring that students can build a complete knowledge system. Through systematic explanations and exercises, students can gradually master the basic methods and problem-solving skills of advanced mathematics, and improve their ability to analyze and solve problems.

[0003] Advanced mathematics tutoring not only focuses on imparting theoretical knowledge, but also emphasizes the cultivation of practical application ability, so that students can use the knowledge they have learned to solve practical problems. At the same time, the tutoring process also focuses on cultivating students' logical thinking and abstract thinking abilities, providing strong support for their future academic research and career development.

[0004] However, there are some problems with the existing technology: higher mathematics learning often adopts a unified teaching schedule and teaching content, ignoring students' individual differences and learning needs, resulting in some students feeling that learning is difficult, while other students may find the content too simple. In addition, students need to find and screen learning materials, which is not only time-consuming and laborious, but also may find that the resources they find do not meet their learning needs or interests due to lack of experience or information. Therefore, we propose a high-level mathematics intelligent tutoring method and system based on personalized learning paths. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides a method and system for intelligent tutoring of high mathematics based on personalized learning paths. The method of the present invention dynamically screens data resources and designs personalized learning paths according to the user's personal data and learning needs, thereby improving the pertinence and effectiveness of learning. Secondly, by utilizing the convolutional neural network (CNN) model for feature extraction and prediction, it can accurately capture the user's learning style and preference, and provide the user with more accurate learning resource recommendations and solution steps.

[0006] The object of the present invention is achieved by: a method and system for intelligent tutoring of advanced mathematics based on personalized learning paths, comprising the following steps:

[0007] S1, the system receives personal data transmitted by the user and performs multi-dimensional diagnosis and data analysis;

[0008] S2, dynamically screens data resources and designs personalized learning paths based on user needs;

[0009] S3, based on database retrieval, outputs the problem-solving steps and tracks the learning progress in real time;

[0010] S4, promptly reminds users who are lagging behind or experiencing poor performance, and provides personalized improvement suggestions.

[0011] Furthermore, the acquisition of personal data in step S1 includes: questionnaire survey, test questions and learning history data import;

[0012] Questionnaire: Provide online questionnaires including multiple-choice, fill-in-the-blank and open-ended questions to collect users' learning experience, interests and hobbies, and learning goals;

[0013] Test questions: Provide test questions and give corresponding scores to evaluate the user's mathematical foundation, problem-solving ability and learning style;

[0014] The test questions include basic contents in multiple mathematics fields;

[0015] Importing learning history data: allows users to import learning history data, including historical grades, homework completion status, and online course learning records.

[0016] Furthermore, the specific process of step 2 is as follows:

[0017] S11, pre-processing user personal data;

[0018] S12, using the convolutional layer and pooling layer of the CNN model, extracts key features from personal data and outputs them to the fully connected layer;

[0019] S13, applies nonlinear activation function in the fully connected layer to increase the nonlinear expression ability of the model;

[0020] S14, integrating and classifying or performing regression prediction on the extracted features to obtain a prediction result of the user's learning needs;

[0021] S15, post-processing the prediction results and outputting learning resource recommendations or learning path planning.

[0022] Furthermore, in step S11, the preprocessing includes cleaning data, converting formats, normalizing, and text processing;

[0023] Cleaning data: removing duplicate, invalid or abnormal data records;

[0024] Conversion format: convert data from different sources into a unified format and standard;

[0025] Normalization: standardize or normalize numerical data to eliminate dimensional differences;

[0026] Text processing: perform word segmentation, stop word removal, stemming, or word embedding on text data.

[0027] Furthermore, the algorithm formula for key extraction in step S12 is specifically:

[0028]

[0029] in: The user's mastery of a specific concept or knowledge point;

[0030] For user personal data The value on

[0031] is the convolution kernel weight for a specific user;

[0032] is a bias item for a specific user;

[0033] It is a personalization factor used to adjust the influence of user personal data on system output;

[0034] It is a function based on user portrait, which is used to extract personalized features of users in the process of learning advanced mathematics.

[0035] Furthermore, the convolution kernel weight criteria include: weight analysis criteria, personalized learning resource recommendation criteria and personalized learning path optimization criteria;

[0036] The weight analysis criterion performs quantitative analysis on the extracted weights to identify the weights of the weighted knowledge points;

[0037] The personalized learning resource recommendation criteria include:

[0038] 1) If the weight value of a knowledge point is large, more relevant learning materials and exercises are recommended;

[0039] If the weight value of a knowledge point is small, the recommendation of related resources will be reduced;

[0040] 2) When recommending learning resources, output a variety of learning resources, including video tutorials, text materials, and exercises;

[0041] 3) Recommend appropriate resource types based on the user’s learning preferences and habits.

[0042] Furthermore, the specific process of step 3 is as follows:

[0043] S31, recognizes the advanced mathematics related questions input by the user and converts them into a data text format that can be processed by a computer;

[0044] S32, using NLP technology to convert data text into feature vectors;

[0045] S33, using the extracted feature vector to perform similarity matching calculation with the features of each entry in the database;

[0046] S34, based on the similarity score, retrieve the knowledge points, examples and solution steps most relevant to the user's question from the database;

[0047] S35, converting the search data into a user-friendly format and outputting a matching list.

[0048] Furthermore, in step S35, the search data is converted into a user-friendly format, specifically:

[0049] Direct display: If the search results are in plain text format, they will be displayed directly;

[0050] Formatting: For text containing mathematical formulas or specific symbols, LaTeX mathematical typesetting language is used for formatting;

[0051] Paragraphing and numbering: If the solution is long, divide it into multiple paragraphs and add numbers to each paragraph;

[0052] Image generation: If the solution steps include diagrams or complex mathematical formulas, they will be displayed in the form of pictures, and the diagrams and formula pictures will be generated through the graphics drawing library;

[0053] Video link: If the solution step contains a video tutorial, the link to the relevant video will be output to the user interface;

[0054] Related video recommendations: Display other video tutorials related to the user’s current learning topic on the video playback page.

[0055] A high-level intelligent tutoring system based on personalized learning paths, deployed on a server, is used to implement the above method, including:

[0056] Data processing module: responsible for receiving personal data transmitted by users and pre-processing the collected data;

[0057] CNN model: uses convolutional and pooling layers to extract key features from preprocessed personal data and predict users’ learning needs and preferences;

[0058] Learning path design module: dynamically screen data resources and design personalized learning paths based on model prediction results;

[0059] Question answering module: recognizes the advanced mathematics questions input by users and retrieves the most relevant knowledge points, examples and answering steps in the database;

[0060] Learning progress tracking module: real-time tracking of user's learning progress, providing timely feedback and reminders.

[0061] User interaction module: provides a user-friendly interactive interface and displays the output results of other modules;

[0062] Feedback module: monitors the system's operating status and continuously optimizes the system's performance and learning path design based on users' learning data and feedback.

[0063] Furthermore, the CNN model has a built-in model training module, specifically:

[0064] Data input: receiving pre-processed data;

[0065] Feature extraction: Use the convolutional layer and pooling layer of CNN to extract key features from the input data;

[0066] The convolution layer generates feature maps through convolution operations to capture local features of the data;

[0067] The pooling layer downsamples the feature map to reduce the data dimension and retain important features;

[0068] Model parameter update: During the training process, the gradient of the model parameters is calculated according to the loss function, and the model parameters are continuously updated to gradually reduce the loss function.

[0069] Compared with the prior art, the beneficial effects of the present invention are as follows: the method of the present invention dynamically screens data resources and designs personalized learning paths according to the user's personal data and learning needs, thereby improving the pertinence and effectiveness of learning. Secondly, by using the convolutional neural network CNN model for feature extraction and prediction, it can accurately capture the user's learning style and preferences, and provide users with more accurate learning resource recommendations and answer steps. In addition, the method can also track the user's learning progress in real time, promptly remind users who are lagging behind or have poor results, and provide personalized improvement suggestions, which helps users to adjust their learning strategies in a timely manner and improve learning results. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0071] Figure 1 It is a flow chart of the tutoring method provided by the present invention.

[0072] Figure 2 It is a schematic diagram of the related question retrieval process provided by the present invention.

[0073] Figure 3 It is a schematic diagram of the user learning demand prediction process provided by the present invention. DETAILED DESCRIPTION

[0074] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0075] like Figures 1 to 3 As shown, an embodiment of the present invention provides a method and system for intelligent tutoring of advanced mathematics based on a personalized learning path, comprising the following steps:

[0076] S1, the system receives personal data transmitted by the user and performs multi-dimensional diagnosis and data analysis;

[0077] S2, dynamically screens data resources and designs personalized learning paths based on user needs;

[0078] S3, based on database retrieval, outputs the problem-solving steps and tracks the learning progress in real time;

[0079] S4, promptly reminds users who are lagging behind or experiencing poor performance, and provides personalized improvement suggestions.

[0080] Furthermore, through multi-dimensional diagnosis and data analysis, we can gain an in-depth understanding of the user's learning status and lay the foundation for personalized learning. Secondly, we can dynamically screen data resources and design personalized learning paths to make the learning content more in line with user needs and improve learning efficiency. At the same time, we can track learning progress in real time and provide answers to questions to ensure that users get timely help during the learning process.

[0081] Finally, personalized improvement suggestions are provided for users who are lagging behind or have poor results, helping them overcome learning obstacles. This method is user-centered and uses intelligent means to achieve personalized learning, effectively improving the learning effect of advanced mathematics.

[0082] Specifically, the acquisition of personal data in step S1 includes: questionnaire survey, test questions and learning history data import;

[0083] Questionnaire: Provide online questionnaires including multiple-choice, fill-in-the-blank and open-ended questions to collect users' learning experience, interests and hobbies, and learning goals;

[0084] Test questions: Provide test questions and give corresponding scores to evaluate the user's mathematical foundation, problem-solving ability and learning style;

[0085] The test questions include basic contents in multiple mathematics fields;

[0086] Importing learning history data: allows users to import learning history data, including historical grades, homework completion status, and online course learning records.

[0087] Furthermore, through questionnaires, test questions and the import of learning history data, users' personal learning data is collected comprehensively and in depth. Questionnaires help understand users' learning experience, interests and goals, test questions accurately evaluate users' mathematical foundation, problem-solving ability and learning style, and the import of learning history data provides an objective record of users' past learning performance. The comprehensive application of these data enables the system to more accurately grasp users' learning characteristics and needs, providing strong support for the design of subsequent personalized learning paths, thereby greatly improving the pertinence and effectiveness of learning.

[0088] Specifically, the specific process of step 2 is:

[0089] S11, pre-processing user personal data;

[0090] S12, using the convolutional layer and pooling layer of the CNN model, extracts key features from personal data and outputs them to the fully connected layer;

[0091] S13, applies nonlinear activation function in the fully connected layer to increase the nonlinear expression ability of the model;

[0092] S14, integrating and classifying or performing regression prediction on the extracted features to obtain a prediction result of the user's learning needs;

[0093] S15, post-processing the prediction results and outputting learning resource recommendations or learning path planning.

[0094] Furthermore, in step S2, the deep learning CNN model intelligently processes the user's personal data.

[0095] Data quality is ensured through preprocessing, and the convolutional and pooling layers of CNN efficiently extract key features. The fully connected layer is combined with nonlinear activation functions to enhance the model's expressiveness. It not only accurately captures users' learning needs and preferences, but also achieves a deep understanding of users' learning needs through integration, classification or regression prediction.

[0096] Finally, post-processing of the prediction results provides users with personalized learning resource recommendations and learning path planning, significantly improving the pertinence and efficiency of learning and promoting the development of personalized learning.

[0097] Specifically, in step S11, the preprocessing includes cleaning data, converting formats, normalizing, and text processing;

[0098] Cleaning data: removing duplicate, invalid or abnormal data records;

[0099] Conversion format: convert data from different sources into a unified format and standard;

[0100] Normalization: standardize or normalize numerical data to eliminate dimensional differences;

[0101] Text processing: perform word segmentation, stop word removal, stemming, or word embedding on text data.

[0102] Furthermore, in the S11 preprocessing step, the quality and availability of user personal data can be ensured by cleaning data, converting formats, normalizing, and text processing. Cleaning data removes redundant and abnormal information, converting formats unifies data standards, normalizing eliminates dimensional differences in numerical data, and text processing enhances the analytical value of text data.

[0103] The above preprocessing measures lay a solid foundation for subsequent data analysis and model training, effectively improve the accuracy and efficiency of data processing, ensure the scientificity and pertinence of personalized learning path design, and thus enhance the user's learning experience and effect.

[0104] Specifically, the algorithm formula for key extraction in step S12 is:

[0105]

[0106] in: The user's mastery of a specific concept or knowledge point;

[0107] For user personal data The value on

[0108] is the convolution kernel weight for a specific user;

[0109] is a bias item for a specific user;

[0110] It is a personalization factor used to adjust the influence of user personal data on system output;

[0111] It is a function based on user portrait, which is used to extract personalized features of users in the process of learning advanced mathematics.

[0112] Furthermore, The construction steps of a function based on user portrait are as follows:

[0113] First, a comprehensive user profile needs to be built, which usually includes basic information of students (such as age, gender, educational background, etc.), learning history (such as math courses taken, grades, completion of exercises, etc.), interest preferences (such as which math branches are of interest, preferred learning methods, etc.), learning speed and ability, etc. These data are collected through various methods such as students' learning records, online tests, questionnaires, and learning trajectories.

[0114] Secondly, when designing the function When extracting user profiles, it is necessary to clarify which features to extract from user portraits. These features should be able to reflect the personalized needs and characteristics of students in the process of learning advanced mathematics, such as learning difficulties, learning progress, learning interests, etc. The extraction of these features helps the system provide students with more accurate learning suggestions and recommendations.

[0115] Then, based on the above requirements, the function It is designed as a function that can process user profile data and extract personalized features. This function may contain multiple sub-functions or modules, each of which is responsible for processing one or more aspects of the user profile and extracting corresponding features.

[0116] For example, the function It may contain the following sub-functions:

[0117] Learning history processing function: Process the student's learning history data and extract features such as the math courses that have been learned, grades, and completion status of exercises;

[0118] Interest preference processing function: Process students’ interest preference data and extract features such as which branches of mathematics they are interested in and their preferred learning methods;

[0119] Learning speed and ability processing function: Process students’ learning speed and ability data to extract features such as learning speed and ability to master new knowledge;

[0120] The outputs of these sub-functions or modules will be combined into a feature vector as a function The final result is this feature vector. This feature vector will be used in subsequent recommendation, prediction or personalized suggestion models to generate personalized learning paths and resources for each student.

[0121] In practical applications, the function The design and implementation of may need to be optimized and adjusted multiple times according to the actual situation. For example, it may be necessary to add new feature extraction sub-functions or modules to process new user portrait data; it may be necessary to adjust the parameters of existing sub-functions or modules to improve the accuracy of feature extraction; it may be necessary to use more advanced machine learning algorithms or models to improve the quality of feature vectors, etc.

[0122] Specifically, the convolution kernel weight criteria include: weight analysis criteria, personalized learning resource recommendation criteria and personalized learning path optimization criteria;

[0123] The weight analysis criterion performs quantitative analysis on the extracted weights to identify the weights of the weighted knowledge points;

[0124] The personalized learning resource recommendation criteria include:

[0125] 1) If the weight value of a knowledge point is large, more relevant learning materials and exercises are recommended;

[0126] If the weight value of a knowledge point is small, the recommendation of related resources will be reduced;

[0127] 2) When recommending learning resources, output a variety of learning resources, including video tutorials, text materials, and exercises;

[0128] 3) Recommend appropriate resource types based on the user’s learning preferences and habits;

[0129] Furthermore, the determination of convolution kernel weights follows the weight analysis criteria, the personalized learning resource recommendation criteria, and the personalized learning path optimization criteria.

[0130] These criteria not only ensure the scientificity and accuracy of weight extraction, but also directly guide the recommendation of personalized learning resources. By quantitatively analyzing the weights of knowledge points, the system can intelligently recommend relevant learning materials and exercises based on the weight values, while ensuring the diversity of learning resources and meeting user learning preferences. This precise resource recommendation mechanism has greatly improved user learning efficiency and satisfaction.

[0131] Specifically, the specific process of step 3 is as follows:

[0132] S31, recognizes the advanced mathematics related questions input by the user and converts them into a data text format that can be processed by a computer;

[0133] S32, using NLP technology to convert data text into feature vectors;

[0134] S33, using the extracted feature vector to perform similarity matching calculation with the features of each entry in the database;

[0135] S34, based on the similarity score, retrieve the knowledge points, examples and solution steps most relevant to the user's question from the database;

[0136] S35, converting the search data into a user-friendly format and outputting a matching list.

[0137] Furthermore, in the detailed process of step S3, the system first accurately identifies the advanced mathematics-related questions input by the user and efficiently converts them into a data text format that can be processed by the computer. Subsequently, using advanced NLP (natural language processing) technology, these data texts are further converted into feature vectors, laying a solid foundation for subsequent similarity matching calculations.

[0138] By calculating the similarity between the user's question feature vector and the features of each item in the database, the system can quickly retrieve the knowledge points, examples and detailed answer steps that are most closely related to the user's question. Finally, these search results are carefully converted and presented in a user-friendly format, including a clear matching list, to ensure that users can easily obtain the information they need.

[0139] Furthermore, in order to continuously optimize the user experience, the system can also dynamically adjust the algorithm parameters and retrieval strategies of similarity matching based on user feedback and usage data to achieve more accurate personalized matching. At the same time, the system can also continuously update and enrich the database content to ensure that users can obtain the latest and most comprehensive higher mathematics learning resources and answers.

[0140] Specifically, in step S35, the search data is converted into a user-friendly format, specifically:

[0141] Direct display: If the search results are in plain text format, they will be displayed directly;

[0142] Formatting: For text containing mathematical formulas or specific symbols, LaTeX mathematical typesetting language is used for formatting;

[0143] Paragraphing and numbering: If the solution is long, divide it into multiple paragraphs and add numbers to each paragraph;

[0144] Image generation: If the solution steps include diagrams or complex mathematical formulas, they will be displayed in the form of pictures, and the diagrams and formula pictures will be generated through the graphics drawing library;

[0145] Video link: If the solution step contains a video tutorial, the link to the relevant video will be output to the user interface;

[0146] Related video recommendations: Display other video tutorials related to the user’s current learning topic on the video playback page.

[0147] Furthermore, in step S35, the system converts the retrieved data into a friendly format that highly matches the user's needs. For plain text results, the system displays them directly to ensure the intuitiveness of the information. For text containing mathematical formulas or specific symbols, the system uses LaTeX mathematical typesetting language to format them to improve the professionalism and clarity of reading. For longer solution steps, the system divides and numbers them into sections to make the information more organized and easy to understand.

[0148] When the solution steps include charts or complex mathematical formulas, the system uses the graphics drawing library to generate charts and formula images and displays them in the form of pictures to ensure the accuracy and readability of the information. If the solution includes a video tutorial, the system will output the relevant video link for users to watch and learn. In addition, on the video playback page, the system will also intelligently recommend other video tutorials related to the user's current learning topic, further expanding the user's learning resources and improving learning effects.

[0149] A high-level intelligent tutoring system based on personalized learning paths, deployed on a server, is used to implement the above method, including:

[0150] Data processing module: responsible for receiving personal data transmitted by users and pre-processing the collected data;

[0151] CNN model: uses convolutional and pooling layers to extract key features from preprocessed personal data and predict users’ learning needs and preferences;

[0152] Learning path design module: dynamically screen data resources and design personalized learning paths based on model prediction results;

[0153] Question answering module: recognizes the advanced mathematics questions input by users and retrieves the most relevant knowledge points, examples and answering steps in the database;

[0154] Learning progress tracking module: real-time tracking of user's learning progress, providing timely feedback and reminders.

[0155] User interaction module: provides a user-friendly interactive interface and displays the output results of other modules;

[0156] Feedback module: monitors the system's operating status and continuously optimizes the system's performance and learning path design based on users' learning data and feedback.

[0157] Specifically, the CNN model has a built-in model training module, specifically:

[0158] Data input: receiving pre-processed data;

[0159] Feature extraction: Use the convolutional layer and pooling layer of CNN to extract key features from the input data;

[0160] The convolution layer generates feature maps through convolution operations to capture local features of the data;

[0161] The pooling layer downsamples the feature map to reduce the data dimension and retain important features;

[0162] Model parameter update: During the training process, the gradient of the model parameters is calculated according to the loss function, and the model parameters are continuously updated to gradually reduce the loss function.

[0163] In summary: the method of the present invention dynamically screens data resources and designs personalized learning paths according to the user's personal data and learning needs, thereby improving the pertinence and effectiveness of learning. Secondly, by using the convolutional neural network (CNN) model for feature extraction and prediction, it can accurately capture the user's learning style and preference, and provide users with more accurate learning resource recommendations and solution steps.

[0164] In addition, this method can also track the user's learning progress in real time, promptly remind users who are lagging behind or have poor results, and provide personalized improvement suggestions, which helps users adjust their learning strategies in time and improve learning results.

[0165] In short, this method not only improves the efficiency and accuracy of advanced mathematics learning, but also brings users a more personalized and intelligent learning experience. It is an intelligent tutoring method with broad application prospects in modern education. Through continuous optimization and improvement, this method is expected to play a greater role in the future and bring convenience and progress in learning to more users.

[0166] The above embodiments are only used to help understand the method and core idea of ​​the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made to the present invention without departing from the principles of the present invention, and these improvements and modifications also fall within the scope of protection of the claims of the present invention.

Claims

1. A high-level intelligent tutoring method based on personalized learning paths, characterized in that: The steps include: S1, the system receives personal data transmitted by the user and performs multi-dimensional diagnosis and data analysis; S2, dynamically screens data resources and designs personalized learning paths based on user needs; S3, based on database retrieval, outputs the problem-solving steps and tracks the learning progress in real time; S4, promptly reminds users who are lagging behind or experiencing poor performance, and provides personalized improvement suggestions.

2. According to claim 1, a method for intelligent tutoring of advanced mathematics based on personalized learning paths is characterized by: The acquisition of personal data in step S1 includes: questionnaire survey, test questions and learning history data import; Questionnaire: Provide online questionnaires including multiple-choice, fill-in-the-blank and open-ended questions to collect users' learning experience, interests and hobbies, and learning goals; Test questions: Provide test questions and give corresponding scores to evaluate the user's mathematical foundation, problem-solving ability and learning style; The test questions include basic contents in multiple mathematics fields; Importing learning history data: allows users to import learning history data, including historical grades, homework completion status, and online course learning records.

3. The intelligent tutoring method for advanced mathematics based on personalized learning paths according to claim 1 is characterized by: The specific process of step 2 is as follows: S11, pre-processing user personal data; S12, using the convolutional layer and pooling layer of the CNN model, extracts key features from personal data and outputs them to the fully connected layer; S13, applies nonlinear activation function in the fully connected layer to increase the nonlinear expression ability of the model; S14, integrating and classifying or performing regression prediction on the extracted features to obtain a prediction result of the user's learning needs; S15, post-processing the prediction results and outputting learning resource recommendations or learning path planning.

4. The intelligent tutoring method for advanced mathematics based on personalized learning paths according to claim 3 is characterized by: In step S11, the preprocessing includes cleaning data, converting formats, normalizing and text processing; Cleaning data: removing duplicate, invalid or abnormal data records; Conversion format: convert data from different sources into a unified format and standard; Normalization: standardize or normalize numerical data to eliminate dimensional differences; Text processing: perform word segmentation, stop word removal, stemming, or word embedding on text data.

5. The intelligent tutoring method for advanced mathematics based on personalized learning paths according to claim 3 is characterized by: The algorithm formula for key extraction in step S12 is specifically: ； in: The user's mastery of a specific concept or knowledge point; For user personal data The value on is the convolution kernel weight for a specific user; is a bias item for a specific user; It is a personalization factor used to adjust the influence of user personal data on system output; It is a function based on user portrait, which is used to extract personalized features of users in the process of learning advanced mathematics.

6. The intelligent tutoring method for advanced mathematics based on personalized learning paths according to claim 5 is characterized by: The convolution kernel weight criteria include: weight analysis criteria, personalized learning resource recommendation criteria and personalized learning path optimization criteria; The weight analysis criterion performs quantitative analysis on the extracted weights to identify the weights of the weighted knowledge points; The personalized learning resource recommendation criteria include: 1) If the weight value of a knowledge point is large, more relevant learning materials and exercises are recommended; If the weight value of a knowledge point is small, the recommendation of related resources will be reduced; 2) When recommending learning resources, output a variety of learning resources, including video tutorials, text materials, and exercises; 3) Recommend appropriate resource types based on the user’s learning preferences and habits.

7. The intelligent tutoring method for advanced mathematics based on personalized learning paths according to claim 1 is characterized by: The specific process of step 3 is as follows: S31, recognizes the advanced mathematics related questions input by the user and converts them into a data text format that can be processed by a computer; S32, using NLP technology to convert data text into feature vectors; S33, using the extracted feature vector to perform similarity matching calculation with the features of each entry in the database; S34, based on the similarity score, retrieve the knowledge points, examples and solution steps most relevant to the user's question from the database; S35, converting the search data into a user-friendly format and outputting a matching list.

8. The intelligent tutoring method for advanced mathematics based on personalized learning paths according to claim 7 is characterized by: In step S35, the search data is converted into a user-friendly format, specifically: Direct display: If the search results are in plain text format, they will be displayed directly; Formatting: For text containing mathematical formulas or specific symbols, LaTeX mathematical typesetting language is used for formatting; Paragraphing and numbering: If the solution is long, divide it into multiple paragraphs and add numbers to each paragraph; Image generation: If the solution steps include diagrams or complex mathematical formulas, they will be displayed in the form of pictures, and the diagrams and formula pictures will be generated through the graphics drawing library; Video link: If the solution step contains a video tutorial, the link to the relevant video will be output to the user interface; Related video recommendations: Display other video tutorials related to the user’s current learning topic on the video playback page.

9. An intelligent tutoring system for advanced mathematics based on personalized learning paths, deployed on a server, characterized by: The system is used to implement the method described in any one of claims 1 to 8, including: Data processing module: responsible for receiving personal data transmitted by users and pre-processing the collected data; CNN model: uses convolutional and pooling layers to extract key features from preprocessed personal data and predict users’ learning needs and preferences; Learning path design module: dynamically screen data resources and design personalized learning paths based on model prediction results; Question answering module: recognizes the advanced mathematics questions input by users and retrieves the most relevant knowledge points, examples and answering steps in the database; Learning progress tracking module: real-time tracking of user's learning progress, providing timely feedback and reminders; User interaction module: provides a user-friendly interactive interface and displays the output results of other modules; Feedback module: monitors the system's operating status and continuously optimizes the system's performance and learning path design based on users' learning data and feedback.

10. The intelligent tutoring system for advanced mathematics based on personalized learning paths according to claim 9, characterized in that: The CNN model has built-in model training components, specifically: Data input: receiving pre-processed data; Feature extraction: Use the convolutional layer and pooling layer of CNN to extract key features from the input data; The convolution layer generates feature maps through convolution operations to capture local features of the data; The pooling layer downsamples the feature map to reduce the data dimension and retain important features; Model parameter update: During the training process, the gradient of the model parameters is calculated according to the loss function, and the model parameters are continuously updated to gradually reduce the loss function.

Images