Artificial intelligence adaptive education system based on big data

Abstract

The invention discloses an artificial intelligence self-adaptive education system based on big data, and relates to the field of intelligent education, the artificial intelligence self-adaptive education system comprises an education resource management module, a student data management module, a self-adaptive learning recommendation module, an auxiliary teaching decision module and a security reinforcement module, and a cloud data management library is used for classifying and integrating education resources. The method comprises the steps of establishing a knowledge graph, performing knowledge point classification on resources by constructing the knowledge graph, performing accurate matching on personal portraits of students and the knowledge graph through an incremental self-adaptive matching model, dynamically adjusting recommended learning resources in real time, analyzing collective learning problems through a big data intelligent analysis model, and generating collective teaching suggestions. A three-layer security firewall is adopted to ensure the security of the system; the problem that an existing self-adaptive education system neglects the learning progress and weak links of classes or groups is solved, and intelligent management, personalized learning recommendation and collective problem analysis of education resources can be achieved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent education, and particularly to an artificial intelligence adaptive education system based on big data. Background Art

[0002] With the progress of educational concepts and the deepening of the understanding of individual differences among learners, learners hope to obtain customized learning resources and teaching strategies according to their interests, abilities, and learning progress, quickly master the required knowledge, improve learning efficiency, and explore their own potential. However, traditional teaching methods often fail to meet the personalized learning needs of students with different learning abilities and interests, lack personalized teaching methods, and cannot give full play to the potential of students. With the continuous development of technology, the education industry is facing huge challenges of transformation, and artificial intelligence adaptive education systems are classified as one of the development directions of educational intelligence.

[0003] However, existing adaptive education systems focus on optimizing the personalized learning of individual students while ignoring the problems of collective teaching. Since the learning progress and weak links of the entire class or group are not well reflected in the adaptive education system, it is difficult for teachers to comprehensively understand the collective learning problems and weak links of the entire class or group. It is difficult for teachers to grasp the overall learning situation of the class and take corresponding measures to help the disadvantaged groups in the class overcome difficulties.

[0004] Therefore, there is a need for an artificial intelligence adaptive education system based on big data to solve the above problems. Summary of the Invention

[0005] In view of the deficiencies of the prior art, the present invention discloses an artificial intelligence adaptive education system based on big data, which realizes the intelligent management of educational resources, personalized learning recommendation, collective problem analysis, and accurate teaching decision-making support through the deep integration of big data and artificial intelligence technologies.

[0006] The present invention adopts the following technical solutions: An artificial intelligence adaptive education system based on big data, comprising: An educational resource management module, which classifies and integrates educational resources using a cloud data management library. The cloud data management library classifies and stores educational resources according to knowledge points by constructing a knowledge graph, and divides the educational resources with the same knowledge points into different difficulty levels; A student data management module, which includes a student data collection unit and a student data analysis unit. The student data collection unit is used to collect student data, and the student data analysis unit analyzes the learning styles, interest points, and learning difficulties of students based on the collected student data, and generates a personal portrait of the student; An adaptive learning recommendation module that uses an incremental adaptive matching model to adaptively match a student's personal profile with educational resources in a knowledge graph to obtain personalized learning resources for the student. The incremental adaptive matching model dynamically adjusts the recommended learning resources in real time according to the updated changes in the knowledge graph information of the student's personal profile and educational resources. An auxiliary teaching decision-making module that uses a big data intelligent analysis model to analyze and mine student data with the same learning background to obtain collective learning problems, and generates collective teaching suggestions based on the obtained collective learning problems. A security reinforcement module that realizes system security, kernel platform security, and system service security through a three-layer security firewall.

[0007] Furthermore, the cloud data management library collects educational resources from multiple data sources, and uses a natural language processor to identify knowledge points and the relationships between knowledge points in the collected educational resources. Then, the identified knowledge points and the relationships between knowledge points are stored in a structured database to form a knowledge graph. Users query the educational resources corresponding to the knowledge points in the knowledge graph through a query interface, and use an inference engine to mine the internal connections and logical relationships between knowledge points. The knowledge graph is updated incrementally in real time using an incremental window.

[0008] Furthermore, the knowledge graph includes a data preprocessing unit, a knowledge point classification unit, an entity recognition and linking unit, a relationship extraction unit, a knowledge organization and representation unit, and an incremental update unit. The data preprocessing unit cleans and organizes educational resource data through a natural language processor. The knowledge point classification unit mines the types of knowledge points in the educational resource data through a text domain mining model LDA to classify the educational resource data. The entity recognition and linking unit automatically recognizes entities in the educational resource data through a named entity recognition task, and links the entities involved in the educational resource data to the corresponding entities in the knowledge base through an entity linking task. The relationship extraction unit automatically extracts the relationships between entities in the educational resource data through a relationship extraction algorithm and stores them in the knowledge graph. The knowledge organization and representation unit stores the educational resource data in a graph database or triple storage format, and indexes and optimizes the educational resource data. The incremental update unit updates the knowledge graph in real time through a time window mechanism.

[0009] Further, the student data collection unit automatically collects the learning behavior information and learning result information of students in the online learning platform and school management system through the application programming interface, and collects the personal information, self-learning evaluation and learning psychology of students through information entry and questionnaires. The student data analysis unit extracts the student learning style, interest points and learning difficulty features by using an autoencoder, and constructs a student personal portrait through an unsupervised learning method.

[0010] Further, the working method of the incremental adaptive matching model includes the following steps: Step 1, input the initial student personal portrait feature vector , initialize the educational resource feature vector and the covariance matrix of the initial student personalized learning resource matching . The output function formula of the covariance matrix of the initial student personalized learning resource matching is: (1) In formula (1), is the initial student personal portrait feature vector, is the initial matched educational resource feature vector, represents the transpose operation of the matrix; Step 2, according to the student personal portrait feature vector at time k - 1, the educational resource feature vector in the knowledge graph at time k - 1, and the matching equation, obtain the student personalized learning resource at time (2) In formula (2), is the matched student personalized learning resource at time, is the student personal portrait feature vector at time k - 1, is the matched student personalized learning resource at time, is the covariance matrix of the student personalized learning resource matching at time k - 1, is the student personal portrait feature vector weight matrix, is the student personalized learning resource matching weight matrix, is the covariance matrix coefficient of the student personalized learning resource matching at time k; Step 3, obtain the feedback data of the student on the matched learning resource at time , and according to Feedback data of students on matching learning resources at a certain moment Update Feature vector of students' personal portraits at a certain moment And Covariance matrix of personalized learning resource matching for students at a certain moment , the update equation is expressed as: (3) In formula (3), is Feature update vector of students' personal portraits at a certain moment, is Covariance matrix of personalized learning resource matching for students at a certain moment, C is Update weight matrix of students' personal portrait features at a certain moment.

[0011] Furthermore, the big data intelligent analysis model includes an application layer, a test generation layer, a test execution layer, an adaptive control layer, a fitness function layer, and a genetic algorithm layer. The working method of the big data intelligent analysis model includes the following steps: S1. Obtain students' data, perform preprocessing operations on the obtained data, and then obtain mining parameters, constraints, and analysis objectives through the application layer. The mining parameters are students' data, the constraints are the same learning background, and the analysis objective is collective learning problems; S2. The test generation layer generates test cases for detecting collective learning problems based on students' historical data, learning background, mining parameters, constraints, and analysis objectives to evaluate students' mastery of different knowledge points; S3. The test execution layer deploys the generated test cases into the actual environment using a simulation test executor to determine each student's performance in each test dimension, record the test execution results, and obtain collective learning problems; S4. The adaptive control layer adaptively adjusts the learning path and teaching strategy according to students' performance in the test; S5. The fitness function layer evaluates the learning effect of students and the effectiveness of the teaching strategies recommended by the model by setting a fitness function. The fitness function comprehensively evaluates the learning effect of students and the effectiveness of the teaching strategies based on students' score changes, answer accuracy rates, and learning durations; S6. The genetic algorithm layer optimizes the teaching strategies and students' learning paths using selection, crossover, and mutation operations, generates new test cases for simulation test execution, and repeats the operations of S4, S5, and S6 until the optimal teaching strategies and students' learning paths are obtained.

[0012] Further, the three-layer security firewall includes an application layer firewall, a software firewall, and a hardware firewall. The application layer firewall filters and manages cloud traffic based on application protocols, and checks and controls data packets, data streams, and data content through a protocol decoder and regular expressions. The software firewall filters and manages network traffic by monitoring and controlling network connections to and from the host operating system. The hardware firewall achieves security protection by screening and filtering inbound and outbound data packets. The software firewall and the hardware firewall use an SSL secure socket layer acceleration card to reduce the load on the internal firewall server. The SSL secure socket layer acceleration card reduces the load on the internal firewall server by accelerating the processing of secure socket layer and transport layer connections.

[0013] Further, the educational resources at least include course videos, e-textbooks, teaching courseware, and e-exercises. The student data includes students' learning behavior information, learning outcome information, personal information, self-learning evaluation, and learning psychology.

[0014] The beneficial effects of the present invention are as follows: 1. Through the incremental adaptive matching model, the present invention matches the student's personal profile with educational resources in the knowledge graph, thereby realizing personalized learning resource recommendation. Students can obtain customized learning content according to their interests, learning styles, learning progress, etc., effectively improving students' learning motivation and learning effect. The incremental adaptive matching model dynamically adjusts the recommended learning resources according to the real-time update of the knowledge graph information of the student's personal profile and educational resources, ensuring the timeliness and relevance of the learning content, and avoiding the obsolescence or irrelevance of resource recommendation.

[0015] 2. The present invention adopts a big data intelligent analysis model to analyze data of students with the same learning background, and can discover collective learning problems. This enables teachers to not only pay attention to the needs of individual students, but also timely discover and solve common learning problems in the entire class or learning group. Based on big data analysis, collective teaching suggestions are provided to help teachers make accurate teaching plans and improve the overall teaching effect.

[0016] 3. The present invention classifies and integrates educational resources through a cloud data management library, and stores the resources by classifying knowledge points through a knowledge graph, ensuring more efficient organization and invocation of educational resources. The resources are divided according to difficulty levels, which helps the system provide appropriate learning materials according to the learning progress and ability of students, and avoids students accessing resources that are too simple or too difficult. By collecting and analyzing student data, the learning difficulties, interests, and learning styles of each student can be accurately identified. By generating a student's personal profile, teachers can more comprehensively understand the learning needs of students and formulate more targeted learning plans for each student.

[0017] 4. The present invention ensures the security of the system through a three - layer security firewall, covering the security of the system itself, the security of the kernel platform, and the security of system services. This helps to ensure the security of the personal data of students and teachers, prevent problems such as data leakage and cyber attacks, and improve the reliability and trust of the entire system. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a schematic diagram of the overall system architecture of the present invention; Figure 2 It is a schematic diagram of the working process of the incremental adaptive matching model in the present invention; Figure 3 It is a schematic diagram of the working process of the big data intelligent analysis model in the present invention; Figure 4 It is an architecture diagram of the three - layer security firewall in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings from... to... Figure 1 to... Figure 4 Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0020] The embodiments of the present invention disclose an artificial intelligence adaptive education system based on big data, including: An educational resource management module that manages educational resources using a cloud - based data management library to ensure the efficient storage and retrieval of educational resources. By constructing a knowledge graph, educational resources are classified according to knowledge points, and resources with the same knowledge points are divided into difficulty levels. These resources include courseware, videos, exercises, articles, cases, etc., and all these contents are organized according to the hierarchical system of knowledge points, facilitating students to switch between different difficulty levels. The cloud - based data management library is implemented using a cloud platform (such as AWS, Azure, or Google Cloud), and a database system such as a NoSQL database (e.g., MongoDB) or a graph database (e.g., Neo4j) is used to store and manage educational resources. The construction of the knowledge graph uses natural language processing (NLP) technology to analyze contents such as textbooks, lecture notes, and exercise sets, automatically extract key concepts and knowledge points, and construct the knowledge graph. And through machine learning methods, the relationships and hierarchies in the graph are continuously optimized. By using big data analysis of students' learning behaviors and performance feedback, difficulty labels are set for each educational resource, and the difficulty level is dynamically adjusted.

[0021] Student Data Management Module. Through the student data collection unit, the system can automatically or manually collect students' learning data, including grades, learning duration, learning frequency, homework submission status, exam performance, etc. The analysis unit then uses machine learning algorithms (such as clustering, regression analysis, etc.) to analyze students' learning styles, interests, and learning difficulties, and then generates personalized learning portraits of students. The student data collection unit collects students' behavioral data, such as video watching duration, number of discussions participated in, homework submission status, etc., through learning platforms, online classrooms, learning management systems (LMS), etc. The student data analysis unit uses data mining and machine learning technologies (such as clustering analysis, deep learning, etc.) to analyze students' learning data, identify their learning patterns, preferences, interests, strengths, and weaknesses, so as to create personal learning portraits. Based on the collected data, a personal portrait of the student will be generated, which will include learning styles (such as visual, auditory, etc.), interests (such as interests in a certain subject area), and weak links (for example, the mastery of certain knowledge points is not high).

[0022] Adaptive Learning Recommendation Module. Through the incremental adaptive matching model, the system matches the personal portrait of the student with the knowledge graph of educational resources to recommend personalized learning resources. The recommendation system dynamically updates the learning content according to the students' learning needs, interests, and progress, and adjusts the recommendation strategy to ensure the accuracy and timeliness of the learning content. The incremental adaptive matching model adopts recommendation algorithms based on deep learning (such as collaborative filtering, content-based recommendation, etc.), and continuously updates the personal portrait of the student and the knowledge graph of educational resources to adjust the recommended learning resources in real time. According to information such as the difficulties encountered by students in the learning process, error rates, and progress speeds, the system will dynamically adjust the recommendation strategy and push new learning content or review materials. The system recommends the most suitable learning resources, such as videos, graphic textbooks, practice questions, interactive discussions, etc., according to the students' learning progress, interests, and learning styles.

[0023] Auxiliary Teaching Decision-making Module. Based on the big data intelligent analysis model, it analyzes the learning situations of students with the same learning background (such as the same grade, the same course, etc.), and mines collective learning problems. These collective problems can be uneven mastery of knowledge points, high error rates for certain questions, etc. Based on these collective problems, teaching suggestions are generated to provide decision-making support for teachers. Methods such as clustering analysis and association rule mining are used to identify the problems commonly existing in the student group, such as high difficulty of certain knowledge points or high error rates for certain types of questions. Through data mining algorithms, collective teaching suggestions based on big data analysis are generated, such as the need to adjust teaching focuses, strengthen the review of certain knowledge points, and increase specific types of practice questions.

[0024] Security reinforcement module. To ensure the security of the system, a three-layer security firewall is designed to protect the system itself, the kernel platform, and system services respectively, ensuring the security of each layer and preventing data leakage and attacks. The three-layer security firewall takes different security measures at different levels. For example, at the network level, traditional firewalls and intrusion detection systems (IDS) are used for monitoring; at the application level, all transmitted data is encrypted and authenticated; at the data storage level, sensitive data is encrypted for storage. To ensure the security of the underlying operating system, database, middleware, and other platforms of the system, automated security vulnerability detection tools are used to repair system vulnerabilities in a timely manner. To ensure the security of cloud services and third-party interfaces, an API gateway is used for access control to ensure that only authorized users can access data and services.

[0025] As shown in the Figure 1 appendix, the output end of the educational resource management module is connected to the input end of the adaptive learning recommendation module. The output end of the student data management module is connected to the input end of the adaptive learning recommendation module and the input end of the auxiliary teaching decision-making module. The output end of the adaptive learning recommendation module is connected to the input end of the auxiliary teaching decision-making module. The security reinforcement module is connected to the entire system to ensure the security of the system.

[0026] The system combines technologies such as big data, artificial intelligence, adaptive learning, and security reinforcement, and can achieve personalized learning recommendations, dynamic teaching decision support, and comprehensive security guarantees. By intelligently analyzing student data, generating a student personal profile and matching it with educational resources, it can provide a customized learning plan; at the same time, with the help of big data analysis, it helps teachers improve teaching methods and enhance the overall education quality; the security reinforcement measures ensure the stable and secure operation of the system.

[0027] The cloud data management library first collects educational resources from multiple data sources (such as textbooks, academic papers, online courses, educational websites, etc.). These resources can be structured data (such as database tables, CSV files) or unstructured data (such as articles, videos, lecture notes, etc.). The collected educational resources are preprocessed, including removing redundant data, correcting format problems, and extracting key information (such as keywords, titles, abstracts, etc.).

[0028] Use natural language processing (NLP) technology to analyze the text in educational resources and identify key knowledge points. This typically includes part-of-speech tagging, named entity recognition (NER), dependency parsing, etc. Based on the identification of knowledge points, use relationship extraction algorithms to identify the logical relationships and hierarchical structures between various knowledge points. For example, identify the relationship between "mathematical formulas" and "application problems", or the association between "concept definitions" and "example applications". Combine the context and use deep learning models (such as BERT, GPT, etc.) to deeply understand the semantics in educational resources to ensure accurate extraction of content related to knowledge points.

[0029] Take the identified knowledge points as nodes and the relationships between knowledge points (such as causal relationships, inclusion relationships, super-subordinate relationships, etc.) as edges to construct a knowledge graph. Each node in the knowledge graph represents an independent knowledge point, and the edges represent the semantic connections between these knowledge points. Store the constructed knowledge graph in a structured database. Commonly used databases include graph databases (such as Neo4j, ArangoDB, etc.) or relational databases (such as PostgreSQL), etc. These databases support efficient query and update operations. Provide a query interface for users. Through the graph database or other storage systems, users can query the node information in the knowledge graph and obtain educational resources related to these knowledge points. Queries can be simple text queries or complex graph-based queries. Users can obtain information such as the definitions, applications, and examples of specific knowledge points through queries.

[0030] The inference engine discovers potential logical associations and internal connections by analyzing the knowledge points in the knowledge graph and their relationships. For example, the inference engine can discover hidden connections between certain knowledge points (such as the relationship between mathematical theorems and application problems) and push new educational resources based on these associations. The inference engine can also continuously optimize and update the knowledge points and relationships in the knowledge graph according to the user's interactions and feedback, further improving the accuracy and integrity of the knowledge graph.

[0031] The knowledge graph uses an incremental window mechanism to perform incremental updates on the data source periodically. This means that the system will identify and process newly added educational resources and add new knowledge points and relationships to the existing knowledge graph without having to reconstruct the entire graph. To maintain the timeliness of the knowledge graph, the system adopts a real-time update mechanism. Whenever new educational resources are collected, the system will automatically identify the newly added knowledge points and update the knowledge graph in real time to ensure that users can access the latest educational resources. Use a structured database (such as MySQL) and a graph database (such as Neo4j) to store the nodes and relationships in the knowledge graph. The database needs to have high scalability and high-performance query capabilities.

[0032] The implementation process of the cloud data management library ensures that the knowledge graph can be updated in real time and continuously optimized from data collection, preprocessing to knowledge graph construction, querying, reasoning, and then to the incremental update mechanism, so as to provide users with accurate educational resources and reasoning services. Through such a system, users can efficiently obtain the required knowledge, and at the same time, potential knowledge associations can be mined through the reasoning engine to promote learning and in-depth understanding of knowledge.

[0033] The construction process of the knowledge graph includes multiple functional modules. Each module is responsible for specific data processing tasks and works together to finally form an efficient and dynamically updated educational resource knowledge graph. The following is a detailed description of the functions of each module: The data preprocessing unit is responsible for cleaning, formatting, removing noise, correcting errors, etc. of the original educational resource data to ensure data quality. Natural language processing technologies (such as word segmentation, stop word removal, syntactic analysis, etc.) are used to prepare for the subsequent construction of the knowledge graph. For example, cleaning irrelevant information in the text, formatting the text, and removing meaningless words.

[0034] The knowledge point classification unit classifies the educational resource data according to the content in it and divides it into different knowledge point types. Text mining technologies such as the LDA (Latent Dirichlet Allocation) model are used to automatically analyze the topics and content in the text data, identify different types of knowledge points, and classify the data into corresponding categories. LDA can extract topics from a large amount of educational texts and assign category labels to different educational resources.

[0035] The entity recognition and linking unit is responsible for automatically identifying relevant entities (such as people, places, times, events, etc.) from the educational resource data and linking these entities to the existing knowledge base to ensure data accuracy and context consistency. Named entity recognition (NER) technology is used to extract meaningful entities from the text, such as book names, authors, schools, courses, etc. By matching the entities in the educational resource data with the entities in the knowledge base, the accuracy of the link is ensured. Commonly used technologies include entity alignment methods based on similarity matching or deep learning models.

[0036] The relationship extraction unit automatically identifies the relationships between different entities in the educational resource data (such as "teachers teach courses", "students learn knowledge points") and extracts these relationships and stores them in the knowledge graph. Relationship extraction algorithms (such as rule-based, supervised learning, or deep learning methods) are used to mine the association information between entities from the text. The relationships may be simple (such as "professor - course") or complex (such as "student - exam score - course"), and relationship extraction needs to consider context and semantic information.

[0037] The knowledge organization representation unit organizes and stores all educational resource data and the extracted relationships, forms a structured knowledge graph, and indexes and optimizes the data for efficient querying and accessing. It uses a graph database (such as Neo4j) or triple storage (such as RDF format) to represent entities and relationships as nodes and edges, forming a graph structure. Each node represents an entity, and each edge represents the relationship between entities. By indexing the nodes and edges in the knowledge graph, the query efficiency can be improved. At the same time, the data can be optimized to reduce storage overhead and query latency.

[0038] The incremental update unit ensures that the knowledge graph can be continuously updated as new educational resource data arrives and maintains the latest knowledge through a real-time incremental update mechanism. It uses a time window mechanism (such as regular updates or event-triggered updates) to perform incremental updates on the knowledge graph. Whenever new data flows in, only the new data needs to be processed, and the newly identified entity, relationship and other information are merged with the existing knowledge graph. Issues such as data consistency, deduplication, and efficient incremental reasoning need to be considered during the update process to ensure that the update of the graph does not affect the correctness of the existing data.

[0039] Implementation steps: First, collect data from different educational resources (such as textbooks, academic papers, online courses, Q&A, etc.) and clean it. Then apply text mining methods such as LDA to classify the cleaned data into different knowledge point types. Then use named entity recognition technology to identify entities and match these entities with the existing knowledge base through entity linking. Then analyze the relationships between entities through relationship extraction algorithms and convert them into edges in the graph database. In addition, use a graph database or triple storage format to construct a knowledge graph and optimize the storage and query methods. Finally, update the knowledge graph regularly or in real time to process new educational resource data and ensure that the knowledge graph always remains up-to-date.

[0040] Through this multi-level processing method, the system can effectively extract structured knowledge from a large amount of educational resources, organize and represent it in the form of a graph, and finally achieve efficient knowledge querying and recommendation.

[0041] In the student data management module, the student data collection unit performs API integration with online learning platforms (such as MOOCs, online homework systems, online test platforms, etc.) and school management systems (such as student information systems, course management systems, etc.). It automatically collects students' learning behavior information (such as login time, learning duration, course access frequency, homework submission status, etc.) and learning result information (such as test scores, course grades, etc.) through the API. And collect students' personal information (such as name, age, grade, class, etc.), self-learning evaluation (such as learning satisfaction, learning difficulties, etc.) and learning psychology (such as learning motivation, learning attitude, etc.) through input information and questionnaires: The student data analysis unit performs preprocessing operations such as cleaning, deduplication, and normalization on the collected student data to improve data quality and the accuracy of feature extraction. An autoencoder model is constructed using deep learning frameworks (such as TensorFlow, PyTorch, etc.). The preprocessed student data is used as training samples to train the autoencoder so that it can learn effective representations of features such as students' learning styles, interest points, and learning difficulties. After training, the autoencoder is used to extract features from new student data, obtaining feature vectors that characterize students' learning styles, interest points, and learning difficulties. Clustering algorithms such as K-means and DBSCAN are used to perform clustering analysis on the extracted feature vectors. According to the clustering results, students are divided into different groups, and each group has similar features such as learning styles, interest points, and learning difficulties. Combining the clustering analysis results with students' basic information, questionnaire survey data, etc., a personal portrait of each student is constructed. The personal portrait should include multiple dimensions such as students' learning styles, interest points, learning difficulties, and basic information to provide support for subsequent personalized learning recommendations and auxiliary teaching decision-making. As students' learning behaviors continue and new data is continuously collected, the personal portraits of students are updated in real time. Ensure that the portraits can accurately reflect students' current learning status and needs.

[0042] Through the above implementation steps, the learning data of students can be effectively collected and analyzed, and accurate and comprehensive personal portraits of students can be constructed, providing strong support for subsequent personalized learning recommendations and auxiliary teaching decision-making. At the same time, we should also pay attention to data privacy protection and security to ensure that students' personal information is not leaked and misused.

[0043] The working method of the incremental adaptive matching model is as shown in Figure 2 the appendix and includes the following steps: Step 1: Initialize the student personal portrait feature vector, educational resource feature vector, and covariance matrix Input the initial student personal portrait feature vector , initialize the educational resource feature vector and the covariance matrix for the initial student personalized learning resource matching . The initial student personal portrait feature vector represents information such as the initial learning status, interests, and abilities of students. The initialized educational resource feature vector represents the characteristics of educational resources, such as difficulty, theme, type, etc. The covariance matrix for the initial student personalized learning resource matching describes the correlation between the student personal portrait feature vector and the educational resource feature vector. The output function formula of the covariance matrix for the initial student personalized learning resource matching is as follows: (1) In formula (1), is the initial student personal portrait feature vector, is the initial matched educational resource feature vector, represents the transpose operation of the matrix; Step 2, calculate the personalized learning resources at the current moment through the matching equation According to the student personal portrait feature vector at time k-1 , the educational resource feature vector in the knowledge graph at time k-1 and the matching equation, obtain the student personalized learning resources at time , and the matching equation is expressed as: (2) In formula (2), is the matched student personalized learning resources at time is the student personal portrait feature vector at time k-1, is the matched student personalized learning resources at time is the covariance matrix matched by the student personalized learning resources at time k-1, is the student personal portrait feature vector weight matrix, is the student personalized learning resource matching weight matrix, is the covariance matrix coefficient of the student personalized learning resource matching at time k; Formula (2) is used to calculate the student personalized learning resources at time k. This equation combines the student personal portrait, educational resource features, matched covariance matrix, and weight matrix at time k-1 to calculate the personalized learning resources at the current moment. Note that the weight matrices (student personal portrait feature vector weight matrix and student personalized learning resource matching weight matrix) and covariance matrix coefficient in formula (2) need to be set or learned according to the actual situation; Step 3, obtain the feedback of the student on the matched learning resources, and update the personal portrait feature vector and the covariance matrix of the learning resource matching Obtain the feedback data of the student on the matched learning resources at time , and according to the feedback data of the student on the matched learning resources at time update the student personal portrait feature vector at time and the covariance matrix of the student personalized learning resource matching at time , and the update equation is expressed as: (3) In formula (3), be the updated vector of the student's personal portrait features at time be the covariance matrix of the personalized learning resource matching of the student at time , and C is the update weight matrix of the student's personal portrait feature vector at time . Formula (3) is used to update the student's personal portrait feature vector and the matching covariance matrix at time . This equation takes into account the student's feedback on learning resources and the update weight matrix C. The updated student's personal portrait feature vector will better reflect the student's current learning state, while the updated covariance matrix will more accurately describe the correlation between the student's personal portrait feature vector and the educational resource feature vector. The updated student's personal portrait feature vector and the matching covariance matrix are used as the input for the next iteration, and steps 2 and 3 are continued until a predetermined number of iterations is reached or a certain convergence condition is met.

[0044] Through the implementation of the above steps, the incremental adaptive matching model can dynamically adjust the student's personal portrait and matching strategy according to the student's feedback and the characteristics of learning resources, so as to achieve personalized learning resource recommendation.

[0045] The big data intelligent analysis model includes an application layer, a test generation layer, a test execution layer, an adaptive control layer, a fitness function layer, and a genetic algorithm layer. The working method of the big data intelligent analysis model is as shown in Figure 3 and includes the following steps: S1: Obtain student data and perform preprocessing, and set analysis objectives, mining parameters, and limiting conditions Collect relevant data of students from multiple sources. This data may include students' basic information, learning records, grade data, classroom interaction data, etc. Perform preprocessing on the collected data, including steps such as missing value filling, data cleaning, and data standardization, to ensure data quality and facilitate subsequent analysis.

[0046] S2: Generate test cases based on students' historical data and background, and evaluate collective learning problems By analyzing students' historical data, including their learning grades, learning backgrounds (such as course participation, exam scores, etc.), and known learning bottlenecks, infer students' learning patterns and problems. According to students' historical data and preset analysis objectives (such as collective learning problems), the test generation layer generates test cases related to these objectives. The design of test cases will focus on evaluating students' mastery of each knowledge point to help discover collective learning problems.

[0047] S3: Execute test cases in the actual environment, record students' performance, and obtain collective learning problem data The generated test cases will be deployed to the actual teaching environment through a simulated test executor. The executor here may be an online examination system, an interactive teaching platform, etc. The executor will record each student's performance in the test, including data such as their scores in various test dimensions, the correctness of their answers, and the duration of answering questions. Through these execution results, the system can identify collective learning problems, that is, knowledge gaps or learning difficulties commonly existing among the student group.

[0048] S4: Adjust the learning path and teaching strategies according to the students' test performance Based on the test results, the adaptive control layer will adjust the learning path according to each student's specific performance, and recommend learning content and difficulty suitable for them to the students. The control layer will also dynamically adjust the teaching strategies, such as changing teaching methods, increasing or decreasing interaction sessions, providing more personalized tutoring, etc., to improve the students' learning effect.

[0049] S5: Evaluate the students' learning effect and the effectiveness of teaching strategies through a fitness function The fitness function layer evaluates the students' learning effect by setting a series of evaluation criteria (such as students' score changes, answer correctness rates, and learning durations, etc.). In addition to evaluating the students' learning effect, the fitness function will also evaluate the effectiveness of the adjusted teaching strategies to ensure that they have a positive effect on the students' learning progress.

[0050] S6: Use the genetic algorithm to optimize teaching strategies and learning paths, generate new test cases, and iterate repeatedly until the optimal result is obtained The genetic algorithm simulates the process of natural selection and adopts operations such as selection, crossover, and mutation to continuously optimize teaching strategies and students' learning paths. For example, those most effective strategies can be selected and crossed and mutated to produce new and improved strategies. The genetic algorithm will also generate new test cases and deploy them to the teaching environment to further evaluate the impact of the optimized teaching strategies on students' learning. By continuously executing adaptive control, fitness evaluation, and genetic algorithm optimization in a loop, the most suitable teaching strategies and learning paths for the student group can be finally obtained to solve collective learning problems.

[0051] The big data intelligent analysis model collaborates through different levels, from data acquisition to genetic algorithm optimization. The ultimate goal is to discover and solve problems in collective learning. By dynamically adjusting test cases and optimizing strategies, the detection accuracy can be gradually improved, helping educators identify knowledge weaknesses or behavioral deviations in group learning, and thus making targeted improvement measures.

[0052] The three - layer security firewall system includes an application - layer firewall, a software firewall, and a hardware firewall. Each layer has unique functions and roles, and together they provide multi - level security protection for the network.

[0053] An application-layer firewall is located at the application layer of the OSI model and is specifically responsible for managing and filtering network traffic based on application protocols. It can identify and process application-layer protocols (such as HTTP, FTP, SMTP, etc.) and use protocol decoders and regular expressions to check the data stream and packet content. By parsing the application-layer protocols with protocol decoders, it can detect and filter potential malicious packets. The application-layer firewall can detect deeper types of attacks, such as SQL injection, cross-site scripting (XSS), etc. Regular expressions are used to match and identify specific malicious patterns or malicious code, thus achieving more refined traffic screening.

[0054] A software firewall runs within the host operating system and monitors and controls inbound and outbound network traffic in real time. It can filter incoming and outgoing data according to predefined rules. The software firewall monitors and manages the communication between the host and the outside world, controlling which applications can access the network and which connections need to be blocked. The software firewall can restrict illegal inbound requests from the outside or prevent malicious traffic from being sent by internal host applications. It is usually integrated into the operating system and can utilize the kernel resources of the operating system to achieve efficient traffic management.

[0055] A hardware firewall is an independent hardware device, usually deployed at the network edge. It provides security protection by screening and filtering packets entering and leaving the network. The hardware firewall mainly relies on packet filtering and state detection to determine whether a packet is allowed to pass. The hardware firewall analyzes the packets in the network and filters the data stream based on preset rules (such as source IP address, destination port, etc.). The hardware firewall can monitor the state of connections, ensure that the data stream complies with the security policy, and prevent illegal access. As the first line of defense for the network, it can effectively block the intrusion of external attackers.

[0056] When dealing with encrypted traffic, the firewall needs to parse the SSL / TLS-encrypted data stream. The introduction of an SSL acceleration card can significantly improve the efficiency of processing these encrypted connections and reduce the burden on the internal servers of the firewall. The SSL acceleration card is specifically used to accelerate the decryption and encryption processes of SSL / TLS encrypted connections and can significantly improve the processing speed of encrypted traffic. Since the main processor of the firewall can hand over the SSL decryption task to the acceleration card for processing, the load on the internal servers of the firewall is reduced, thereby improving the overall network security performance and efficiency.

[0057] Specific implementation process: First, the application layer firewall conducts in-depth analysis on the incoming traffic to check for the existence of malicious code or attack patterns. At this time, it screens the data according to protocols and regular expressions. At the host level, the software firewall monitors all network connections according to system rules and intercepts traffic that does not conform to the security policy. The hardware firewall is deployed at the network edge and directly processes the data streams entering and leaving the network. It ensures that only legitimate traffic can enter the internal network through packet filtering and state detection. If the traffic is encrypted (such as HTTPS), the SSL acceleration card intervenes to decrypt the data stream and then passes it to the firewall for further analysis and filtering. This can avoid overloading the internal servers of the firewall when processing encrypted traffic.

[0058] As shown in the Figure 4 accompanying drawings, through the collaborative work of these three layers of firewalls, the entire system can efficiently manage network traffic, prevent malicious attacks, maintain system performance, and at the same time ensure the secure processing of encrypted data.

[0059] The educational resources at least include course videos, e-textbooks, teaching courseware, and e-exercises. The student data includes students' learning behavior information, learning result information, personal information, self-learning evaluation, and learning psychology. The learning behavior information at least includes learning progress, learning frequency, learning duration, question interaction frequency, and homework submission frequency. The learning result information includes assessment results and homework completion quality. The students' personal information includes students' gender, age, hobbies, and learning goals. The self-learning evaluation and learning psychology include self-learning evaluation, learning pressure, and academic difficulties.

[0060] So far, the technical solution of the present invention has been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the protection scope of the present invention.

Claims

1. An artificial intelligence adaptive education system based on big data, characterized in that: include: An educational resource management module, which uses a cloud data management library to classify and integrate educational resources. The cloud data management library classifies and stores educational resources according to knowledge points by constructing a knowledge graph, and divides educational resources with the same knowledge points into difficulty levels; A student data management module, the student data management module includes a student data collection unit and a student data analysis unit, the student data collection unit is used to collect student data, the student data analysis unit analyzes the student's learning style, interests and learning difficulties based on the collected student data, and generates a student personal portrait; An adaptive learning recommendation module, which uses an incremental adaptive matching model to adaptively match student profiles with educational resources in the knowledge graph to obtain personalized learning resources for students. The incremental adaptive matching model dynamically adjusts the recommended learning resources in real time according to the updated changes in the student profiles and the knowledge graph information of educational resources; An auxiliary teaching decision-making module, which uses a big data intelligent analysis model to analyze and mine student data with the same learning background to obtain collective learning problems, and generates collective teaching suggestions based on the obtained collective learning problems; A security reinforcement module implements system security, kernel platform security and system service security through a three-layer security firewall.

2. According to claim 1, an artificial intelligence adaptive education system based on big data is characterized in that: The cloud-based data management library collects educational resources from multiple data sources, and uses a natural language processor to identify knowledge points and the relationships between knowledge points in the collected educational resources. Then, the identified knowledge points and the relationships between knowledge points are stored in a structured database to form a knowledge graph. Users query the educational resources corresponding to the knowledge points in the knowledge graph through a query interface, and use the reasoning engine to mine the intrinsic connections and logical relationships between knowledge points. The knowledge graph uses an incremental window for real-time incremental updates.

3. The artificial intelligence adaptive education system based on big data according to claim 2 is characterized in that: The knowledge graph includes a data preprocessing unit, a knowledge point classification unit, an entity recognition and linking unit, a relationship extraction unit, a knowledge organization representation unit and an incremental update unit. The data preprocessing unit cleans and organizes the educational resource data through a natural language processor. The knowledge point classification unit mines the types of knowledge points in the educational resource data through the text domain mining model LDA, and classifies the educational resource data. The entity recognition and linking unit automatically recognizes entities in the educational resource data through a named entity recognition task, and links the entities involved in the educational resource data to the corresponding entities in the knowledge base through an entity linking task. The relationship extraction unit automatically extracts the relationship between entities in the educational resource data through a relationship extraction algorithm, and stores it in the knowledge graph. The knowledge organization representation unit stores the educational resource data in a graph database or a triple storage format, and indexes and optimizes the educational resource data. The incremental update unit updates the knowledge graph in real time through a time window mechanism.

4. The artificial intelligence adaptive education system based on big data according to claim 1 is characterized in that: The student data collection unit automatically collects the learning behavior information and learning results information of students in the online learning platform and the school management system through the application program interface, and collects the personal information, self-learning evaluation and learning psychology of students through input information and questionnaires. The student data analysis unit uses an autoencoder to extract the characteristics of students' learning style, interests and learning difficulties, and constructs a personal portrait of students through unsupervised learning.

5. The artificial intelligence adaptive education system based on big data according to claim 1, characterized in that: The working method of the incremental adaptive matching model comprises the following steps: Step 1: Input the initial student profile feature vector , initialize the educational resource feature vector And the covariance matrix of the initial student personalized learning resource matching , the covariance matrix of the initial student personalized learning resource matching The output function formula is: （1） In formula (1), is the initial student personal portrait feature vector, is the initial matching educational resource feature vector, Represents the transpose operation of a matrix; Step 2: Based on the student's personal portrait feature vector at time k-1 , the educational resource feature vector in the knowledge graph at time k-1 and matching equations, we get Personalized learning resources for students at all times , the matching equation is expressed as: （2） In formula (2), for Personalized learning resources for students that match every moment. is the student’s personal portrait feature vector at time k-1, for Personalized learning resources for students that match every moment. is the covariance matrix of personalized learning resource matching for students at time k-1, is the weight matrix of the student's personal portrait feature vector, Match the weight matrix for students’ personalized learning resources. is the covariance matrix coefficient of personalized learning resource matching for students at time k; Step 3: Get Students’ feedback data on matching learning resources at all times , and according to Students’ feedback data on matching learning resources at all times renew Characteristic vector of student's personal portrait at each moment and Covariance matrix of personalized learning resource matching for students at every moment , the update equation is expressed as: （3） In formula (3), for The student's personal portrait feature update vector at all times, for The covariance matrix of the personalized learning resource matching of students at each moment, C is Update the weight matrix of the student’s personal portrait features at all times.

6. The artificial intelligence adaptive education system based on big data according to claim 1 is characterized in that: The big data intelligent analysis model includes an application layer, a test generation layer, a test execution layer, an adaptive control layer, a fitness function layer and a genetic algorithm layer. The working method of the big data intelligent analysis model includes the following steps: S1. Obtain student data and perform preprocessing operations on the acquired data, and then obtain mining parameters, constraints and analysis targets through the application layer. The mining parameters are student data, the constraints are the same learning background, and the analysis target is collective learning problems; S2, the test generation layer generates test cases for detecting collective learning problems based on students' historical data, learning background, mining parameters, constraints and analysis goals to evaluate students' mastery of different knowledge points; S3, the test execution layer uses a simulated test executor to deploy the generated test cases into the actual environment to determine the performance of each student on each test dimension and record the test execution results to obtain collective learning problems; S4, the adaptive control layer adaptively adjusts the learning path and teaching strategy according to the students’ performance in the test; S5, the fitness function layer evaluates the learning effect of students and the effectiveness of the teaching strategy recommended by the model by setting the fitness function. The fitness function comprehensively evaluates the learning effect of students and the effectiveness of the teaching strategy based on the changes in students' grades, the accuracy of answering questions and the learning time. S6, the genetic algorithm layer uses selection, crossover and mutation operations to optimize the teaching strategy and student learning path, and generates new test cases for simulation test execution, repeating S4, S5 and S6 operations until the optimal teaching strategy and student learning path are obtained.

7. The artificial intelligence adaptive education system based on big data according to claim 1 is characterized in that: The three-layer security firewall includes an application layer firewall, a software firewall and a hardware firewall. The application layer firewall filters and manages cloud traffic based on the application protocol, and inspects and controls data packets, data streams and data content through protocol decoders and regular expressions. The software firewall filters and manages network traffic by monitoring and controlling network connections in and out of the host operating system. The hardware firewall implements security protection by screening and filtering inbound and outbound data packets. The software firewall and hardware firewall use SSL secure socket layer acceleration cards to reduce the load of the firewall's internal server. The SSL secure socket layer acceleration card reduces the load of the firewall's internal server by accelerating the processing of secure socket layer and transport layer connections.

8. The artificial intelligence adaptive education system based on big data according to claim 1, characterized in that: The educational resources include at least course videos, electronic teaching materials, teaching courseware and electronic exercises, and the student data includes students' learning behavior information, learning results information, personal information, self-learning evaluation and learning psychology.

Images