A student answer prediction method, system, device and storage medium

Abstract

The present invention discloses a method, system, device and storage medium for predicting students' answers to questions. The relevant methods include: obtaining a knowledge point hierarchy, test questions and the correspondence between test questions and knowledge points; performing Bayesian network modeling based on the knowledge point hierarchy to infer students' knowledge point mastery patterns; characterizing the test question state in combination with the difficulty of the test question in each knowledge point, utilizing the correspondence between the test question and the knowledge point, converting the students' knowledge point mastery patterns and test question states through a parameter converter, obtaining student characteristics and test question characteristics, and inputting them into a cognitive diagnosis model to predict the students' correct answer rate to the test questions. Using the above-mentioned scheme provided by the present invention, it is possible to efficiently and accurately model students' knowledge point mastery patterns, making up for the drawback that the existing method parameter restrictions are too strong, resulting in the inability to flexibly model the knowledge point hierarchy, and also ensuring the accuracy of the prediction results.

Description

Technical Field

[0001] The present invention relates to the fields of artificial intelligence and intelligent education, and in particular to a method, system, device and storage medium for predicting student answers. Background Art

[0002] Response prediction is a fundamental task in the fields of artificial intelligence and intelligent education. Its purpose is to estimate a student's probability of a correct answer or score based on their understanding of knowledge points (knowledge attributes). In intelligent education systems, student response prediction solutions have broad application prospects in intelligent education services, as student response performance serves as a reference for many important services, such as question recommendation, curriculum design, and ability diagnosis. Because learning dependencies exist between knowledge points during the student learning process, incorporating a knowledge point hierarchy into response prediction solutions is crucial for predicting student responses.

[0003] At present, the application methods of answer prediction scheme in answer prediction mainly include the following methods:

[0004] (1)Answer prediction scheme based on item response theory.

[0005] The answer prediction scheme based on item response theory uses a single-dimensional scalar or low-dimensional vector to abstractly represent the student's cognitive level, and uses low-dimensional parameters to characterize the test characteristics; then a logistic-like function is used to calculate the probability that the student will answer the question correctly; finally, parameter optimization methods such as MCMC (Markov Chain Monte Carlo method) are used to optimize the model and optimize the answer prediction accuracy.

[0006] (2)Answer prediction scheme based on cognitive theory.

[0007] Response prediction schemes based on cognitive theory use variables to directly model students' cognitive level at each knowledge point. In this approach, each question examines a subset of knowledge points, and the relationship between the question and knowledge points is represented using the question Q matrix. Students' accuracy rates for each question are estimated using an interaction function. Finally, parameters are optimized using methods such as Gibbs sampling and gradient descent to estimate students' expected cognitive level at each knowledge point, thereby more accurately predicting their performance. Representative approaches include response prediction schemes based on neurocognitive diagnostic models.

[0008] At present, the main methods for applying the knowledge point hierarchy in question answer prediction are as follows:

[0009] (1) Rule-based attribute hierarchy approach.

[0010] Some scholars have proposed using the hierarchical structure of knowledge points. First, a rule is set requiring students' binary cognitive level at any parent knowledge point to be no less than their cognitive level at the child knowledge point; then the possible cognitive level combinations of students at all knowledge points are enumerated; finally, the traditional cognitive diagnosis model is used to calculate the cognitive level combination that generates the student's answer record with the maximum probability, and this cognitive level combination is used to calculate the student's answer probability on the new question.

[0011] The above method has two limitations: first, the time complexity is unbearable when the number of knowledge points is large. Second, the assumption used to model the knowledge point hierarchy is too strong, which may lose some information about the knowledge point hierarchy. It is also not flexible enough to adapt to all situations. Summary of the Invention

[0012] The purpose of the present invention is to provide a student answer prediction method, system, device and storage medium, which can efficiently and accurately model the student's knowledge point mastery pattern, make up for the disadvantages of the existing method that the parameters are too restricted and the knowledge point hierarchy cannot be flexibly modeled, and can also ensure the accuracy of the prediction results.

[0013] The purpose of the present invention is achieved through the following technical solutions:

[0014] A method for predicting student answers, comprising:

[0015] Obtain the knowledge point hierarchy, test questions, and the correspondence between test questions and knowledge points;

[0016] Based on the hierarchical structure of the knowledge points, Bayesian network modeling is performed to infer the students' knowledge point mastery patterns;

[0017] The test question status is represented by combining the difficulty of the test question in each knowledge point. The correspondence between the test question and the knowledge point is used to convert the student's knowledge point mastery pattern and the test question status through a parameter converter to obtain student characteristics and test question characteristics, and input them into the cognitive diagnosis model to predict the student's correct answer rate to the test question.

[0018] A student answer prediction system, comprising:

[0019] A data acquisition unit, used to obtain the knowledge point hierarchy, test questions, and the correspondence between test questions and knowledge points;

[0020] A network building and pattern inference unit, configured to perform Bayesian network modeling based on the knowledge point hierarchy and infer students' knowledge point mastery patterns;

[0021] The student answer prediction unit is used to characterize the test question status in combination with the difficulty of the test question in each knowledge point, and use the correspondence between the test question and the knowledge point to convert the student's knowledge point mastery mode and the test question status through a parameter converter to obtain student characteristics and test question characteristics, and input them into the cognitive diagnosis model to predict the student's answer accuracy to the test question.

[0022] A processing device comprising: one or more processors; a memory for storing one or more programs;

[0023] When the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned method.

[0024] A readable storage medium stores a computer program, which implements the aforementioned method when the computer program is executed by a processor.

[0025] It can be seen from the technical solution provided by the present invention that the modeling method based on Bayesian networks makes full use of the hierarchical structure of knowledge points, can flexibly and efficiently model the hierarchical relationship of knowledge points, and accurately model the mastery mode of student users on knowledge points, making up for the shortcomings of the existing methods that are too time complex and too strong in parameter restrictions, which lead to the inability to flexibly model the knowledge point hierarchy, and can also ensure the accuracy of the prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0027] Figure 1 A flow chart of a method for predicting student answers provided by an embodiment of the present invention;

[0028] Figure 2 A schematic diagram of a student answer prediction system provided by an embodiment of the present invention;

[0029] Figure 3 A schematic diagram of a processing device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0030] The following is a clear and complete description of the technical solutions in the embodiments of the present invention, in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0031] First, the following terms may be used in this article:

[0032] The terms "include," "comprises," "contains," "has," or other similar expressions should be interpreted as non-exclusive. For example, "including certain technical features (such as raw materials, components, ingredients, carriers, dosage forms, materials, dimensions, parts, components, mechanisms, devices, steps, procedures, methods, reaction conditions, processing conditions, parameters, algorithms, signals, data, products, or manufactured articles, etc.) should be interpreted as including not only the technical features explicitly listed, but also other technical features known in the art that are not explicitly listed.

[0033] The AAAA method provided by the present invention is described in detail below. Any content not described in detail in the examples of the present invention belongs to the prior art known to those skilled in the art. Where specific conditions are not specified in the examples of the present invention, the method is carried out according to conventional conditions in the art or the conditions recommended by the manufacturer. Where the manufacturer of the reagents or instruments used in the examples of the present invention is not specified, they are all conventional products that can be purchased commercially.

[0034] Example 1

[0035] The embodiment of the present invention provides a method for predicting student answers. Figure 1 As shown, it mainly includes:

[0036] Step 1: Obtain the knowledge point hierarchy, test questions, and the correspondence between test questions and knowledge points.

[0037] In an embodiment of the present invention, the knowledge point hierarchy includes: the ID of each knowledge point, whether there are directed edges between different knowledge points, which are pre-marked by educational experts or teachers; specifically, the knowledge point hierarchy can be represented as a directed acyclic graph G = (V, E), where V represents a node set, each node corresponds to a knowledge point; E represents an edge set, and each directed edge connection represents a cognitive dependency relationship between knowledge points, from a parent knowledge point to a child knowledge point, indicating that the child knowledge point depends on the parent knowledge point.

[0038] The set corresponding to the test question is recorded as J = {e1, e2, ..., e M}, where e j Represents the jth test question, j = 1, 2,…, M, and M is the total number of test questions.

[0039] The correspondence between test questions and knowledge points is expressed using the test question Q matrix, which is expressed as Q = (q jv ) M×K ;q jv Represents the element value of the jth row and vth column in the test matrix Q, q jv =1 means test question e j Examined knowledge points v, q jv =0 means test question e j Knowledge points v and q were not examined. jv =I(Test question e j Consider knowledge point v), I(·) is an indicator function. The function value is 1 if and only if the event in the brackets is true, otherwise the function value is 0.

[0040] Furthermore, we need to obtain student answer history data (real data) for the training phase. Let R = {(s, e, y) | s∈S, e∈J, y∈{0,1}} represent the student answer history data set, where y is a binary variable representing the answer score, with 0 indicating an error and 1 indicating a correct answer. (s, e, y) indicates that student s scored y on question e.

[0041] The above information can be obtained from the online learning platform.

[0042] Step 2: Perform Bayesian network modeling based on the knowledge point hierarchy to infer students' knowledge point mastery patterns.

[0043] The preferred implementation of this step is as follows:

[0044] A Bayesian network is established that is isomorphic to the knowledge point hierarchy, wherein each node of the Bayesian network represents a knowledge point, each directed edge represents a cognitive dependency relationship between two knowledge points, and each student corresponds to a Bayesian network.

[0045] At the node level, for a student i and a knowledge point v, use m iv =P(Θ iv =1) represents the probability that student i has mastered knowledge point v, m iv is the cognitive state of student i on knowledge point v (i.e., mastery level), and the vector formed by the cognitive state of student i on all knowledge points is the knowledge point mastery model of student i; where, Θ iv Indicates whether student i has mastered knowledge point v, Θ iv =1 means student i has mastered knowledge point v, Θ iv =0 means that student i has not mastered the knowledge point v, Θ iv =I(student i has mastered knowledge point v), I(·) is an indicator function. The function value is 1 if and only if the event in the brackets is true, otherwise the function value is 0.

[0046] At the directed edge level, for student i and a directed edge (u, v), the directed edge (u, v) points from knowledge point u to knowledge point v, where knowledge point u and knowledge point v are the parent knowledge point and child knowledge point respectively; define the conditional mastery probability and It represents the probability that student i can master knowledge point v when he masters knowledge point u. It represents the probability that student i can master knowledge point v without mastering knowledge point u, and limits 1≤i≤M,(u,v)∈E, where, Θ iu =1 means student i has mastered knowledge point u, Θ iu =0 means that student i has not mastered knowledge point u.

[0047] The topological order of the knowledge point hierarchy is used to infer the student's knowledge point mastery pattern. For a knowledge point v, based on the position of the knowledge point v and the number of parent knowledge points of the knowledge point v, the cognitive state m of the student i's knowledge point v is inferred. iv .

[0048] In the embodiment of the present invention, three cases are mainly considered: (1) Knowledge point v is a root node. In this case, it has no parent knowledge point. (2) Knowledge point v is not a root node and has only a single parent knowledge point. (3) Knowledge point v is not a root node and has only multiple (two or more) parent knowledge points. The following describes the inference of the cognitive state m of knowledge point v of student i for each case. iv Way.

[0049] (1) Knowledge point v is the root node. Knowledge point v has no parent knowledge point, that is, the parent knowledge point set p(v) is an empty set. Then the cognitive state m of student i’s knowledge point v is iv Expressed as:

[0050]

[0051] Among them, the parameters Indicates m iv The optimal value when predicting the answer score of student i.

[0052] (2) Knowledge point v is not a root node, and the number of parent knowledge points is 1. and They represent the probability that student i can master the child knowledge point v when he / she has mastered / not mastered u, and the cognitive state of student i on the parent knowledge point u is represented by probability m iu Therefore, student i’s cognitive state m of sub-knowledge point v is iv Expressed as and The weighted expectation of iu and (1-m iu ):

[0053]

[0054] In addition, for any directed edge (u, v)∈E, the higher the level of cognition of the parent knowledge point, the higher the level of cognition of the child knowledge point, that is, Used to maintain the interpretability of students' cognitive states.

[0055] (3) Knowledge point v is not a root node, and the number of parent knowledge points is z, z>1.

[0056] Let the parent knowledge point set of knowledge point v p(v)={u1,…,u z},u l Represents the lth parent knowledge point, l = 1,…,z, then the cognitive state m of student i’s knowledge point v iv Represents all The joint function of:

[0057]

[0058] in, represents a joint function; Indicates that student i has mastered the lth parent knowledge point u l probability; Indicates that student i has mastered the lth parent knowledge point u l The probability of mastering knowledge point v under the condition of Indicates that student i has not mastered the lth parent knowledge point u l Under the condition of , the probability of mastering knowledge point v.

[0059] In this case, first use To approximate the joint conditional probability that student i has mastered sub-knowledge point v:

[0060]

[0061] in, The specific values ​​of each parent knowledge point indicate whether it is mastered or not. A value of 0 indicates that the corresponding parent knowledge point is not mastered, and a value of 1 indicates that the corresponding parent knowledge point is mastered.

[0062] Approximate function The following properties need to be satisfied: 1) The function value is monotonically increasing with any independent variable component. 2) The function value is sensitive to smaller independent variable components, that is, if at least one independent variable component is small, the function value is small.

[0063] To satisfy the above properties, use The harmonic mean of fits the joint conditional mastery probability of the sub-knowledge points, that is:

[0064]

[0065] Second, fit using the independence assumption:

[0066]

[0067] Finally, the joint conditional mastery probability of the sub-knowledge points combined with the independence hypothesis is used to fit the cognitive state of student i on the sub-knowledge point v:

[0068]

[0069] Step 3: Combine the difficulty of the test questions in each knowledge point to represent the test question status, use the correspondence between the test questions and knowledge points, and convert the student's knowledge point mastery pattern and the test question status through a parameter converter to obtain student characteristics and test question characteristics, and input them into the cognitive diagnosis model to predict the student's answer accuracy rate to the test questions.

[0070] The preferred implementation of this step is as follows:

[0071] (1) Characterize the test question status: For test question e j , using the parameter Represents the state of the test question, where Indicates test question e j The difficulty of each knowledge point, K is the number of knowledge points; β j It is a set of optional parameters, which can be set according to the selected cognitive diagnosis model. j , for example, the discrimination of the test questions can be Included in β j In cognitive diagnosis, the greater the discriminability of the test questions, the greater the difference in scores between two students when the difference in their mastery of the knowledge points is fixed. j The existence of β makes the present invention adaptable to a variety of cognitive diagnosis modeling methods. For example, some cognitive diagnosis models model the test questions as "overall difficulty" and "overall discrimination", then β j It can be set as the “overall discrimination” in scalar form; some cognitive diagnostic models model the test questions as “knowledge point difficulty” and “knowledge point discrimination”, so β j It can be set as the “knowledge point discrimination” in vector form; if the cognitive diagnosis model only considers the difficulty attribute on the question side, then β j Just set it to empty. j All components of are parameters that can be trained based on the student's answer records. In the initial stage, custom initialization ψ is used. j The values ​​of all components are updated during training.

[0072] (2) Let student i’s knowledge point mastery pattern be recorded as m i =(m i1 ,…,m iK ) T , m iv represents the cognitive state of student i’s knowledge point v, v = 1,…,K, T is the transposition symbol; the transposition of the jth row in the test matrix Q is represented as q j =(q j1 ,…,q jK ) T ; Convert through the following formula to obtain student characteristics and test feature v j :

[0073]

[0074]

[0075] Among them, ⊙ represents the multiplication of corresponding elements; W s 、W E 、b s 、b E are all trainable parameters of parameter converter, W s With W E are two weight matrices, b s with b E are two bias terms. Specifically, W s and W E They are two trainable D×K matrices, where D is a positive integer and is the feature dimension given in advance, and b s and b E are two trainable D×1 bias terms. In addition, during training, the restriction W s All elements are not less than 0.

[0076] (3) The cognitive diagnosis model is used to predict the correct rate of students’ answers to test questions, which can be expressed as:

[0077]

[0078] in, Represents the predicted answer of student i to question e j The correct answer rate of the question; F(·) represents the interaction function of the cognitive diagnosis model. Any existing cognitive diagnosis model interaction function that meets the above form can be selected, for example:

[0079]

[0080] The above is the main steps of the above solution of the embodiment of the present invention. In the application, the parameters of the Bayesian network, the test question state, and the parameter converter need to be trained in advance. j , the question state uses the parameter ψ j Characterization; the parameters in the parameter converter include the weight matrix and the bias term (W s 、W E 、b s 、b E ); The parameters in the Bayesian network include: when the knowledge point v is the root node, it is used to determine the cognitive state m of the knowledge point v of student i iv Parameters And the conditional mastery probability when knowledge point v is not the root node and The parameters that need to be trained in advance are represented by the symbol Ω, and the training loss function is expressed as:

[0081]

[0082] Where j = 1, 2, ..., M, 1 ≤ i ≤ N, N represents the number of students; y ij Indicates that student i has a good score on test question e j The real answer score is 1 if the answer is correct, otherwise it is 0, obtained from the student's answer history record data; Represents the predicted answer of student i to question e j The correct answer rate; λ is a hyperparameter, J(Ω) is a penalty term, used to achieve This restriction is expressed as:

[0083]

[0084] in, It represents the probability that student i can master knowledge point v when he masters knowledge point u; It represents the probability that student i can master knowledge point v without mastering knowledge point u. Knowledge point u and knowledge point v are the parent knowledge point and child knowledge point respectively, that is, knowledge point v depends on knowledge point u.

[0085] The above-mentioned scheme of the embodiment of the present invention uses a modeling method based on Bayesian networks, fully utilizing the knowledge point hierarchy and the student user's answer records, and can flexibly and efficiently model the knowledge point hierarchy relationship, while accurately modeling the student user's cognitive level on the knowledge points, making up for the disadvantage of the existing method's excessive parameter restrictions that make it impossible to flexibly model the knowledge point hierarchy.

[0086] Example 2

[0087] The present invention also provides a student answer prediction system, which is mainly implemented based on the method provided in the above embodiment. Figure 2 As shown, the system mainly includes:

[0088] A data acquisition unit, used to obtain the knowledge point hierarchy, test questions, and the correspondence between test questions and knowledge points;

[0089] A network building and pattern inference unit, configured to perform Bayesian network modeling based on the knowledge point hierarchy and infer students' knowledge point mastery patterns;

[0090] The student answer prediction unit is used to characterize the test question status in combination with the difficulty of the test question in each knowledge point, and use the correspondence between the test question and the knowledge point to convert the student's knowledge point mastery mode and the test question status through a parameter converter to obtain student characteristics and test question characteristics, and input them into the cognitive diagnosis model to predict the student's answer accuracy to the test question.

[0091] Those skilled in the art will clearly understand that for the convenience and brevity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the system can be divided into different functional modules to complete all or part of the functions described above.

[0092] The technical details involved in the above system have been described in detail in the previous embodiment 1, so they will not be repeated here.

[0093] Example 3

[0094] The present invention also provides a processing device, such as Figure 3 As shown, it mainly 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 the method provided by the aforementioned embodiment.

[0095] Furthermore, the processing device further includes at least one input device and at least one output device; in the processing device, the processor, memory, input device, and output device are connected via a bus.

[0096] In the embodiment of the present invention, the specific types of the memory, input device, and output device are not limited; for example:

[0097] The input device can be a touch screen, image acquisition device, physical button or mouse;

[0098] The output device may be a display terminal;

[0099] The memory may be a random access memory (RAM) or a non-volatile memory, such as a disk memory.

[0100] Example 4

[0101] The present invention also provides a readable storage medium storing a computer program, which implements the method provided in the above embodiment when the computer program is executed by a processor.

[0102] In the embodiments of the present invention, the computer-readable storage medium may be provided in the aforementioned processing device, for example, as a memory in the processing device. Alternatively, the computer-readable storage medium may be a USB flash drive, a removable hard drive, a read-only memory (ROM), a magnetic disk, or an optical disk, among other media capable of storing program code.

[0103] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A method for predicting student answers, characterized in that: include: Obtain the knowledge point hierarchy, test questions, and the correspondence between test questions and knowledge points; Based on the hierarchical structure of the knowledge points, Bayesian network modeling is performed to infer the students' knowledge point mastery patterns; The difficulty level of each knowledge point in the test question is used to represent the test question status. The correspondence between the test question and the knowledge point is used to convert the student's knowledge point mastery pattern and the test question status through a parameter converter to obtain student characteristics and test question characteristics. These characteristics are input into the cognitive diagnosis model to predict the student's correct answer rate for the test question, including: Represent the state of the test question: let the test question be J = {e1, e2, ..., e M }, where e j represents the jth test question, j=1,2,…,M, M is the total number of test questions; for test question e j , using the parameter Represents the state of the test question, where Indicates test question e j The difficulty of each knowledge point, K is the number of knowledge points; β j It is a set of optional parameters, which are set according to the cognitive diagnosis model; By using the correspondence between the test questions and knowledge points, the student's knowledge point mastering mode and the test question status are converted through the parameter converter: the knowledge point mastering mode of student i is recorded as m i =(m i1 ,…,m iK ) T , m iv Represents the cognitive state of student i’s knowledge point v, v = 1,…,K, T is the transposition symbol; the correspondence between test questions and knowledge points is represented by the test question Q matrix, which is expressed as Q = (q jv ) M×K ;q jv Represents the element value of the jth row and vth column in the test matrix Q, q jv =1 means test question e j Examined knowledge points v, q jv =0 means test question e j Knowledge points v and q were not examined. jv =I(Test question e j In the knowledge point v), I(·) is an indicator function. If and only if the event in the brackets is true, the function value is 1, otherwise the function value is 0; the transpose of the jth row in the test matrix Q is expressed as q j =(q j1 ,…,q jK ) T ; Convert through the following formula to obtain student characteristics and test feature v j : Among them, ⊙ represents the multiplication of corresponding elements; W s 、W E 、b s 、b E are all trainable parameters of parameter converter, W s With W E are two weight matrices, b s with b E are two bias terms; The cognitive diagnosis model is used to predict the correct rate of students’ answers to test questions, which can be expressed as: in, Represents the predicted answer of student i to question e j The correct answer rate of the question; F(·) represents the interaction function of the cognitive diagnosis model.

2. A student answer prediction method according to claim 1, characterized in that: The Bayesian network modeling based on the knowledge point hierarchy to infer the student's knowledge point mastery pattern includes: The knowledge point hierarchy is represented as a directed acyclic graph G = (V, E), where V represents a node set, each node corresponds to a knowledge point; E represents an edge set, and each directed edge connection represents a cognitive dependency relationship between knowledge points, from a parent knowledge point to a child knowledge point, indicating that the child knowledge point depends on the parent knowledge point; Establishing a Bayesian network isomorphic to the knowledge point hierarchy, wherein each node of the Bayesian network represents a knowledge point, and each directed edge represents a cognitive dependency relationship between two knowledge points; For student i and a knowledge point v, use m iv =P(Θ iv =1) represents the probability that student i has mastered knowledge point v, m iv is the cognitive state of student i on knowledge point v, and the vector formed by the cognitive state of student i on all knowledge points is the knowledge point mastery pattern of student i; iv Indicates whether student i has mastered knowledge point v, Θ iv =1 means student i has mastered knowledge point v, Θ iv =0 means that student i has not mastered the knowledge point v, Θ iv = I(student i mastered knowledge point v), I(·) is an indicator function, which takes the value 1 if and only if the event in the brackets is true, and takes the value 0 otherwise; For student i and a directed edge (u, v), the directed edge (u, v) points from knowledge point u to knowledge point v, where knowledge point u and knowledge point v are the parent knowledge point and child knowledge point respectively; define the conditional mastery probability and It represents the probability that student i can master knowledge point v when he masters knowledge point u. It represents the probability that student i can master knowledge point v without mastering knowledge point u, and limits Among them, Θ iu =1 means student i has mastered knowledge point u, Θ iu =0 means student i has not mastered knowledge point u; The topological order of the knowledge point hierarchy is used to infer the student's knowledge point mastery pattern. For a knowledge point v, based on the position of the knowledge point v and the number of parent knowledge points of the knowledge point v, the cognitive state m of the student i's knowledge point v is inferred. iv .

3. A student answer prediction method according to claim 2, characterized in that: The cognitive state m of the knowledge point v of student i is inferred based on the position of the knowledge point v and the number of parent knowledge points of the knowledge point v. iv include: When knowledge point v is the root node, knowledge point v has no parent knowledge point, that is, the parent knowledge point set p(v) is an empty set, then student i’s knowledge point v mastery mode m iv Expressed as: Among them, the parameters Indicates m iv is the optimal parameter for predicting student i’s answer score.

4. A student answer prediction method according to claim 2, characterized in that: The cognitive state m of the knowledge point v of student i is inferred based on the position of the knowledge point v and the number of parent knowledge points of the knowledge point v. iv include: When knowledge point v is not a root node and the number of parent knowledge points is 1, student i’s knowledge point v mastery mode m iv Expressed as: Among them, m iu represents the probability that student i masters knowledge point u.

5. A student answer prediction method according to claim 2, characterized in that: The cognitive state m of the knowledge point v of student i is inferred based on the position of the knowledge point v and the number of parent knowledge points of the knowledge point v. iv include: When the knowledge point v is not the root node and the number of parent knowledge points is z, z>1, let the parent knowledge point set of knowledge point v p(v)={u1,…,u z },u l Represents the lth parent knowledge point, l=1,…,z, then student i’s knowledge point v’s mastery mode m iv Represents all The joint function of: in, represents a joint function; Indicates that student i has mastered the lth parent knowledge point u l probability; Indicates that student i has mastered the lth parent knowledge point u l The probability of mastering knowledge point v under the condition of Indicates that student i has not mastered the lth parent knowledge point u l Under the condition of , the probability of mastering knowledge point v.

6. A student answer prediction method according to claim 1 or 2, characterized in that: The method also includes: the parameters of the Bayesian network, the test question state, and the parameter converter need to be trained in advance, wherein, for the test question e j , the question state uses the parameter ψ j Representation; the parameters in the parameter converter include weight matrix and bias term; the parameters in the Bayesian network include: when the knowledge point v is the root node, it is used to determine the cognitive state m of the knowledge point v of student i iv Parameters And the conditional mastery probability when knowledge point v is not the root node and The parameters that need to be trained in advance are represented by the symbol Ω, and the training loss function is expressed as: Where j = 1, 2, ..., M, M is the total number of test questions; 1 ≤ i ≤ N, N is the number of students; v = 1, ..., K, K is the number of knowledge points; y ij Indicates that student i has the best score on test question e j The real answer score is 1 if the answer is correct, otherwise it is 0; Represents the predicted answer of student i to question e j The correct answer rate; λ is a hyperparameter, J(Ω) is a penalty term, used to achieve This restriction is expressed as: in, It represents the probability that student i can master knowledge point v when he masters knowledge point u; It represents the probability that student i can master knowledge point v without mastering knowledge point u. Knowledge point u and knowledge point v are the parent knowledge point and child knowledge point respectively, that is, knowledge point v depends on knowledge point u.

7. A student answer prediction system, characterized in that: The method according to any one of claims 1 to 6 is implemented, and the system comprises: A data acquisition unit, used to obtain the knowledge point hierarchy, test questions, and the correspondence between test questions and knowledge points; A network building and pattern inference unit, configured to perform Bayesian network modeling based on the knowledge point hierarchy and infer students' knowledge point mastery patterns; The student answer prediction unit is used to characterize the test question status in combination with the difficulty of the test question in each knowledge point, and use the correspondence between the test question and the knowledge point to convert the student's knowledge point mastery mode and the test question status through a parameter converter to obtain student characteristics and test question characteristics, and input them into the cognitive diagnosis model to predict the student's answer accuracy to the test question.

8. A processing device, characterized in that include: one or more processors; a memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 6.

9. A readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images