Student learning cognition prediction method and device based on bias tensor decomposition

Abstract

The invention discloses a student learning cognition prediction method and device based on bias tensor decomposition. The method specifically comprises the following steps: S1, collecting historical student test score data from various educational activities from a server and storing the historical student test score data; s2, constructing a student test score data tensor according to the stored historical student test score data; s3, constructing a first linear deviation vector according to the stored historical student test score data; s4, according to the constructed student test score data tensor and the first linear deviation vector, constructing a target loss function, and extracting a hidden feature matrix; and S5, outputting and storing a learning cognition score predicted value by using the LSTM and the extracted hidden feature matrix.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and apparatus for predicting student learning cognition based on bias tensor decomposition. Background Technology

[0002] With the rapid development of information technology and artificial intelligence, the education industry has begun to utilize advanced technologies to improve teaching quality and learning efficiency. In particular, the widespread adoption of online education platforms has made a large amount of student learning data available for analysis and processing. How to accurately assess and predict students' mastery of various knowledge points has become a key issue.

[0003] Students' cognitive scores, as an important indicator of learning effectiveness, play a crucial role in the development of personalized education plans. However, traditional methods for predicting cognitive scores are mostly based on simple statistics or static data analysis. These methods often ignore the dynamic changes in learning behavior over time, thus failing to accurately reflect students' learning progress and real-time mastery. For example, while analyzing historical score data using traditional matrix factorization techniques can predict unknown score data, these methods do not consider the time factor, often resulting in predictions that lack timeliness and accuracy.

[0004] Currently, many researchers are exploring dynamic data processing methods, attempting to integrate time series analysis to improve the accuracy and practicality of predictions. How to effectively combine temporal dynamics and individual differences in predicting student learning scores, and how to utilize this information to develop more precise teaching strategies and learning recommendations, has become an urgent problem to be solved in the field of educational technology. Summary of the Invention

[0005] To address the problem of low accuracy in predicting students' cognitive scores in existing technologies, this invention proposes a method and apparatus for predicting student learning cognition based on bias tensor decomposition.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A student learning cognition prediction method based on bias tensor decomposition specifically includes the following steps:

[0008] S1: Collect and store historical student test score data from various educational activities from the server;

[0009] S2: Construct a student test score data tensor based on the stored historical student test score data;

[0010] S3: Construct the first linear deviation vector based on the stored historical student test score data;

[0011] S4: Based on the constructed student test score data tensor and the first linear deviation vector, construct the target loss function and extract the latent feature matrix;

[0012] S5: Using LSTM and the extracted latent feature matrix, output and store the predicted learning cognitive score.

[0013] Preferably, in step S1, the received historical student test score data is stored in the form of a quadruple, which is represented as X = (s, d, t, u), where s represents the student, d represents the test question completed by the student, t represents the time period during which student s completed test question d, and u represents the score of student s in completing test question d during the time period t.

[0014] Preferably, in step S2, the student test score data tensor is: X∈R I×J×K R represents the set of real numbers, I is the number of students, J is the number of test questions, K is the number of time periods, 1≤i≤I, 1≤j≤J, 1≤k≤K.

[0015] Preferably, S3 includes:

[0016] S3-1: Based on the stored historical student test score data, create a linear deviation vector a for all students s;

[0017] S3-2: Based on the stored historical student test score data, create a linear deviation vector b for all test items d.

[0018] Preferably, S4 includes:

[0019] S4-1: Initialize the process parameters involved in predicting student learning cognitive diagnostic scores;

[0020] S4-2: Construct the tensor decomposition objective loss function based on the constructed student test score data tensor, process parameters, and the first linear deviation vector;

[0021] S4-3: Train and iterate the tensor decomposition objective loss function to extract the latent feature matrix.

[0022] Preferably, in step S4-1, the process parameters include:

[0023] Student test score data tensor X, latent feature matrices S, D, T, latent feature dimension M, maximum number of training iterations R, iteration number control variable r during training, convergence termination threshold τ, regularization factor λ a , λ b , λ s , λ d , λ t .

[0024] Preferably, in step S4-2, the tensor decomposition objective loss function constructed using the Lp norm is:

[0025]

[0026] In formula (1), ε represents the target loss function; a represents the linear bias vector of the students; b represents the linear bias vector of the test items; S represents the user latent feature matrix; D represents the test item latent feature matrix; T represents the time latent feature matrix; (i,j,k)∈Γ represents the set of indices of the known elements contained in the tensor X; p represents the order of the Lp norm; x ijk Let X represent a known entity in the student test score data tensor, namely, the score of the i-th student on the j-th test item in the k-th time period; a i b represents the linear deviation vector of the i-th student; j Let M represent the linear deviation vector of the j-th test item; M represents the number of columns in the three latent feature matrices of the student test score tensor X decomposition; s im d represents the value in the i-th row and m-th column of the user latent feature matrix S; jm t represents the value in the j-th row and m-th column of the latent feature matrix D of the test items; km λ represents the value in the k-th row and m-th column of the temporal latent feature matrix T. a , λ b Let λ represent the regularization parameters of a and b, respectively. s , λ d , λ t represent the regularization parameters of the latent feature matrices S, D, and T, respectively.

[0027] Preferably, in S4-3, the formula for training the iteration by calculating the derivative and updating using non-negative multiplication is as follows:

[0028]

[0029] In formula (2), a i b represents the linear deviation vector of the i-th student; j s represents the linear deviation vector of the j-th test item; im d represents the value in the i-th row and m-th column of the user latent feature matrix S; jm t represents the value in the j-th row and m-th column of the latent feature matrix D of the test items; km λ represents the value in the k-th row and m-th column of the temporal latent feature matrix T. a , λ b Let λ represent the regularization parameters of a and b, respectively. s , λ d , λ tLet S, D, and T represent the regularization parameters of the latent feature matrices S, D, and T, respectively; j,k:(i,j,k)∈Γ, i,k:(i,j,k)∈Γ, and i,j:(i,j,k)∈Γ represent the sets of indices of elements contained in the slice matrix of each dimension of the tensor X, respectively; x ijk This represents a known entity in the student test score data tensor X, namely the score of the i-th student on the j-th test item in the k-th time period; This represents the student's test score calculated based on the latent feature matrix and the linear deviation vector.

[0030] Preferably, S5 includes:

[0031] S5-1: Construct input feature vectors for each student using the latent feature matrix. T represents the time latent feature matrix, and S represents the user latent feature matrix;

[0032]

[0033] In formula (4), l i,k S represents the latent feature vector of the i-th student in the k-th time period; J represents the number of test questions; S represents the number of test questions. i D represents the latent feature vector of the i-th student. j Let T represent the latent feature vector of the j-th test. k This represents the latent feature vector for the k-th time period;

[0034] S5-2: Construct the input feature vector Inputting the data into a Long Short-Term Memory (LSTM) network yields the hidden sequence h corresponding to the student. i,k :

[0035] h i,k =LSTM9l i,k-1 ,h i,k-1 (5)

[0036] In formula (5), h i,k The hidden sequence of the i-th student in the k-th time period is represented by LSTM; LSTM represents a Long Short-Term Memory network; i,k-1 h represents the latent feature vector of the i-th student in the (k-1)-th time period; i,k-1 This represents the hidden sequence of the i-th student in the (k-1)-th time period;

[0037] S5-3: The hidden sequence h obtained after the last iteration using the Long Short-Term Memory network unit. i,T+1 The student's cognitive prediction score is calculated and stored using the calibrated test knowledge point mapping matrix G.

[0038] v i,n =G n·sigmoid(Wh i,T+1 +r) (6)

[0039] In formula (6), v i,n G represents the degree of mastery of the i-th student over the n-th knowledge point, i.e., the student's cognitive prediction score; n The mapping matrix for the nth knowledge point represents the known state; sigmoid represents the activation function; h i,T+1 denoted as the hidden sequence obtained after the last iteration of the Long Short-Term Memory network unit; W represents the weight matrix at each iteration; and r represents the linear deviation at each iteration.

[0040] The present invention also provides a student learning cognition prediction device based on bias tensor decomposition, including a data acquisition module, a storage module, a tensor construction module, a linear bias construction module, a latent feature extraction module, and an output module;

[0041] The data acquisition module is used to receive historical student test score data from various educational activities from the server.

[0042] The storage module is used to store the received historical student test scores and student learning cognitive prediction scores;

[0043] The tensor construction module is used to construct student test score data tensors based on stored historical student test score data.

[0044] The linear deviation construction module is used to construct the first linear deviation vector based on the stored historical student test score data;

[0045] The latent feature extraction module is used to construct a target loss function based on the student test score data tensor and the first linear deviation vector, and extract the latent feature matrix.

[0046] The output module is used to output the student's learning and cognitive prediction score based on the extracted latent features.

[0047] In summary, by adopting the above technical solution, the present invention has at least the following beneficial effects compared with the prior art:

[0048] This invention is specifically designed for student learning behavior data. The method can predict student cognitive scores with high accuracy, conforming to educational psychology and learning statistics, thus addressing the dynamic diagnostic problem of student learning data containing time-series information. It can be widely applied in online education platforms, personalized learning recommendation systems, and academic performance assessment. Attached Figure Description

[0049] Figure 1This is a schematic diagram of a student learning cognition prediction method based on bias tensor decomposition according to an exemplary embodiment of the present invention.

[0050] Figure 2 This is a schematic diagram of a student learning cognition prediction device based on bias tensor decomposition according to an exemplary embodiment of the present invention.

[0051] Figure 3 This is a schematic diagram of a storage module according to an exemplary embodiment of the present invention.

[0052] Figure 4 This is a schematic diagram of a latent feature extraction module according to an exemplary embodiment of the present invention.

[0053] Figure 5 This is a schematic diagram of the output module according to an exemplary embodiment of the present invention. Detailed Implementation

[0054] The present invention will be further described in detail below with reference to embodiments and specific implementation methods. However, this should not be construed as limiting the scope of the above-described subject matter of the present invention to the following embodiments; all technologies implemented based on the content of the present invention fall within the scope of the present invention.

[0055] In the description of this invention, it should be understood that the terms "longitudinal", "lateral", "up", "down", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0056] like Figure 1 As shown, this invention provides a student learning cognition prediction method based on bias tensor decomposition, specifically including the following steps:

[0057] S1: Receives instructions from the server to predict students' cognitive learning scores, collects historical student test score data from various educational activities from the server, and stores it.

[0058] The server may send instructions to the student test score data receiving module periodically, or according to a notification from the device, or according to a notification from a server, by sending an instruction to predict student test score data.

[0059] In this embodiment, the received historical student test score data is stored in the form of quadruples; the quadruple is represented as X = (s, d, t, u), where s represents the student, d represents the test question completed by the student, t represents the time period during which student s completed test question d, and u represents the score of student s in completing test question d during the time period t.

[0060] S2: Construct a student test score data tensor based on the stored historical student test score data.

[0061] In this embodiment, the student test score data tensor X∈R is constructed using student test score data X=(s,d,t,u). I×J×K (R represents the set of real numbers); where I is the number of students, J is the number of test questions, and K is the number of time periods, 1≤i≤I, 1≤j≤J, 1≤k≤K. Since it is virtually impossible for each student to complete all the test questions in a given time period, and the score is always positive, the constructed student test score data tensor X is a non-negative sparse tensor. The known data set Γ represents the set of all data in the student test score data tensor X, including the number of students I, the number of test questions J, the number of time periods K, etc.

[0062] S3: Based on the stored historical student test score data, create and store the first linear deviation vector.

[0063] S3-1: Based on the stored historical student test score data, create a linear deviation vector a for all students s that is randomly initialized within a reasonable non-negative range;

[0064] S3-2: Based on the stored historical student test score data, create a randomly initialized linear deviation vector b for all test items d within a reasonable non-negative range.

[0065] S4: Based on the constructed student test score data tensor and the first linear deviation vector, construct the latent features of the student learning cognitive diagnostic score and the loss function of the actual linear deviation target of the student learning cognitive diagnostic score, and output the student test score value.

[0066] S4-1: Initialize the process parameters involved in the prediction of student learning cognitive diagnostic scores.

[0067] In this embodiment, the process parameters include: student test score data tensor X, latent feature matrices S, D, and T, latent feature dimension M, maximum number of training iterations R, iteration number control variable r during training, convergence termination threshold τ, and regularization factor λ. a , λ b , λ s , λ d , λ t .

[0068] in:

[0069] The latent feature dimension M determines the latent feature space dimension of the latent feature matrices S, D, and T, and is initialized to a positive integer;

[0070] The size of the latent feature matrices S, D, and T is determined by the value of each dimension of the corresponding student test score data tensor X and the latent feature dimension M. That is, S is a latent feature matrix with I rows and M columns, D is a latent feature matrix with J rows and M columns, and T is a latent feature matrix with K rows and M columns. The three latent feature matrices are initialized with small random positive numbers respectively.

[0071] The maximum number of training iterations R is a variable that controls the upper limit of the iteration process, and is initialized to a large positive integer;

[0072] The iteration round number control variable r is initialized to 0;

[0073] The convergence termination threshold τ is a parameter used to determine whether the iterative process has converged, and it is initialized with a very small positive number.

[0074] Regularization factor λ a , λ b , λ s , λ d , λ t It is a constant that controls the regularization effect of the first linear deviations a, b and the related elements of the latent feature matrices S, D, T during the extraction iteration process, and is initialized to a small positive number.

[0075] S4-2: Based on the constructed student test score data tensor, process parameters, and first linear deviation vectors a and b, train students to learn the values ​​of the latent feature matrices S, D, and T of cognitive diagnostic scores and the actual linear deviation values ​​of the first linear deviation vectors a and b.

[0076] Among them, the latent feature matrices S, D, and T of the obtained student learning cognitive diagnostic scores and the actual linear deviation values ​​of the first linear deviation vectors a and b are: the known data set Γ of the student test score data tensor X, and the tensor decomposition target loss function ε is constructed, expressed by the following formula:

[0077]

[0078] In formula (1), ε represents the target loss function; a represents the linear bias vector of the students; b represents the linear bias vector of the test items; S represents the user latent feature matrix; D represents the test item latent feature matrix; T represents the time latent feature matrix; (i,j,k)∈Γ represents the set of indices of the known elements contained in the tensor X; p represents the order of the Lp norm; x ijk Let X represent a known entity in the student test score data tensor, namely, the score of the i-th student on the j-th test item in the k-th time period; ai b represents the linear deviation vector of the i-th student; j Let M represent the linear deviation vector of the j-th test item; M represents the number of columns in the three latent feature matrices of the student test score tensor X decomposition; s im d represents the value in the i-th row and m-th column of the user latent feature matrix S; jm t represents the value in the j-th row and m-th column of the latent feature matrix D of the test items; km λ represents the value in the k-th row and m-th column of the temporal latent feature matrix T. a , λ b Let λ represent the regularization parameters of a and b, respectively. s , λ d , λ t represent the regularization parameters of the latent feature matrices S, D, and T, respectively.

[0079] In this embodiment, the Lp norm is used as the optimization objective; Tikhonov regularization is used to constrain the optimization process and prevent overfitting during the optimization process.

[0080] S4-3: To ensure non-negativity during the update process, a multiplicative update rule is used to iteratively optimize the target loss function ε to minimize its value. The training iteration formula is shown below:

[0081]

[0082] In formula (2), a i b represents the linear deviation vector of the i-th student; j s represents the linear deviation vector of the j-th test item; im d represents the value in the i-th row and m-th column of the user latent feature matrix S; jm t represents the value in the j-th row and m-th column of the latent feature matrix D of the test items; km λ represents the value in the k-th row and m-th column of the temporal latent feature matrix T. a , λ b Let λ represent the regularization parameters of a and b, respectively. s , λ d , λ t Let S, D, and T represent the regularization parameters of the latent feature matrices S, D, and T, respectively; j,k:(i,j,k)∈Γ, i,k:(i,j,k)∈Γ, and i,j:(i,j,k)∈Γ represent the sets of indices of elements contained in the slice matrix of each dimension of the tensor X, respectively; x ijk This represents a known entity in the student test score data tensor X, namely the score of the i-th student on the j-th test item in the k-th time period; This represents the student test score calculated based on the latent feature matrix and the linear deviation vector. Let represent the score of the i-th student on the j-th test item during the k-th time period, which is the score of the (p-1)-th order.

[0083]

[0084] In formula (3), Indicates the student's test score; a i b represents the linear deviation vector of the i-th student; j s represents the linear deviation vector of the j-th test item; im d represents the value in the i-th row and m-th column of the user latent feature matrix S; jm t represents the value in the j-th row and m-th column of the latent feature matrix D of the test items; km This represents the value in the k-th row and m-th column of the temporal latent feature matrix T; M represents the number of columns (i.e., the dimension).

[0085] S4-4: Determine whether the training iteration process of the target loss function ε on the known data set Γ has reached the termination condition. If the termination condition is reached, terminate the process and output the latent feature matrices S, D, and T.

[0086] In this step, there are two cases where the training iteration process of the target loss function ε on the known data set Γ reaches the termination condition. The first is that the value of the training iteration number control variable r is incremented by 1 for each iteration of ε. When the value of r reaches the maximum training iteration number L, ε stops training. The second is that during the training process of ε, when the absolute value of the difference between the ε value calculated after the end of the current iteration and the ε value of the previous iteration is less than the convergence termination threshold, ε stops training.

[0087] S5: Using LSTM (Long Short-Term Memory) and the latent feature matrix, output and store the predicted student learning cognitive scores.

[0088] S5-1: After the target loss function ε converges on the known data set Γ, the latent feature matrix is ​​used to construct the input feature vector for each student. l i,k Let T represent the latent feature vector of the i-th student in the k-th time period, T represent the time latent feature matrix, and S represent the user (student) latent feature matrix.

[0089]

[0090] In formula (4), l i,k S represents the latent feature vector of the i-th student in the k-th time period; J represents the number of test questions; S represents the number of test questions. i D represents the latent feature vector of the i-th student. j Let T represent the latent feature vector of the j-th test. k This represents the latent feature vector for the k-th time period;

[0091] S5-2: Construct the input feature vector The input vector is fed into the Long Short-Term Memory (LSTM) layer. The LTM layer receives the input vector and updates its internal state. After a series of recursive gating mechanisms in the LTM layer, the hidden sequence h corresponding to the student is obtained. i,k The expression is:

[0092] h i,k =LSTM(l i,k-1 ,h i,k-1 (5)

[0093] In formula (5), h i,k The hidden sequence of the i-th student in the k-th time period is represented by LSTM; LSTM represents a Long Short-Term Memory network; i,k-1 h represents the latent feature vector of the i-th student in the (k-1)-th time period; i,k-1 Let represent the hidden sequence of the i-th student in the (k-1)-th time period.

[0094] S5-3: The hidden sequence h obtained after the last iteration using the Long Short-Term Memory network unit. i,T+1 The predicted values ​​of students' learning and cognitive scores are calculated and stored using the test knowledge point mapping matrix G, which has been calibrated by experts.

[0095] v i,n =G n ·sigmoid(Wh i,T+1 +r) (6)

[0096] In formula (6), v i,n G represents the degree of mastery of the i-th student over the n-th knowledge point, i.e., the student's cognitive prediction score; n The mapping matrix for the nth knowledge point represents the known state; sigmoid represents the activation function; h i,T+1 denoted as the hidden sequence obtained after the last iteration of the Long Short-Term Memory network unit; W represents the weight matrix at each iteration; and r represents the linear deviation at each iteration.

[0097] Based on the above, a method for predicting student learning cognition based on bias tensor decomposition is proposed, such as... Figure 2 As shown, the present invention also provides a student learning cognition prediction device based on bias tensor decomposition, including a data acquisition module 1, a storage module 2, a tensor construction module 3, a linear bias construction module 4, a latent feature extraction module 5, and an output module 6.

[0098] The output of data acquisition module 1 is connected to the first input of storage module 2. The output of storage module 2 is connected to the input of tensor construction module 3 and the input of linear deviation construction module 4, respectively. The output of tensor construction module 3 and the output of linear deviation construction module 4 are connected in parallel to the input of latent feature extraction module 5. The output of latent feature extraction module 5 is connected to the input of output module 6. The output of output module 6 is connected to the second input of storage module 2.

[0099] The data acquisition module is used to receive historical student test score data from various educational activities from the server.

[0100] Storage module 2 is used to store the received historical student test score data and student learning cognitive prediction scores.

[0101] In this embodiment, as Figure 3 As shown, the storage module 2 includes a first storage unit 21 and a second storage unit 22;

[0102] The first storage unit 21 is used to store the received historical student test score data, and stores it in the form of a quadruple; the quadruple is represented as X = (s, d, t, u), where s represents the student, d represents the test question completed by the student, t represents the time period during which student s completed test question d, and u represents the score of student s in completing test question d during the time period t.

[0103] The second storage unit 22 is used to store the student's learning cognitive prediction score, stored in the form of a triple. The representation is Q = (s, n, v), where s represents the student, n represents the knowledge point number, and v represents the student's mastery of the knowledge point.

[0104] Tensor construction module 3 is used to construct a student test score data tensor: X∈R based on the stored historical student test score data. I×J×K (R represents the set of real numbers); where I is the number of students, J is the number of test questions, K is the number of time periods, 1≤i≤I, 1≤j≤J, 1≤k≤K.

[0105] The linear deviation construction module 4 is used to create a first linear deviation vector based on the stored historical student test score data. The first linear deviation vector includes the linear deviation vector a of all students s and the linear deviation vector b of all test items d.

[0106] The latent feature extraction module 5 is used to construct a target loss function based on the student test score data tensor and the first linear deviation vector to extract latent features.

[0107] like Figure 4As shown, the latent feature extraction module 5 includes an initialization unit 51, a tensor decomposition target loss function construction unit 52, and a training unit 53;

[0108] Initialization unit 51 is used to initialize the process parameters involved in the prediction of students' learning cognitive diagnostic scores.

[0109] Tensor decomposition target loss function construction unit 52 is used to construct tensor decomposition target loss function based on student test score data tensor, first linear deviation vector and process parameters;

[0110] Training unit 53 is used to train and optimize the tensor decomposition target loss function and extract latent features.

[0111] Output module 6 is used to output the student's learning and cognitive prediction score based on the extracted latent features.

[0112] like Figure 5 As shown, the output module 6 includes an input feature vector construction unit 61, a hidden sequence construction unit 62, and an output unit 63.

[0113] Among them, the input feature vector construction unit 61 is used to construct an input feature vector for each student using the extracted latent features;

[0114] Hidden sequence construction unit 62 is used to construct the hidden sequence of the corresponding student based on LSTM (Long Short-Term Memory) network and input feature vector;

[0115] Output unit 63 is used to calculate the student's learning cognitive prediction score based on the student's hidden sequence and the test knowledge point mapping matrix G that has been calibrated by experts.

[0116] Embodiment 2 of this invention provides a method and apparatus for student learning cognitive diagnosis and prediction based on tensor decomposition and long short-term memory networks, specifically designed for student learning behavior data. This method can predict student learning cognitive scores with high accuracy, conforming to educational psychology and learning statistics, thus addressing the dynamic diagnosis problem of student learning data containing temporal information. It can be widely applied in online education platforms, personalized learning recommendation systems, and academic performance evaluation.

[0117] The present invention also provides an electronic device, the electronic device including a processor, the processor being configured to run a computer program stored in a memory, such that the electronic device implements the steps of a student learning cognitive prediction method based on bias tensor decomposition as described in the above embodiments.

[0118] The present invention also provides a computer-readable storage medium storing a computer program, which, when run on a processor, implements the steps of a student learning cognitive prediction method based on bias tensor decomposition as described in the above embodiments.

[0119] Computer programs include computer program code, which can be in the form of source code or...

[0120] Examples of computer-readable media include code, executable files, and certain intermediate forms. Computer-readable media can include at least: any entity or device capable of carrying computer program code to an electronic device, recording media, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media. Examples include USB flash drives, portable hard drives, magnetic disks, or optical discs. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.

[0121] Those skilled in the art will understand that the above embodiments are specific examples of implementing the present invention, and in practical applications, various changes in form and detail may be made without departing from the spirit and scope of the present invention.

Claims

1. A student learning cognition prediction method based on bias tensor decomposition, characterized in that, Specifically, the following steps are included: S1: Collect and store historical student test score data from various educational activities from the server; S2: Construct a student test score data tensor based on the stored historical student test score data; S3: Construct the first linear deviation vector based on the stored historical student test score data; S4: Based on the constructed student test score data tensor and the first linear deviation vector, construct the target loss function and extract the latent feature matrix; S5: Using LSTM and the extracted latent feature matrix, output and store the predicted learning cognitive score.

2. The student learning cognition prediction method based on bias tensor decomposition as described in claim 1, characterized in that, In step S1, the received historical student test score data is stored in the form of a quadruple, which is represented as X = (s, d, t, u), where s represents the student, d represents the test question completed by the student, t represents the time period during which student s completed test question d, and u represents the score of student s in completing test question d during the time period t.

3. The student learning cognition prediction method based on bias tensor decomposition as described in claim 1, characterized in that, In S2, the student test score data tensor is: X∈R I×J×K R represents the set of real numbers, I is the number of students, J is the number of test questions, K is the number of time periods, 1≤i≤I, 1≤j≤J, 1≤k≤K.

4. The student learning cognition prediction method based on bias tensor decomposition as described in claim 1, characterized in that, S3 includes: S3-1: Based on the stored historical student test score data, create a linear deviation vector a for all students s; S3-2: Based on the stored historical student test score data, create a linear deviation vector b for all test items d.

5. The student learning cognition prediction method based on bias tensor decomposition as described in claim 1, characterized in that, S4 includes: S4-1: Initialize the process parameters involved in predicting student learning cognitive diagnostic scores; S4-2: Construct the tensor decomposition objective loss function based on the constructed student test score data tensor, process parameters, and the first linear deviation vector; S4-3: Train and iterate the tensor decomposition objective loss function to extract the latent feature matrix.

6. The student learning cognition prediction method based on bias tensor decomposition as described in claim 5, characterized in that, In S4-1, the process parameters include: Student test score data tensor X, latent feature matrices S, D, T, latent feature dimension M, maximum number of training iterations R, iteration number control variable r during training, convergence termination threshold τ, regularization factor λ a , λ b , λ s , λ d , λ t .

7. The student learning cognition prediction method based on bias tensor decomposition as described in claim 5, characterized in that, In S4-2, the tensor decomposition objective loss function constructed using the Lp norm is as follows: In formula (1), ε represents the target loss function; a represents the linear deviation vector of the students; b represents the linear deviation vector of the test items; S represents the user latent feature matrix; D represents the test item latent feature matrix; and T represents the time latent feature matrix. (i,j,k)∈Γ represents the set of indices of the known elements contained in tensor X; p represents the order of the Lp norm; x ijk Let X represent a known entity in the student test score data tensor, namely, the score of the i-th student on the j-th test item in the k-th time period; a i b represents the linear deviation vector of the i-th student; j Let M represent the linear deviation vector of the j-th test item; M represents the number of columns in the three latent feature matrices of the student test score tensor X decomposition; s im d represents the value in the i-th row and m-th column of the user latent feature matrix S; jm t represents the value in the j-th row and m-th column of the latent feature matrix D of the test items; km λ represents the value in the k-th row and m-th column of the temporal latent feature matrix T. a , λ b Let λ represent the regularization parameters of a and b, respectively. s , λ d , λ t represent the regularization parameters of the latent feature matrices S, D, and T, respectively.

8. The student learning cognition prediction method based on bias tensor decomposition as described in claim 5, characterized in that, In S4-3, the formula for training the iteration by calculating the derivative and updating using non-negative multiplication is as follows: In formula (2), a i b represents the linear deviation vector of the i-th student; j s represents the linear deviation vector of the j-th test item; im d represents the value in the i-th row and m-th column of the user latent feature matrix S; jm t represents the value in the j-th row and m-th column of the latent feature matrix D of the test items; km λ represents the value in the k-th row and m-th column of the temporal latent feature matrix T. a , λ b Let λ represent the regularization parameters of a and b, respectively. s , λ d , λ t Let S, D, and T represent the regularization parameters of the latent feature matrices S, D, and T, respectively; j,k:(i,j,k)∈Γ, i,k:(i,j,k)∈Γ, and i,j:(i,j,k)∈Γ represent the sets of indices of elements contained in the slice matrix of each dimension of the tensor X, respectively; x ijk This represents a known entity in the student test score data tensor X, namely the score of the i-th student on the j-th test item in the k-th time period; This represents the student's test score calculated based on the latent feature matrix and the linear deviation vector.

9. The student learning cognition prediction method based on bias tensor decomposition as described in claim 5, characterized in that, S5 includes: S5-1: Construct input feature vectors for each student using the latent feature matrix. T represents the time latent feature matrix, and S represents the user latent feature matrix; In formula (4), l i,k S represents the latent feature vector of the i-th student in the k-th time period; J represents the number of test questions; S represents the number of test questions. i D represents the latent feature vector of the i-th student. j Let T represent the latent feature vector of the j-th test. k This represents the latent feature vector for the k-th time period; S5-2: Construct the input feature vector Inputting the data into a Long Short-Term Memory (LSTM) network yields the hidden sequence h corresponding to the student. i,k : h i,k =LSTM9l i,k-1 ,h i,k-1 ) (5) In formula (5), h i,k The hidden sequence of the i-th student in the k-th time period is represented by LSTM; LSTM represents a Long Short-Term Memory network; i,k-1 h represents the latent feature vector of the i-th student in the (k-1)-th time period; i,k-1 This represents the hidden sequence of the i-th student in the (k-1)-th time period; S5-3: The hidden sequence h obtained after the last iteration using the Long Short-Term Memory network unit. i,T+1 The student's cognitive prediction score is calculated and stored using the calibrated test knowledge point mapping matrix G. v i,n =G n ·sigmoid(Wh i,T+1 +r) (6) In formula (6), v i,n G represents the degree of mastery of the i-th student over the n-th knowledge point, i.e., the student's cognitive prediction score; n The mapping matrix for the nth knowledge point represents the known state; sigmoid represents the activation function; h i,T+1 denoted as the hidden sequence obtained after the last iteration of the Long Short-Term Memory network unit; W represents the weight matrix at each iteration; and r represents the linear deviation at each iteration.

10. A student learning cognition prediction device based on bias tensor decomposition, based on the method of any one of claims 1-9, characterized in that, It includes a data acquisition module, a storage module, a tensor construction module, a linear deviation construction module, a latent feature extraction module, and an output module; The data acquisition module is used to receive historical student test score data from various educational activities from the server. The storage module is used to store the received historical student test scores and student learning cognitive prediction scores; The tensor construction module is used to construct student test score data tensors based on stored historical student test score data. The linear deviation construction module is used to construct the first linear deviation vector based on the stored historical student test score data; The latent feature extraction module is used to construct a target loss function based on the student test score data tensor and the first linear deviation vector, and extract the latent feature matrix. The output module is used to output the student's learning and cognitive prediction score based on the extracted latent features.

Images