Method and system for constructing student mutual assistance network and dividing groups based on knowledge graph
Through Bayesian knowledge tracking model and graph convolutional neural network, the student mutual aid network is constructed, and combined with Manhattan distance and k-means clustering, the problem of difficult to construct student mutual aid relationship and low group division accuracy in existing technology is solved, and dynamic and accurate learning group division and mutual aid effect are improved.
Patent Information
- Application Number
- CN202510699612.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-28
- Publication Date
- 2025-09-02
- Estimated Expiration
- 2045-05-28
AI Technical Summary
The existing learning status assessment and student group division methods cannot effectively construct mutual assistance relationships between students, and lack of in-depth analysis, resulting in unsatisfactory mutual assistance results and low group division accuracy, so it is impossible to dynamically respond to changes in learning states.
The Bayesian knowledge tracking model is used to calculate the degree of students' mastery of knowledge points, combine Manhattan distance to build a mutual aid network, and use graph convolutional neural network to mine students' features, and group division is performed through the k-means clustering method.
Dynamically build mutual aid relationships and deeply explore students' characteristics, improve the scientificity and adaptability of learning group division, ensure the consistency of internal characteristics of the group, and improve the efficiency and flexibility of mutual aid learning.
Smart Images
Figure CN120217031B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of knowledge graph technology, and in particular relates to a method and system for constructing a student mutual assistance network and dividing groups based on a knowledge graph. Background Art
[0002] With the development of educational informatization, online learning platforms provide students with rich learning resources and personalized learning experiences. They also accumulate a large amount of student learning behavior data, such as answer time, submission time, and correctness / failure rate. How to effectively utilize this data to assess students' learning status, establish dynamic and accurate student support networks, and further group students with different learning characteristics to achieve data-driven personalized education has become a key research direction in the field of education.
[0003] Currently, existing learning status assessment and student grouping methods have the following technical problems:
[0004] 1. Difficulty in effectively building knowledge gaps and mutual assistance relationships among students: Current approaches to building mutual assistance relationships often focus on differences in knowledge mastery, lack multi-level data analysis of student learning behaviors, and fail to effectively explore knowledge gaps among students. Simple mutual assistance relationships make it difficult to form a reasonable learning network and establish targeted mutual assistance relationships between students with complementary learning needs, thus limiting knowledge sharing and mutual learning among students.
[0005] 2. Lack of in-depth analysis of student characteristics: Traditional clustering methods are mostly based on static characteristics or superficial learning behavior characteristics of students, failing to deeply explore the underlying characteristics of students in the dynamic learning process. However, the characteristics of students in a student mutual assistance network depend not only on their own knowledge acquisition but also on mutual assistance relationships and network structure. Therefore, clustering methods based solely on superficial characteristics are difficult to accurately divide into groups, resulting in inconsistent characteristics among students within learning groups and suboptimal mutual assistance results.
[0006] 3. Low accuracy in dynamic learning group division: Due to the lack of dynamic analysis of students’ knowledge structure, mastery level, and mutual assistance relationships during the learning process, traditional group division methods cannot respond to changes in students’ learning status in real time. The accuracy and flexibility of group division are insufficient, making it difficult to form optimal learning groups. Summary of the Invention
[0007] To solve the above technical problems, the present invention provides a method for constructing a student mutual assistance network and dividing groups based on a knowledge graph, comprising the following steps:
[0008] include:
[0009] Step S1: Obtain student answer data from the online learning platform or smart learning system to build a student learning profile;
[0010] Step S2: Using the Bayesian knowledge tracking model, calculate the student's mastery of each knowledge point based on the student's answer record and form a knowledge point mastery matrix;
[0011] Step S3: Based on the knowledge point mastery matrix, Manhattan distance is used to calculate the knowledge point mastery differences between students to construct a knowledge point mastery difference matrix; and a student mutual assistance network is constructed based on the knowledge point mastery difference matrix;
[0012] Step S4: Based on the student mutual assistance network, a graph convolutional neural network is used to fully mine student characteristics, and the k-means clustering method is used to divide students into groups to form learning groups with different characteristics.
[0013] Beneficial effects:
[0014] 1. Effectively build dynamic mutual aid relationships: This paper uses a Bayesian knowledge tracking model combined with a knowledge graph to dynamically calculate students' mastery of each knowledge point. It also uses Manhattan distance to calculate knowledge differences between students, automatically forming mutual aid relationships among students with complementary learning needs. This allows for a more efficient construction of a student mutual aid network, supporting mutual learning and knowledge sharing among students.
[0015] 2. Deeper student feature mining: This paper uses graph convolutional neural networks to perform deep feature extraction on the student mutual aid network, representing the features of students in the student mutual aid network as vectors, which helps to capture the potential features of each student in the network and thus improve the accuracy of student feature mining.
[0016] 3. The division of learning groups is more targeted: By extracting the deep features of students and using the k-means clustering algorithm for group division, the present invention can divide different learning groups according to the similarity of student characteristics, ensuring the similarity of student characteristics within the group, making mutual learning more efficient, and improving the scientificity and rationality of learning group division.
[0017] 4. Strong dynamic response capability and high adaptability: The present invention can dynamically update the knowledge graph and mutual assistance network according to the real-time data of students in the learning process, adapt to the changes in students' learning status, ensure the continuous optimization of learning mutual assistance relationships, realize the dynamic adjustment of learning group division, and improve the adaptability and flexibility of the system. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] Figure 1 This is a flow chart of a method for building a student mutual assistance network and dividing groups based on a knowledge graph according to the present invention;
[0019] Figure 2 Provide students with a learning portrait diagram;
[0020] Figure 3 This is a schematic diagram of the Bayesian knowledge tracing model;
[0021] Figure 4 A diagram of the student mutual aid network;
[0022] Figure 5 Schematic diagram for dividing students into groups;
[0023] Figure 6 This is a structural block diagram of a student mutual assistance network construction and group division system based on knowledge graph of the present invention. DETAILED DESCRIPTION
[0024] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only intended to illustrate the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
[0025] Example 1
[0026] like Figure 1 As shown, the embodiment of the present invention provides a method for constructing a student mutual assistance network and dividing groups based on a knowledge graph, including the following steps:
[0027] Step S1: Obtain student answer data from the online learning platform or smart learning system to build a student learning profile;
[0028] Step S2: Using the Bayesian knowledge tracking model, calculate the student's mastery of each knowledge point based on the student's answer record and form a knowledge point mastery matrix;
[0029] Step S3: Based on the knowledge point mastery matrix, Manhattan distance is used to calculate the knowledge point mastery differences between students, and a knowledge point mastery difference matrix is constructed; and a student mutual assistance network is constructed based on the knowledge point mastery difference matrix;
[0030] Step S4: Based on the student mutual assistance network, the graph convolutional neural network is used to fully mine the student characteristics, and the k-means clustering method is used to divide the students into groups to form learning groups with different characteristics.
[0031] In one embodiment, the above step S1: constructing a student mutual assistance network and dividing groups based on the knowledge graph specifically includes:
[0032] Step S11: Obtain the online answer data of each student during the learning process through the learning platform as the original answer data for constructing the student learning profile, including: the student's answer time, answer content, answer result, knowledge point label and other information;
[0033] In this embodiment, the goal of data collection is to obtain students' answering behavior data on the online learning platform and organize these data into a structured input matrix for subsequent analysis and processing. The minimum input data is assumed to be a two-dimensional matrix, denoted as , the dimension of the matrix is , where each row represents an answer record, recording the student's answer behavior and related information. The specific data content is as follows:
[0034] 1. Data structure definition
[0035] In the matrix In the , each row represents a student's answer record, which contains the following four fields:
[0036] 1) User ID ( ): indicates the The student's unique identifier for the record. This field can be used to uniquely identify the student to whom the answer record belongs.
[0037] 2) Knowledge point ID ( ): indicates the The knowledge point involved in each record. The knowledge point ID identifies the knowledge point corresponding to the question answered by the student, so as to facilitate subsequent analysis of the student's mastery of different knowledge points.
[0038] 3) Correctness of answers ( ): indicates the The correctness of the answer in the record, with a value of 1 (correct) or 0 (incorrect). This field is used to assess the student's mastery of the knowledge point and help calculate the degree of knowledge mastery.
[0039] 4) Timestamp ( ): indicates the The answer time sequence of each record is in serial numbers, starting from 1 and increasing in increments. Timestamps are used to record the chronological order of answering behaviors, helping to analyze students' learning progress and the dynamic changes in knowledge mastery.
[0040] 2. Data Collection Process
[0041] In the embodiment of the present invention, the process of data collection is as follows:
[0042] 1) Log in to the online learning platform: During the learning process, students log in to their accounts through the online learning platform and practice answering questions on the knowledge points. Each student's answer and the corresponding knowledge point information will be automatically recorded by the system.
[0043] 2) Real-time recording of answering behavior: After a student answers a question, the system generates and stores a record in real time, including user ID, knowledge point ID, answer accuracy, and timestamp. This information is accumulated chronologically to form time series data, facilitating subsequent analysis of students' learning progress and knowledge acquisition dynamics.
[0044] 3) Construct the answer data matrix :Based on the above answer records, the system integrates all answer behaviors into a two-dimensional matrix , where each row represents a record, and each field is user ID, knowledge point ID, answer correctness and timestamp, forming the four column fields of the matrix.
[0045] 4) Data quality check: During the data collection process, the input matrix Perform data quality checks to ensure that each record is complete and accurate, avoid missing key information such as user ID or knowledge point ID, and verify whether the value of the answer correctness field is within the specified range (1 or 0).
[0046] 3. Sample Data
[0047] The sample data includes student ID, knowledge point ID, answer correctness, and timestamp. For example, student 1001 correctly answers knowledge point 102 at timestamp 1. This data provides a raw record of behavior for subsequent steps, supporting analysis of student knowledge mastery and the construction of a mutual assistance network.
[0048] Step S12: Preprocessing the original answer data, including data cleaning, formatting, and data standardization to ensure data consistency and accuracy; at the same time, knowledge point mapping is performed on the original answer data according to the knowledge point labels;
[0049] Step S13: Use the knowledge graph to associate students' answering behaviors with knowledge points and construct a student learning profile for each student. The specific method includes: according to the answer content and knowledge point labels, each student's answering behavior is associated with the corresponding knowledge point; the graph contains three types of nodes, students, knowledge points and questions; the correct and timestamp features are embedded in the question nodes, indicating whether the answer is correct and the order of the answers respectively.
[0050] like Figure 2As shown, the graph consists of three types of nodes. Taking node 1001 as an example, its subordinate nodes are knowledge point nodes, including 101, 102, and 103. The subordinate nodes of these three knowledge point nodes correspond to the questions of the knowledge point. Because there is a one-to-many relationship between knowledge point nodes and question nodes, a knowledge point can have multiple subordinate question nodes. Each question node is embedded with the correct and timestamp features, which respectively indicate whether the answer is correct and the order of the answers. Figure 2 A part of the student learning portrait is shown. The complete graph structure of the embodiment of the present invention includes 27066 student nodes, 266 knowledge point nodes and 2541201 question nodes.
[0051] In one embodiment, the above step S2: using the Bayesian knowledge tracking model, based on the student's answer record, calculates the student's mastery of each knowledge point and forms a knowledge point mastery matrix, which specifically includes:
[0052] Step S21: Initialize the mastery status of each student's knowledge point based on the student's learning profile: Assume that the initial value of the mastery degree is between 0 and 1, where 0 indicates no mastery at all and 1 indicates complete mastery;
[0053] Step S22: Based on the basic parameters of the Bayesian knowledge tracking model, define the learning transfer probability of the knowledge point, including: guess probability , probability of wrong answer , learning probability and forgetting probability These four basic parameters are obtained by training the Bayesian model with historical answer data and are used to subsequently calculate the update of students' knowledge mastery status during the learning process.
[0054] In this embodiment of the present invention, in order to accurately assess each student's mastery of each knowledge point, a Bayesian knowledge tracking model is used to dynamically model the student's knowledge status. The Bayesian knowledge tracking model continuously updates the student's mastery probability of the knowledge point based on the correctness of the student's answer. The mastery of each knowledge point is modeled based on four parameters:
[0055] In the Bayesian knowledge tracking model, each student's mastery of a certain knowledge point is determined by the following four parameters:
[0056] Learning Probability : It indicates the probability that a student has mastered the knowledge point from never mastering it, and is used to measure the success rate of students in learning the knowledge point.
[0057] Forgetting probability : It indicates the probability that students will forget the knowledge point after mastering it, and is used to reflect the possibility that students may forget the knowledge they have mastered.
[0058] Guess probability : It indicates the probability that students answer the question correctly without mastering the knowledge point, reflecting the possibility of guessing correctly by chance.
[0059] Probability of wrong answer : Indicates the probability that a student will answer a question incorrectly when they have mastered the knowledge point. It is used to handle situations where a student has mastered the knowledge point but occasionally answers incorrectly.
[0060] Step S23: After each student answers a question, the student's mastery of the relevant knowledge points is calculated based on the Bayesian knowledge tracking model, and the status is updated according to the correct or incorrect answer information: if the answer is correct, the mastery of the knowledge point is improved; if the answer is incorrect, the mastery of the knowledge point is reduced; based on the mastery of the knowledge point, a knowledge point mastery matrix P with a dimension of U×S is constructed, where U is the number of students, S is the number of knowledge points, and each element of the matrix is represents the probability that student i has mastered knowledge point j. The mastery probability is calculated as follows:
[0061] ;
[0062] ;
[0063] in, represents the probability that the student answers the question correctly, Indicates the probability that the student answered the question incorrectly, the guess probability is the probability that student i does not master knowledge point j but answers correctly; the probability of wrong answer is the probability that student i masters knowledge point j but answers incorrectly;
[0064] Based on the changes in learning and forgetting, the knowledge point mastery matrix P is updated. The learning and forgetting adjustment formulas are as follows:
[0065]
[0066] ;
[0067] Among them, the learning probability is the probability that student i has never mastered knowledge point j; the forgetting probability is the probability that student i will forget knowledge point j after mastering it. This formula takes into account the dual effects of learning and forgetting, and ensures that the student's mastery level reflects the dynamic changes in the learning process by increasing or decreasing the mastery probability.
[0068] The knowledge point matrix P is as follows:
[0069] For each student And each knowledge point , store their final knowledge mastery probability in the knowledge point mastery matrix P, where the elements Indicates student i’s mastery level of knowledge point j.
[0070] .
[0071] For example, suppose a student's initial mastery probability is , learning probability , forgetting probability , guess the probability , probability of wrong answer .
[0072] If the student's first answer record shows that the student answered the question correctly ( ), the updated mastery probability is calculated using the above formula. This updated mastery probability is then incorporated into the learning and forgetting adjustment formula to be adjusted again to produce the mastery probability for the next round. All student answer records are processed sequentially to gradually determine the student's mastery of each knowledge point.
[0073] Through the above method, this embodiment can dynamically evaluate students' mastery of various knowledge points, generate a knowledge point mastery matrix, and provide data support for subsequent mutual assistance relationship building and group division.
[0074] like Figure 3 The figure above shows the structure of the Bayesian knowledge tracking model, where each time step contains the student's knowledge mastery status and answer results. The figure above shows the state evolution of the knowledge tracking model at different time steps. Each time step has two nodes: knowledge mastery status and answer results Knowledge mastery status is a latent variable that indicates students’ understanding of knowledge points In time The degree of mastery is affected by the mastery state of the previous time step (the arrow indicates the dependency relationship). is an observable variable indicating whether the student's answer at that time step is correct or incorrect, depending on the current state of mastery Through this chained structure of time steps, the model can infer how a student's knowledge mastery progresses over time. The figure below shows the four transition probabilities between mastered and non-mastered states: P(T) is the probability of learning, P(S) is the probability of forgetting, P(G) is the probability of guessing correctly when the state is not mastered, and P(S) is the probability of answering incorrectly when the state is mastered. This model is used to track the changes in a student's knowledge mastery over time.
[0075] The pseudo code of the algorithm for calculating students’ knowledge mastery through Bayesian knowledge tracking is as follows:
[0076] Input: student answer data, Bayesian knowledge tracking model parameters (learn_prob, forget_prob, guess_prob, slip_prob), checkpoint file path checkpoint_file
[0077] function BKTSystem(learn_prob, forget_prob, guess_prob, slip_prob,checkpoint_file)
[0078] Initialize student_skill_models and processed_timestamps
[0079] if checkpoint_file exists then load_checkpoint(checkpoint_file)
[0080] end function
[0081] function get_or_create_model(user_id, skill_id)
[0082] If model does not exist then create a BKTModel instance
[0083] return model
[0084] end function
[0085] function update_from_data(data)
[0086] for each record row in data do
[0087] Skip processed timestamp
[0088] Get user_id, skill_id, correct
[0089] model ← get_or_create_model(user_id, skill_id)
[0090] model.update_knowledge(correct)
[0091] end for
[0092] end function
[0093] function get_knowledge_level(user_id, skill_id)
[0094] return get_or_create_model(user_id, skill_id).get_knowledge_level()
[0095] end function
[0096] function get_knowledge_matrix(student_ids, skill_ids)
[0097] Initialize knowledge_matrix
[0098] for user_id in student_ids, skill_id in skill_ids do
[0099] knowledge_matrix[user_id][skill_id] ← get_knowledge_level(user_id,skill_id)
[0100] end for
[0101] return knowledge_matrix
[0102] end function.
[0103] In one embodiment, the above step S3: calculating the knowledge point mastery differences between students using Manhattan distance based on the knowledge point mastery matrix to construct a knowledge point mastery difference matrix; and constructing a student mutual assistance network based on the knowledge point mastery difference matrix specifically includes:
[0104] Step S31: setting a predetermined knowledge point mastery difference threshold T to determine whether the degree of knowledge difference between students meets the conditions for a mutual assistance relationship;
[0105] Step S32: Using the mastery degree data in the knowledge point mastery matrix, Manhattan distance is used to calculate the difference in knowledge point mastery between each pair of students. The specific method is:
[0106] ;
[0107] Among them, the elements Indicates students and students Here, it is assumed that the knowledge points with a mastery level of 0 are not included in the difference calculation. Is an indicator function, which takes 1 when the condition in the brackets is met, otherwise it takes 0. When the mastery of all the above items is not 0, the difference is included in the total;
[0108] Step S33: Record the knowledge point mastery differences between all students in a knowledge point mastery difference matrix; assuming there are n students, the matrix is n×n, with diagonal elements being zero, i.e., the differences within each student are zero, and the other elements represent the knowledge point mastery differences between any two students;
[0109] Step S34: Based on the knowledge point mastery difference matrix, determine whether the knowledge point mastery difference between each pair of students exceeds a difference threshold T; if the difference between the two students is greater than T, the two students meet the mutual assistance condition, and the pair of students is recorded as a mutual assistance relationship tuple; each tuple represents a pair of students with complementary learning needs;
[0110] Set a difference threshold .if , then in students and An edge is established between two parties. This edge can be bidirectional or unidirectional. Bidirectional means that both parties can get help from each other, while unidirectional means that only one party can provide help to the other party, but not vice versa. The edge is characterized by a collection of knowledge points. indicates that there is a significant difference in this knowledge point.
[0111] Step S35: In the knowledge graph, mutual assistance edges are constructed for student pairs that meet the mutual assistance conditions. Each mutual assistance edge represents a learning mutual assistance relationship between two students, and the student mutual assistance network is recorded in a graph structure.
[0112] like Figure 4 As shown, the student mutual assistance network is constructed based on student learning profiles. Taking student 1001 as an example, their mastery of knowledge points 101, 102, and 103 is 0.9, 0.3, and 0.2, respectively. Through the calculation of Manhattan distance, the mastery differences between student 1001, student 1002, and student 1003 all exceed the difference threshold T. Therefore, a mutual assistance relationship is established between student 1001, student 1002, and student 1003. Specifically, student 1001 can help student 1002 learn knowledge point 101, student 1002 can help student 1001 learn knowledge point 102, and student 1001 can help student 1003 learn knowledge point 101. Figure 4This graph shows a portion of the student mutual assistance network. The complete graph structure contains 27,066 student nodes, 266 knowledge point nodes, and 2,541,201 question nodes. There are also 11,294 mutual assistance edges.
[0113] In one embodiment, the above step S4: based on the student mutual assistance network, using a graph convolutional neural network to fully explore student characteristics, and using the k-means clustering method to group students to form learning groups with different characteristics, specifically includes:
[0114] Step S41: Initialize a feature vector for each student node in the student mutual assistance network. The initial feature vector includes basic information about the student and key learning features extracted from the knowledge graph. Use the initial features as input to the graph convolutional neural network to perform deep feature extraction in the network.
[0115] Step S42: Input the student mutual aid network into the graph convolutional neural network. Through the graph node convolution operation of the graph convolutional neural network, the deep features of the students in the mutual aid network are extracted. The graph convolutional neural network updates the feature vector of each student node by transferring and aggregating the information of adjacent nodes layer by layer, so that the final extracted feature vector can reflect the position and feature relationship of the students in the mutual aid network.
[0116] The specific steps are as follows:
[0117] Use graph convolutional neural network to analyze the student mutual assistance network. Let the initial eigenvector be the knowledge point mastery matrix Each row of features in . After two layers of graph convolutional neural network:
[0118] First convolution layer: ;
[0119] Second convolution layer: ;
[0120] in: is the adjacency matrix, which defines the mutual assistance relationship; is a learnable weight matrix; is the student's initial eigenvector, i.e., the knowledge point mastery matrix ; As the input of group division, students are divided into different group matrices through k-means clustering. .
[0121] Step S43: setting a feature division standard based on the feature vector so as to divide students into different learning groups based on feature similarity during the clustering process;
[0122] Step S44: The feature vector extracted from the graph convolutional neural network is input into the k-means clustering algorithm, and cluster analysis is performed according to the preset number of groups k. The k-means clustering method assigns nodes with similar student characteristics to the same group through iterative optimization, and finally forms learning groups with different characteristics.
[0123] The specific steps are as follows:
[0124] After completing the mutual aid network construction and graph convolution operation, a new feature matrix will be obtained ,Each row of the matrix represents the characteristics of students in terms of knowledge mastery and mutual help potential.
[0125] Set the feature matrix for:
[0126] ;
[0127] in, is the feature dimension (obtained from the graph convolutional neural network output).
[0128] K-means is a commonly used clustering algorithm whose goal is to The samples are divided into groups so that the distance between each sample and the center of the group to which it belongs is minimized. The clustering process is as follows:
[0129] 1. Initialize the group center: random selection samples as the initial cluster centers, expressed as .
[0130] 2. Assign samples to the nearest cluster center: For each sample , calculate its Euclidean distance to all cluster centers and assign it to the nearest cluster center:
[0131] ;
[0132] in represents the Euclidean distance.
[0133] 3. Update cluster centers: Calculate the mean of each cluster and update the cluster centers:
[0134]
[0135] in, Indicates that it belongs to The sample index of the cluster.
[0136] 4. Iteration: Repeat steps 2 and 3 until the cluster center no longer changes (converges) or the maximum number of iterations is reached.
[0137] The final output is the group partition matrix , whose dimensions are , where each element indicates which group a student belongs to.
[0138]
[0139] The pseudo code of this process is as follows:
[0140] function GCN_Model(input_dim, hidden_dim, output_dim)
[0141] Initialize two GCN convolutional layers
[0142] end function
[0143] function forward_pass(x, edge_index)
[0144] x ← conv1(x, edge_index) → relu(x)
[0145] x ← conv2(x, edge_index)
[0146] return x
[0147] end function
[0148] function train_model(data, model, optimizer, loss_fn, epochs)
[0149] for epoch from 1 to epochs do
[0150] Calculate loss (assuming the target is a zero vector)
[0151] Backpropagate and update weights
[0152] if (epoch is a multiple of 10) then print epoch and loss
[0153] end for
[0154] end function
[0155] function perform_clustering(node_features, num_groups)
[0156] Returns k-means cluster labels
[0157] end function
[0158] function main()
[0159] Initialize model parameters and graph data
[0160] Training model train_model()
[0161] Group label ← perform_clustering(model output, num_groups=3)
[0162] Print group labels for each student
[0163] end function
[0164] like Figure 5 The figure shows the student clustering and the resulting student grouping matrix. Students are divided into six groups based on their learning ability, knowledge mastery, and active support. Student 1001 belongs to Group 1, student 1002 belongs to Group 4, and student 1003 belongs to Group 0. Each group has distinct learning characteristics. By analyzing the individual characteristics of each group, teachers can provide targeted instruction and achieve precise teaching.
[0165] Example 2
[0166] like Figure 6 As shown, the embodiment of the present invention provides a student mutual assistance network construction and group division system based on knowledge graph, including the following modules:
[0167] The student learning profile building module 51 is used to obtain student answer data information from the online learning platform or the intelligent learning system to build a student learning profile;
[0168] Constructing a knowledge point mastery matrix module 52, which is used to calculate the student's mastery of each knowledge point based on the student's answer record using the Bayesian knowledge tracking model, and form a knowledge point mastery matrix;
[0169] The student mutual assistance network construction module 53 is used to calculate the knowledge point mastery differences between students using Manhattan distance according to the knowledge point mastery matrix, and to construct the knowledge point mastery difference matrix; and to construct the student mutual assistance network according to the knowledge point mastery difference matrix;
[0170] A learning group module 54 is constructed to fully mine student characteristics based on the student mutual assistance network using a graph convolutional neural network, and to divide students into groups using a k-means clustering method to form learning groups with different characteristics.
[0171] A device for constructing a student mutual aid network and dividing groups based on a knowledge graph includes one or more electronic devices, wherein the one or more electronic devices are used to implement a method, system and device for constructing a student mutual aid network and dividing groups based on a knowledge graph.
[0172] An electronic device includes: one or more processors; a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement a method, system and device for constructing a student mutual assistance network and dividing groups based on a knowledge graph.
[0173] A computer-readable storage medium stores executable instructions, which, when executed by a processor, enable the processor to implement a method, system, and device for constructing a student mutual assistance network and dividing groups based on a knowledge graph.
[0174] A non-transitory computer-readable storage medium stores a computer program, which, when executed by a processor, implements a method, system, and device for constructing a student mutual assistance network and dividing groups based on a knowledge graph.
[0175] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features of the present invention.
Claims
1. A method for constructing a student mutual assistance network and dividing groups based on a knowledge graph, characterized in that: include: Step S1: Obtain student answer data from the online learning platform or smart learning system to build a student learning profile; Step S2: Using the Bayesian knowledge tracking model, calculate the student's mastery of each knowledge point based on the student's answer record and form a knowledge point mastery matrix; Step S3: Based on the knowledge point mastery matrix, Manhattan distance is used to calculate the knowledge point mastery differences between students to construct a knowledge point mastery difference matrix; and based on the knowledge point mastery difference matrix, a student mutual assistance network is constructed, specifically including: Step S31: setting a predetermined knowledge point mastery difference threshold T to determine whether the degree of knowledge difference between students meets the conditions for a mutual assistance relationship; Step S32: Using the mastery degree data in the knowledge point mastery matrix, Manhattan distance is used to calculate the difference in knowledge point mastery between each pair of students. The specific method is: ； Among them, the elements Indicates students and students Differences in knowledge mastery between them; is the indicator function; is an element of the knowledge point mastery matrix, representing the mastery level of student i on knowledge point s; Step S33: Record the knowledge point mastery differences among all students in a knowledge point mastery difference matrix; assuming there are n students, the matrix is n×n; Step S34: judging whether the knowledge point mastery difference between each pair of students exceeds the difference threshold T based on the knowledge point mastery difference matrix; if the difference between the two students is greater than T, the two students meet the mutual assistance condition, and the pair of students is recorded as a mutual assistance relationship tuple; Step S35: In the knowledge graph, construct mutual assistance edges for student pairs that meet the mutual assistance conditions. Each mutual assistance edge represents a learning mutual assistance relationship between two students, and the student mutual assistance network is recorded in a graph structure. Step S4: Based on the student mutual aid network, a graph convolutional neural network is used to extract the feature vector of each student node. According to the preset number of groups k, the k-means clustering method is used to group the feature vectors, and nodes with similar student characteristics are assigned to the same group, ultimately forming learning groups with different characteristics.
2. The method for constructing a student mutual assistance network and dividing groups based on a knowledge graph according to claim 1 is characterized in that: The step S1 specifically includes: Step S11: Obtain the online answer data of each student during the learning process through the learning platform as the original answer data for constructing the student learning profile, including: the student's answer time, answer content, answer result, and knowledge point label information; Step S12: Preprocessing the raw answer data, including data cleaning, formatting, and data standardization to ensure data consistency and accuracy; and mapping the raw answer data to knowledge points based on knowledge point labels. Step S13: Use the knowledge graph to associate students' answering behaviors with knowledge points and construct students' learning portraits. The specific methods include: according to the answer content and knowledge point labels, each student's answering behavior is associated with the corresponding knowledge point; the graph contains three types of nodes, students, knowledge points and questions; the correct and timestamp features are embedded in the question nodes, indicating whether the answer is correct and the order of the answers respectively.
3. The method for constructing a student mutual assistance network and dividing groups based on a knowledge graph according to claim 2 is characterized in that: Step S2: Using the Bayesian knowledge tracking model, based on the student's answer records, calculate the student's mastery of each knowledge point and form a knowledge point mastery matrix, which specifically includes: Step S21: Initializing the mastery status of each student's knowledge point based on the student learning profile: assuming that the initial value of the mastery degree is between 0 and 1, 0 represents no mastery at all, and 1 represents complete mastery; Step S22: Based on the basic parameters of the Bayesian knowledge tracking model, define the learning transfer probability of the knowledge point, including: guess probability , probability of wrong answer , learning probability and forgetting probability ,The parameters are obtained by training the Bayesian model with historical answer data; Step S23: After each student answers a question, the student's mastery of the relevant knowledge points is calculated based on the Bayesian knowledge tracking model, and the status is updated according to the correct or incorrect information of the answer; based on the mastery of the knowledge points, a knowledge point mastery matrix P with a dimension of U×S is constructed, where U is the number of students and S is the number of knowledge points. Each element of the matrix Represents the probability that student i masters knowledge point j: ； 。 4. The method for constructing a student mutual assistance network and dividing groups based on a knowledge graph according to claim 3 is characterized in that: The step S4: based on the student mutual aid network, using a graph convolutional neural network to fully explore student characteristics, and using a k-means clustering method to group students to form learning groups with different characteristics, specifically includes: Step S41: Initializing a feature vector for each student node in the student mutual assistance network, the initial feature vector including basic information of the student and key learning features extracted from the knowledge graph; using the initial features as input to a graph convolutional neural network to perform deep feature extraction in the network; Step S42: Input the student mutual aid network into the graph convolutional neural network, and extract the deep features of the students in the mutual aid network through the graph node convolution operation of the graph convolutional neural network; the graph convolutional neural network updates the feature vector of each student node by layer-by-layer transmission and aggregation of information from adjacent nodes, so that the final extracted feature vector can reflect the position and feature relationship of the students in the mutual aid network; Step S43: setting a feature division standard based on the feature vector so as to divide students into different learning groups based on feature similarity during the clustering process; Step S44: The feature vector extracted from the graph convolutional neural network is input into the k-means clustering algorithm, and cluster analysis is performed according to the preset number of groups k. The k-means clustering method assigns nodes with similar student characteristics to the same group through iterative optimization, and finally forms learning groups with different characteristics.
5. A student mutual assistance network construction and group division system based on knowledge graph, characterized by: Includes the following modules: Build a student learning portrait module to obtain student answer data from online learning platforms or smart learning systems to build student learning portraits; Construct a knowledge point mastery matrix module, which uses the Bayesian knowledge tracking model to calculate the student's mastery of each knowledge point based on the student's answer record and form a knowledge point mastery matrix; A student mutual assistance network module is constructed, which is used to calculate the knowledge point mastery differences between students using Manhattan distance based on the knowledge point mastery matrix and construct a knowledge point mastery difference matrix; and constructing a student mutual assistance network based on the knowledge point mastery difference matrix specifically includes: Step S31: setting a predetermined knowledge point mastery difference threshold T to determine whether the degree of knowledge difference between students meets the conditions for a mutual assistance relationship; Step S32: Using the mastery degree data in the knowledge point mastery matrix, Manhattan distance is used to calculate the difference in knowledge point mastery between each pair of students. The specific method is: ； Among them, the elements Indicates students and student m’s knowledge mastery differences; is the indicator function; is an element of the knowledge point mastery matrix, representing the mastery level of student i on knowledge point s; Step S33: Record the knowledge point mastery differences among all students in a knowledge point mastery difference matrix; assuming there are n students, the matrix is n×n; Step S34: judging whether the knowledge point mastery difference between each pair of students exceeds the difference threshold T based on the knowledge point mastery difference matrix; if the difference between the two students is greater than T, the two students meet the mutual assistance condition, and the pair of students is recorded as a mutual assistance relationship tuple; Step S35: In the knowledge graph, construct mutual assistance edges for student pairs that meet the mutual assistance conditions. Each mutual assistance edge represents a learning mutual assistance relationship between two students, and the student mutual assistance network is recorded in a graph structure. A learning group module is constructed to extract the feature vector of each student node based on the student mutual assistance network using a graph convolutional neural network. The feature vectors are divided into groups according to the preset number of groups k using the k-means clustering method, and nodes with similar student characteristics are assigned to the same group, ultimately forming learning groups with different characteristics.
6. A device for constructing a student mutual assistance network and dividing groups based on a knowledge graph, characterized in that: The method comprises one or more electronic devices, wherein the one or more electronic devices are used to implement the method according to any one of claims 1 to 4.
7. An electronic device, characterized in that: include: one or more processors; A memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors are enabled to implement the method according to any one of claims 1 to 4.
8. A computer-readable storage medium, characterized in that Executable instructions are stored thereon, and when the instructions are executed by a processor, the processor implements the method according to any one of claims 1 to 4.
9. A non-transitory computer-readable storage medium, characterized in that A computer program is stored thereon, and when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.