Tracking prediction method and system based on interaction between knowledge points and learning forgetting

Abstract

The application provides a new tracking and prediction method and system based on interaction between knowledge points and learning forgetting, comprising the following steps: constructing a knowledge structure containing multiple relationships; realizing feature embedding; modeling multiple interactions between knowledge points in the space aspect; modeling learning and forgetting effects in the time aspect; and predicting future answering performance of students. The application improves the graph knowledge tracking model by using technical methods such as a gated graph neural network model and time series modeling, systematically models the space factors and time factors that affect the knowledge mastery level of learners, can predict the answering conditions of learners, and provides a reliable and convincing explanation for understanding the knowledge state of learners.

Description

Technical Field

[0001] This invention relates to the field of computer data processing technology, and in particular to a tracking and prediction method and system that considers multiple interactions between knowledge points and learning forgetting. Background Technology

[0002] The rapid development of Intelligent Tutoring Systems (ITS) has created opportunities for personalized learning. Knowledge tracking and prediction, as a key technology in personalized learning, aims to accurately track learners' knowledge mastery levels and predict their responses, providing technical support for computer-aided educational applications such as the development of ITS, course learning, and learning resource recommendation.

[0003] Learners' knowledge acquisition levels are influenced by space and time, and predicting these factors is a crucial part of knowledge tracking. Figure 1 This demonstrates how learners' knowledge acquisition level is influenced by space and time during the answering process.

[0004] In terms of the spatial relationship of knowledge, the learner's main learning method is meaningful learning, that is, the learner understands the connection between new knowledge and existing knowledge. During the learning process, existing knowledge and new knowledge interact with each other, so that new knowledge acquires new meaning, and existing knowledge is revised to obtain more accurate and richer meaning. Figure 1 As shown, assuming there is knowledge , , , They are respectively "area" "area of ​​rectangle" "Area of ​​a square" ","length " Compare The scope is broader, therefore yes The higher-level knowledge, yes The subordinate knowledge of the two, and there are superordinate and subordinate relationships between them; and There is no inclusion relationship between them, but they often need to be understood in combination, therefore and There is a parallel combination relationship. Learner practice. This can deepen understanding Understanding, strengthening Furthermore, regarding the time-related effects of learning and forgetting, the degree of knowledge mastery diminishes over time. This means that learners' mastery of knowledge can become rusty and forgotten due to a lack of practice, but it can also be consolidated and strengthened through practice.

[0005] like Figure 1As shown, Due to a lack of practice for a long time, the time step hour Knowledge mastery level is lower than time step The level of knowledge mastery at that time.

[0006] As mentioned above, the level of knowledge state is influenced by both the spatial relationships between knowledge and the temporal effects of learning and forgetting. Spatially, the knowledge structure in knowledge-based cognitive tasks contains rich information such as similarities and dependencies between knowledge. Using a clear structure helps improve model performance and interpretability; therefore, many researchers focus on knowledge structure. KSGKT and DGMN identify relationships between knowledge, while GKT and SKT model the mutual influence between knowledge. However, these models are either limited to using knowledge structure to improve problem representation, represent a limited range of knowledge relationships, or are not conducive to modeling the propagation of multiple relationships. Nevertheless, multiple relationships exist between knowledge, and their interactions can affect the level of knowledge mastery. Temporally, considering the time factor helps accurately simulate the evolution of learners' knowledge state; therefore, many researchers integrate temporal information. HawkesKT and RKT use kernel functions to simulate time effects, while LPKT and GFLDKT improve gating mechanisms. However, meaningful learning and forgetting depend on establishing connections between new and old concepts. The models mentioned above mainly focus on the temporal information of individual knowledge and do not model the knowledge relationships involved in the interaction sequence. In summary, existing models have limitations in modeling spatial and temporal effects, and do not simultaneously model spatiotemporal influences. This makes it difficult to reasonably simulate the evolution of learners' knowledge states, resulting in a contradiction between the learners' actual mastery of knowledge and the data simulation. Consequently, the model results lack accuracy and credibility.

[0007] Therefore, it is necessary to study a tracking and prediction method and system that takes into account multiple interactions between knowledge points and learning forgetting. Summary of the Invention

[0008] The main objective of this invention is to provide a solution to the aforementioned technical problems.

[0009] To achieve the above objectives, this invention provides a tracking and prediction method based on the interaction between knowledge points and learning-forgetting, comprising the following steps: S1, construct a knowledge structure G that includes various relationships between knowledge points. , ,node Represented as the corresponding number There are several knowledge points, and the edge e between knowledge point i and knowledge point j is represented by the dependency type r as follows: , This refers to the dependency relationship between any two knowledge points. relation type ,in, Indicates a superior-subordinate relationship. Indicates a subordinate relationship. Indicates a parallel combination relationship; S2, the repetition time interval Sequence time interval and the number of times in the past The temporal feature vector is obtained by connecting and passing through a linear layer. ; Repeat time interval Sequence time interval The forgotten feature vector is obtained after passing through a linear layer. The hidden states and conceptual embeddings of related knowledge points k of knowledge point i are concatenated into an integrated feature vector. , time feature vector Forgotten feature vectors and integrated feature vectors This serves as the input vector for the gated graph neural network model; where t represents the current time step. S3, the gated graph neural network model applies a loss function to learn the parameters during backpropagation. Each time the parameters are updated, the gated graph neural network model is used to simulate the propagation behavior of different relationships between knowledge points based on the knowledge points and edge e in the knowledge structure G, in order to simulate the influence propagation process. S4 uses a gating mechanism to fuse time features to simulate time effects, based on a reset gate. The candidate hidden state at the current time is calculated. Based on the update gate and candidate hidden state The hidden state at the current time is calculated. That is, to obtain the student's knowledge state after answering the questions at time t. Use this state as the initial state for the student to answer the next question, and return to step S2 again; S5, based on the gated graph neural network model obtained after parameter updates and evolution, predicts the correct answer rate of the respondents.

[0010] Preferably, step S1 includes: S11, through formula Calculate the influence of knowledge point i on knowledge point j ,in, This represents the probability of transitioning from knowledge point i to knowledge point j. This represents the knowledge points in the learner's response sequence. In the knowledge points The number of times a correct answer is given; The threshold is set to 0.2 to obtain sparse edges, and the loops between knowledge points in the knowledge structure G are removed to obtain a directed acyclic graph; the incoming edge of each node represents the subordinate relationship; the outgoing edge of each node represents the superior relationship. S12, through formula Calculate the probability that knowledge point i and knowledge point j are both included in the same question. ,in, This represents a question-knowledge matrix, where each row represents the knowledge involved in each question; express The List, Represents element-wise product. Representing the first normal form; delete the edges representing superordinate and subordinate relationships in step S11, and then... Max-min-scaling normalization is performed to obtain We set a threshold of 0.2 to obtain sparse edges, thereby obtaining the edges between two knowledge points represented as parallel combination relationships.

[0011] Preferred, time feature vector The calculation formula is: ; Forgotten feature vector The calculation formula is: ; in, , It is a weight matrix. , It is a bias term. This indicates the concatenation of elements.

[0012] Preferably, the integrated feature vector of knowledge point i and related knowledge point k. The calculation formula is: in, This represents the hidden state of knowledge point k. For knowledge point k at time step The vector representation after time aggregation Indicates the dimension of a vector; In time step The embedding vector contains both knowledge and answer information; For the embedding matrix of the knowledge index, express The OK.

[0013] Preferably, step S3 includes: Based on different edge types, each knowledge point's corresponding node exchanges information with its neighboring nodes, resulting in a vector containing information about neighboring nodes. The calculation formula is as follows: in, This indicates the communication method between nodes; the parameters in the matrix are determined by the edge type. for Regarding nodes The column; Indicates at time step The hidden state of all nodes at that time; Indicates excluding the current node Except for the hidden state, all values ​​are filled with 0.

[0014] Preferably, step S4 includes: Based on the update gate and reset gate ,calculate and The current hidden state is obtained by linear interpolation and dimensionality reduction using a fully connected layer. The calculation formula is as follows: in, and It is a multilayer perceptron (MLP). This represents element-wise multiplication. , and This is the weight matrix. This is a bias term.

[0015] Preferably, step S5 includes: transferring knowledge points Hidden state Feed it into a fully connected layer with Sigmoid activation; then according to the formula Calculate the accuracy rate of answering questions on knowledge point v. ;in, The weight matrix is ​​a learnable matrix. This is a learnable bias term.

[0016] Preferably, before step S5, the method further includes selecting the Adam optimizer based on the existing dataset and using the binary cross-entropy loss function as the objective function to optimize the gated graph neural network model. The objective function formula is as follows: , This indicates the predicted answer performance for knowledge point v. .

[0017] The present invention also provides a tracking and prediction system based on the interaction between knowledge points and learning and forgetting, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the tracking and prediction method based on the interaction between knowledge points and learning and forgetting as described in any of the preceding claims.

[0018] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the tracking and prediction method based on the interaction between knowledge points and learning and forgetting as described in any of the preceding claims: This invention proposes a novel method for solving the knowledge tracking problem, which comprehensively considers the influence of spatiotemporal factors on learners' knowledge status. Spatially, it uses GGNN to simulate the propagation behavior of various relationships between knowledge points and their different effects; temporally, it constructs a GRU-like gate incorporating three temporal features to simulate the impact of learning and forgetting on knowledge mastery. This invention combines high performance and interpretability, enabling fine-grained diagnosis of learners' knowledge status and providing more comprehensive data support for the implementation of precision teaching and personalized learning. Attached Figure Description

[0019] The accompanying drawings, as part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments and descriptions of the invention are used to explain the invention, but do not constitute an undue limitation of the invention. Obviously, the drawings described below are merely some embodiments, and those skilled in the art can obtain other drawings based on these drawings without creative effort. In the drawings: Figure 1 This is a schematic diagram illustrating how space and time influence a learner's level of knowledge acquisition during the answering process in existing technologies.

[0020] Figure 2 This is a schematic diagram of the neural network model structure of a tracking and prediction method based on the interaction between knowledge points and learning and forgetting in one embodiment of the present invention.

[0021] Figure 3 This is a flowchart illustrating a tracking and prediction method based on the interaction between knowledge points and learning-forgetting in one embodiment of the present invention.

[0022] Figure 4 is a schematic diagram of the hardware structure for running a tracking and prediction method based on the interaction between knowledge points and learning and forgetting in one embodiment of the present invention.

[0023] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0024] The technical problems solved by the embodiments of the present invention, the technical solutions adopted, and the technical effects achieved will be clearly and completely described below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other equivalent or obvious variations of embodiments obtained by those skilled in the art without creative effort fall within the protection scope of the present invention. The embodiments of the present invention can be embodied in various different ways as defined and covered by the claims.

[0025] It should be noted that many specific details are given in the following description for ease of understanding. However, it is obvious that the present invention may be implemented without these specific details.

[0026] It should be noted that, in the absence of explicit limitations or conflicts, the various embodiments and their technical features in this invention can be combined with each other to form a technical solution.

[0027] Please combine them together Figure 2-3 This invention proposes a tracking and prediction method based on the interaction between knowledge points and learning-forgetting, including the following steps: S1, construct a knowledge structure G that includes various relationships between knowledge points. , ,node Represented as the corresponding number There are several knowledge points, and the edge e between knowledge point i and knowledge point j is represented by the dependency type r as follows: , This refers to the dependency relationship between any two knowledge points. relation type ,in, Indicates a superior-subordinate relationship. Indicates a subordinate relationship. Indicates a parallel combination relationship; S2, the repetition time interval Sequence time interval and the number of times in the past The temporal feature vector is obtained by connecting and passing through a linear layer. ; Repeat time interval Sequence time interval The forgotten feature vector is obtained after passing through a linear layer. The hidden states and conceptual embeddings of related knowledge points k of knowledge point i are concatenated into an integrated feature vector. , time feature vector Forgotten feature vectors and integrated feature vectors This serves as the input vector for the gated graph neural network model; where t represents the current time step. S3, the gated graph neural network model applies a loss function to learn the parameters during backpropagation. Each time the parameters are updated, the gated graph neural network model is used to simulate the propagation behavior of different relationships between knowledge points based on the knowledge points and edge e in the knowledge structure G, in order to simulate the influence propagation process. S4 uses a gating mechanism to fuse time features to simulate time effects, based on a reset gate. The candidate hidden state at the current time is calculated. Based on the update gate and candidate hidden state The hidden state at the current time is calculated. That is, to obtain the student's knowledge state after answering the questions at time t. Use this state as the initial state for the student to answer the next question, and return to step S2 again; S5, based on the gated graph neural network model obtained after parameter updates and evolution, predicts the correct answer rate of the respondents.

[0028] Specifically, step S1 involves constructing a knowledge structure containing multiple relationships: automatically learning the relationships between knowledge points from the learner's learning sequence using statistical methods, and constructing a knowledge structure accordingly. This includes: S11, through formula Calculate the influence of knowledge point i on knowledge point j ,in, This represents the probability of transitioning from knowledge point i to knowledge point j. This represents the knowledge points in the learner's response sequence. In the knowledge points The number of times a correct answer is given; The threshold is set to 0.2 to obtain sparse edges, and the loops between knowledge points in the knowledge structure G are removed to obtain a directed acyclic graph; the incoming edge of each node represents the subordinate relationship; the outgoing edge of each node represents the superior relationship. S12, through formula Calculate the probability that knowledge point i and knowledge point j are both included in the same question. ,in, This represents a question-knowledge matrix, where each row represents the knowledge involved in each question; express The List, Represents element-wise product. Representing the first normal form; delete the edges representing superordinate and subordinate relationships in step S11, and then... Max-min-scaling normalization is performed to obtain We set a threshold of 0.2 to obtain sparse edges, thereby obtaining the edges between two knowledge points represented as parallel combination relationships.

[0029] In step S2, the time feature vector The calculation formula is: ; Forgotten feature vector The calculation formula is: ; in, , It is a weight matrix. , It is a bias term. This indicates the concatenation of elements.

[0030] Specifically, for the current knowledge point i, concatenate the interactive embedding of knowledge point i and its hidden state. For a relevant knowledge point k, concatenate its knowledge embedding and hidden state. Then, the integrated feature vector of knowledge point i and related knowledge point k. The calculation formula is: in, This represents the hidden state of knowledge point k related to knowledge point i. For knowledge point i, related knowledge point k at time step The vector representation after time aggregation Indicates the dimension of a vector; In time step The embedding vector contains both knowledge and answer information; For the embedding matrix of the knowledge index, express The OK.

[0031] Preferably, in step S3, the gated graph neural network model learns the parameters by applying a loss function during backpropagation. Each time the parameters are updated, the model uses the edge type to represent the relationship between different knowledge points by modeling various interactions between knowledge points in the space, and uses the gated graph neural network model (GGNN network) to model the evolution process of knowledge state affected by spatial factors.

[0032] Specifically, for each parameter update, node features are aggregated based on the knowledge structure and edge type. According to different edge types, each node exchanges information with its neighboring nodes to obtain a vector containing information about neighboring nodes. Each time a parameter update is performed, the hidden state at time t is multiplied by the adjacency matrix containing the three types of relationships, resulting in a vector containing information about neighboring nodes k. In other words, during the spatial evolution process, the current node aggregates information from its neighbor node k, and the neighbor node k aggregates information from the current node. Depending on the relationship between the nodes, the weight vector (i.e., the edge type is a weight vector) used in the aggregation calculation is different.

[0033] Specifically, considering the edge type, the current node Aggregate all neighboring nodes Information, neighbor nodes Aggregate current node From the information, we obtain a vector containing information about neighboring nodes. The calculation formula is as follows: in, This indicates the communication method between nodes; the parameters in the matrix are determined by the edge type. for Regarding nodes The column; Indicates at time step The hidden state of all nodes at that time; Indicates excluding the current node Except for the hidden state, all values ​​are filled with 0.

[0034] Preferably, in step S4, the learning-forgetting effect in terms of modeling time is considered simultaneously with the learning-forgetting patterns based on the spatial effect of interactions between modeled knowledge, and a gating mechanism is used to fuse time features to simulate the time effect. The time effect is simulated by fusing time features using a gating mechanism based on a reset gate. The candidate hidden state at the current time is calculated. Based on the update gate and candidate hidden state The hidden state at the current time is calculated. That is, to obtain the student's knowledge state after answering the questions at time t. This state is used as the initial state for the student to answer the next question, and the process returns to step S2. By fully exploring the characteristics of the learner's learning process from both temporal and spatial perspectives, a more comprehensive and reasonable model of the learner's knowledge state evolution process can be achieved.

[0035] Preferably, step S4 includes: Based on the update gate and reset gate ,calculate and The linear interpolation is obtained by dimensionality reduction through a fully connected layer. The calculation formula is as follows: in, and It is a multilayer perceptron (MLP). This represents element-wise multiplication. , and This is the weight matrix. This is a bias term.

[0036] Formula (7): Based on vectors containing neighboring node information Interactive information x t Time characteristics t t and the hidden state at the previous moment The update gate is calculated using a fully connected layer and the sigmoid function. The update gate is used to determine the learner's level of mastery in updating their knowledge status; Formula (8): Based on the vector m containing neighboring node information t Interactive information x t Forgetting characteristics f t And the hidden state h from the previous moment t-1 The reset gate is calculated using a fully connected layer and the sigmoid function. The reset gate is used to determine the degree to which learners have forgotten previously acquired knowledge; Formula (9): Based on the vector m containing neighboring node information t Reset gate, hidden state h t-1 Calculate the candidate hidden state ; Formula (10): Based on the update gate z t Hidden state ht-1 Candidate hidden state The current hidden state is calculated; Return to step S2.

[0037] This state is used as the initial state for the student to answer the next question, and the knowledge state is continuously updated based on the student's response to the next question. This process is repeated to track changes in the student's knowledge state.

[0038] The training dataset for the gated graph neural network model (the dataset can be selected from ASSIST09, ASSIST12, and ASSIST17 datasets) includes student ID, question ID, knowledge point ID, answer status, and time data. Model training: The Adam optimizer is selected, and the binary cross-entropy loss function is used as the objective function to optimize the model. The objective function formula is (12): , This indicates the predicted answer performance for knowledge point v. .

[0039] Those skilled in the art can select an appropriate objective function for training based on the training needs.

[0040] In step S5, predicting the student's future answer performance: based on tracking the evolution of the learner's knowledge state, predicting the probability of the learner answering the next question correctly. Preferably, step S5 includes: […]. Hidden state Feed it into a fully connected layer with Sigmoid activation; then according to the formula Calculate the accuracy rate of answering questions on knowledge point v. ;in, The weight matrix is ​​a learnable matrix. This is a learnable bias term.

[0041] This invention proposes a novel method for solving the knowledge tracking problem, which comprehensively considers the influence of spatiotemporal factors on learners' knowledge status. Spatially, it uses GGNN to simulate the propagation behavior of various relationships between knowledge points and their different effects; temporally, it constructs a GRU-like gate incorporating three temporal features to simulate the impact of learning and forgetting on knowledge mastery. This invention combines high performance and interpretability, enabling fine-grained diagnosis of learners' knowledge status and providing more comprehensive data support for the implementation of precision teaching and personalized learning.

[0042] The present invention also provides a tracking and prediction system based on the interaction between knowledge points and learning and forgetting, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the tracking and prediction method based on the interaction between knowledge points and learning and forgetting as described in any of the preceding claims.

[0043] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the tracking and prediction method based on the interaction between knowledge points and learning and forgetting as described in any of the preceding claims.

[0044] Figure 4 This is a schematic diagram of the hardware structure for running a tracking and prediction method based on the interaction between knowledge points and learning-forgetting, according to an embodiment of the present invention. Figure 4 As shown, this embodiment / computer 6 includes: a processor 60, a memory 61, and a computer program 62 stored in the memory 61 and executable on the processor 60, such as a program running a computational thinking assessment method based on BP neural network and multi-source data fusion. When the processor 60 executes the computer program 62, it implements the steps in the above-described embodiments of tracking and predicting methods based on the interaction between knowledge points and learning-forgetting. Alternatively, when the processor 60 executes the computer program 62, it implements the functions of each module / unit in the above-described device embodiments.

[0045] For example, the computer program 62 may be divided into one or more modules / units, which are stored in the memory 61 and executed by the processor 60 to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program 62 in the computer 6.

[0046] The computer 6 can be a desktop computer, laptop, handheld computer, cloud server, or other computing device. The computer 6 may include, but is not limited to, a processor 60 and a memory 61. Those skilled in the art will understand that... Figure 4 This is merely an example of computer 6 and does not constitute a limitation on computer 6. It may include more or fewer components than shown, or combine certain components, or different components. For example, computer 6 may also include input / output devices, network access devices, buses, etc.

[0047] The processor 60 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0048] The memory 61 can be an internal storage unit of the computer 6, such as a hard drive or memory. The memory 61 can also be an external storage device of the terminal device 6, such as a plug-in hard drive, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the terminal device. Furthermore, the memory 61 can include both internal storage units and external storage devices of the computer 6. The memory 61 is used to store the computer program and other programs and data required by the terminal device. The memory 61 can also be used to temporarily store data that has been output or will be output.

[0049] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0050] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0051] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0052] In the embodiments provided by this invention, it should be understood that the disclosed apparatus / terminal devices and methods can be implemented in other ways. For example, the apparatus / terminal device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0053] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0054] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0055] If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0056] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A tracking and prediction method based on the interaction between knowledge points and learning-forgetting, characterized in that, Including the following steps: S1, construct a knowledge structure G that includes various relationships between knowledge points. , ,node Represented as the corresponding number There are several knowledge points, and the edge e between knowledge point i and knowledge point j is represented by the dependency type r as follows: , This refers to the dependency relationship between any two knowledge points. relation type ,in, Indicates a superior-subordinate relationship. Indicates a subordinate relationship. Indicates a parallel combination relationship; S2, the repetition time interval Sequence time interval and the number of times in the past The temporal feature vector is obtained by connecting and passing through a linear layer. ; Repeat time interval Sequence time interval The forgotten feature vector is obtained after passing through a linear layer. The hidden states and conceptual embeddings of related knowledge points k of knowledge point i are concatenated into an integrated feature vector. , time feature vector Forgotten feature vectors and integrated feature vectors This serves as the input vector for the gated graph neural network model; where t represents the current time step. S3, the gated graph neural network model applies a loss function to learn the parameters during backpropagation. Each time the parameters are updated, the gated graph neural network model is used to simulate the propagation behavior of different relationships between knowledge points based on the knowledge points and edge e in the knowledge structure G, in order to simulate the influence propagation process. S4 uses a gating mechanism to fuse time features to simulate time effects, based on a reset gate. The candidate hidden state at the current time is calculated. Based on the update gate and candidate hidden state The hidden state at the current time is calculated. That is, to obtain the student's knowledge state after answering the questions at time t. Use this state as the initial state for the student to answer the next question, and return to step S2 again; S5, based on the gated graph neural network model obtained after parameter updates and evolution, predicts the correct answer rate of the respondents.

2. The tracking and prediction method based on the interaction between knowledge points and learning-forgetting as described in claim 1, characterized in that, Step S1 includes: S11, through formula Calculate the influence of knowledge point i on knowledge point j ,in, This represents the probability of transitioning from knowledge point i to knowledge point j. This represents the knowledge points in the learner's response sequence. In the knowledge points The number of times a correct answer is given; The threshold is set to 0.2 to obtain sparse edges, and the loops between knowledge points in the knowledge structure G are removed to obtain a directed acyclic graph; the incoming edge of each node represents the subordinate relationship; the outgoing edge of each node represents the superior relationship. S12, through formula Calculate the probability that knowledge point i and knowledge point j are both included in the same question. ,in, This represents a question-knowledge matrix, where each row represents the knowledge involved in each question; express The List, Represents element-wise product. Representing the first normal form; delete the edges representing superordinate and subordinate relationships in step S11, and then... Max-min-scaling normalization is performed to obtain We set a threshold of 0.2 to obtain sparse edges, thereby obtaining the edges between two knowledge points represented as parallel combination relationships.

3. The tracking and prediction method based on the interaction between knowledge points and learning-forgetting as described in claim 1, characterized in that, Time feature vector The calculation formula is: ; Forgotten feature vector The calculation formula is: ; in, , It is a weight matrix. , It is a bias term. This indicates the concatenation of elements.

4. The tracking and prediction method based on the interaction between knowledge points and learning-forgetting as described in claim 3, characterized in that, The integrated feature vector of knowledge point i related to knowledge point k The calculation formula is: in, This represents the hidden state of knowledge point k related to knowledge point i. For knowledge point i, related knowledge point k at time step The vector representation after time aggregation; In time step The embedding vector contains both knowledge and answer information; For the embedding matrix of the knowledge index, express The OK.

5. The tracking and prediction method based on the interaction between knowledge points and learning-forgetting as described in claim 1, characterized in that, Step S3 includes: Based on different edge types, each knowledge point's corresponding node exchanges information with its neighboring nodes, resulting in a vector containing information about neighboring nodes. The calculation formula is as follows: in, This indicates the communication method between nodes; the parameters in the matrix are determined by the edge type. for Regarding nodes The column; Indicates at time step The hidden state of all nodes at that time; Indicates excluding the current node Except for the hidden state, all values ​​are filled with 0.

6. The tracking and prediction method based on the interaction between knowledge points and learning-forgetting as described in claim 1, characterized in that, Step S4 includes: Based on update gate and reset gate ,calculate and The current hidden state is obtained by linear interpolation and dimensionality reduction using a fully connected layer. The calculation formula is as follows: in, and It is a multilayer perceptron. This represents element-wise multiplication. , and This is the weight matrix. This is a bias term.

7. The tracking and prediction method based on the interaction between knowledge points and learning-forgetting as described in claim 1, characterized in that, Step S5 includes: transferring knowledge points Hidden state Feed it into a fully connected layer with Sigmoid activation; then according to the formula Calculate the accuracy rate of answering questions on knowledge point v. ;in, The weight matrix is ​​a learnable matrix. This is a learnable bias term.

8. The tracking and prediction method based on the interaction between knowledge points and learning-forgetting as described in claim 7, characterized in that, Before step S5, the process also includes selecting the Adam optimizer based on the existing dataset and using the binary cross-entropy loss function as the objective function to optimize the gated graph neural network model. The objective function formula is as follows: , This indicates the predicted answer performance for knowledge point v.

9. A tracking and prediction system based on the interaction between knowledge points and learning-forgetting, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the tracking and prediction method based on the interaction between knowledge points and learning and forgetting as described in any one of claims 1 to 8.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the tracking and prediction method based on the interaction between knowledge points and learning and forgetting as described in any one of claims 1 to 8.

Images