Self-adaptive data enhancement-based depolarization cognitive diagnosis method

Abstract

The invention discloses a depolarization cognitive diagnosis method based on self-adaptive data enhancement, which comprises the following steps of: 1, constructing heterogeneous data which comprises question answering records of students on exercises and a relation matrix of the exercises and knowledge points; 2, generating deviation data through selective data enhancement, inputting the deviation data and student answer records into a deviation cognitive diagnosis model, and outputting a deviation prediction result, thereby constructing symmetric cross entropy loss of the deviation cognitive diagnosis model; 3, based on a prediction result output by the deviation cognitive diagnosis model, adaptively generating unbiased data, inputting the unbiased data and student answer records into the debiased cognitive diagnosis model for processing, and outputting a debiased prediction result so as to construct weighted cross entropy loss of the debiased cognitive diagnosis model; and 4, in a model inference stage, predicting the answering ability of the student by using the established depolarization cognitive diagnosis model. According to the invention, the problem of deviation in cognitive diagnosis can be relieved.

Description

Technical Field

[0001] The present invention belongs to the field of cognitive diagnosis of smart education, and specifically is a debiased cognitive diagnosis method based on adaptive data enhancement. Background Art

[0002] Cognitive diagnosis is a basic task in the field of smart education, which aims to infer students' mastery of knowledge points by using their answer records to exercises. It has been applied to various educational scenarios, such as exercise recommendation, knowledge tracking, and computer adaptive testing.

[0003] Due to the sparsity of answer records, existing cognitive diagnosis methods make full use of answer records to model students. However, these methods inherit the inherent weakness of machine learning methods: they prefer to learn simple features. The simple feature present in the answer record is the student's correct answer rate. For example, the model predicts that students with a higher correct answer rate will always get the exercises right. This will have a negative impact on the accuracy and reliability of cognitive diagnosis, especially when applied to out-of-distribution scenarios. For example, students' regular exams are set by the school, but in the provincial joint exam, the question setter changes, and the question style changes accordingly, resulting in a decrease in the student's correct answer rate. Over-reliance on the student's correct answer rate to model students will lead to biased diagnosis of the model. Therefore, considering the debiasing of cognitive diagnosis for this scenario where there is a bias in modeling student representation using the correct answer rate is a problem worth studying. Summary of the invention

[0004] In order to address the deficiencies in the prior art, the present invention proposes a debiased cognitive diagnosis method based on adaptive data enhancement, in order to alleviate the problem of bias in the cognitive diagnosis model due to modeling based on the accuracy of answering questions, thereby more accurately predicting the weak knowledge points of students with a higher accuracy rate of answering questions and the proficient knowledge points of students with a lower accuracy rate of answering questions, and accurately predicting the students' answering situation.

[0005] In order to achieve the above-mentioned object of the invention, the present invention adopts the following technical scheme:

[0006] The characteristic of the debiasing cognitive diagnosis method based on adaptive data enhancement of the present invention is that it is carried out according to the following steps:

[0007] Step 1: Construct heterogeneous data, including: students’ answer records to exercises And the relationship matrix between exercises and knowledge points ;

[0008] Define the student set as , the exercise set is , the knowledge point set is ;in, Indicates students, Indicates Exercises, Indicates Knowledge points, represents the number of students, represents the number of exercises, represents the number of knowledge points, , , ;

[0009] Record the student's answers to the exercises as ,make Any triple in ,in, Indicates students Answer Exercises The corresponding answer label, if , then it means students Answer Exercises Correct; if , then it means students Answer Exercises mistake;

[0010] The relationship matrix between exercises and knowledge points is recorded as , among which, if , then it means Exercises With Knowledge Points Related; if , then it means Exercises With Knowledge Points Not relevant;

[0011] The first students The set of answered exercises is denoted as ;

[0012] Based on the students' correct answer rate, the correct answer rate exceeds the threshold The students are divided into a group of students with a higher correct answer rate , the correct answer rate does not exceed the threshold The students are divided into a group of students with lower correct answer rates ;

[0013] Will The answer record with the answer label "1" in the corresponding answer record and The answer records with the answer label "0" in the corresponding answer records constitute a typical pattern ;

[0014] Will The answer record with the answer label "0" in the corresponding answer record and The answer records with the answer label "1" in the corresponding answer records constitute an atypical pattern ;

[0015] Step 2: Define the current number of iterations as , and initialize ;

[0016] Step 3: From Typical Mode In the Deviation data of round iteration , and with The bias cognitive diagnosis model is input together for processing and output The deviation prediction results of the first round of iterations are used to construct the deviation cognitive diagnosis model. Symmetric cross entropy loss for round iterations ;

[0017] Step 4: Based on the output of the bias cognitive diagnosis model The prediction results of the first round of iterations are adaptively generated Unbiased data for rounds of iterations , and with The debiased cognitive diagnosis model is input together for processing and the output is The debiased prediction results of the first round of iterations are used to construct the debiased cognitive diagnosis model. Weighted cross entropy loss for each round ;

[0018] Step 5: Use formula (9) to construct The total objective function of the round iteration :

[0019] (9)

[0020] Step 6: Synchronous training of cognitive diagnosis model:

[0021] Step 6.1: Use the Adam optimizer to perform the second optimization on the bias cognitive diagnosis model and the debias cognitive diagnosis model. round of training and minimize the The loss function of the round iteration , to update the model parameter set and obtain The bias cognitive diagnostic model after the first round of training and Debiased cognitive diagnostic model after rounds of training;

[0022] Step 6.2: Assign to Then return to step 6.1 and execute sequentially until Until convergence, thus obtaining a well-established debiased cognitive diagnostic model;

[0023] Step 6.3: In the model inference stage, use the established debiased cognitive diagnosis model to predict students' ability to answer questions, and obtain the optimal answer prediction label corresponding to the students' answers to the exercises to achieve cognitive diagnosis debiasing.

[0024] The characteristic of the debiasing cognitive diagnosis method based on adaptive data enhancement described in the present invention is that step 3 is performed as follows:

[0025] Step 3.1: From the typical mode Randomly select the Answer record of round iteration , will The first iteration students The one-hot encoding and The first iteration Exercise The unique hot encodings of are input into the bias cognitive diagnosis model, and the embedding layer in the bias cognitive diagnosis model outputs the corresponding The first iteration students Characterization and The first iteration Exercise Characterization ;in, Indicates The first iteration students Answer Exercise The corresponding answer label;

[0026] Step 3.2: From the typical mode Randomly select the answer record of the tth iteration , and satisfies , will The first iteration students The one-hot encoding and The first iteration Exercise The unique hot encodings of are input into the bias cognitive diagnosis model, and the embedding layer in the bias cognitive diagnosis model outputs the corresponding The first iteration students Characterization and The first iteration Exercise Characterization ;in, Indicates The first iteration students Answer Exercise The corresponding answer label;

[0027] Step 3.3: Use formula (1) to generate Deviation triples of round iterations , thus obtaining Deviation data of round iteration :

[0028] (1)

[0029] In formula (1), Show obedience The parameters of the distribution, Indicates Deviation students in round iterations The representation of Indicates Deviation Exercises in Iterations The representation of Representing Deviant Students Deviation Exercises Answer record;

[0030] Step 3.4: and They are input into the bias cognitive diagnosis model for processing, and the corresponding output is The first iteration students Answer Exercises The corresponding deviation answer prediction label ;

[0031] Step 3.5: Use formula (2) to construct a bias cognitive diagnosis model in the Symmetric cross entropy loss for round iterations :

[0032] (2)

[0033] In formula (2), Indicates The cross entropy loss of the round iteration, Indicates The reverse cross entropy loss of the round iteration is:

[0034] (3)

[0035] (4).

[0036] Further, step 4 is performed as follows:

[0037] Step 4.1: Calculate the answer record using formula (5) In the Exponential average of iterative cross entropy :

[0038] (5)

[0039] In formula (5), is momentum; Indicates The exponential average of the iterative cross entropy;

[0040] Step 4.2: Use formula (6) to convert the exponential mean Normalize and truncate to get the answer record In the Importance score of the iteration :

[0041] (6)

[0042] In formula (4), Represents the hyperparameter that controls the cutoff threshold;

[0043] Step 4.3: Based on The importance score set of the iteration and answer record , from a multinomial distribution The sampling is obtained Unbiased answer record for each round of iteration ;

[0044] Step 4.4: Unbiased answer record for each round of iteration Randomly select a question record , will The first iteration students The one-hot encoding and The first iteration Exercises The unique hot encodings of are input into the debiased cognitive diagnosis model, and the embedding layer in the debiased cognitive diagnosis model outputs the corresponding The first iteration students Characterization and The first iteration Exercise Characterization ;in, Indicates The first iteration students Answer Exercise The corresponding answer label;

[0045] Step 4.5: Unbiased answer record for each round of iteration Randomly select a question record , and satisfies , will The first iteration students The one-hot encoding and The first iteration Exercise The unique hot encodings of are input into the debiased cognitive diagnosis model, and the embedding layer in the debiased cognitive diagnosis model outputs the corresponding The first iteration students Characterization and The first iteration Exercise Characterization ;in, Indicates The first iteration students answer Exercise The corresponding answer label;

[0046] Step 4.6: Use formula (7) to generate Unbiased triples of round iterations , thus obtaining Unbiased data for rounds of iterations :

[0047] (7)

[0048] In formula (7), Show obedience The parameters of the distribution, Indicates Unbiased student in round iteration The representation of Indicates Unbiased exercises in round iterations The representation of Indicates unbiased students Unbiased Exercises Answer record;

[0049] Step 4.7: and They are input into the debiased cognitive diagnosis model for processing, and the corresponding output is The first iteration students Answer Exercises The corresponding debiased answer prediction label ;

[0050] Step 4.8: Use formula (8) to construct a debiased cognitive diagnosis model in the Weighted cross entropy loss for each round :

[0051] (8).

[0052] The electronic device of the present invention includes a memory and a processor, wherein the memory is used to store a program that supports the processor to execute the debiased cognitive diagnosis method, and the processor is configured to execute the program stored in the memory.

[0053] The present invention provides a computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, and the computer program executes the steps of the debiasing cognitive diagnosis method when the computer program is executed by a processor.

[0054] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0055] 1. The cognitive diagnosis bias model introduced in the present invention can effectively simulate the bias amplification phenomenon: The bias cognitive diagnosis model proposed in the present invention can accurately simulate the situation where the accuracy of students' answers causes deviations in model predictions. In order to address the problem of excessive reliance on answer accuracy in answer modeling, the invention adopts a specific data enhancement strategy to generate biased data, thereby enhancing the model's sensitivity to answer accuracy deviations, and introduces symmetric cross-entropy loss to amplify this bias characteristic. This bias amplification simulation helps educational applications identify model preferences that may lead to misdiagnosis, and helps optimize cognitive diagnosis models to make them more robust.

[0056] 2. The present invention proposes a training strategy for a debiased cognitive diagnostic model assisted by a biased cognitive diagnostic model, which improves the debiasing effect and enhances the accuracy of cognitive diagnosis: in the student modeling process, the cross entropy of each training sample in the biased cognitive diagnostic model is used to identify unbiased samples to guide adaptive data enhancement and sample reweighting, so that the model can more accurately diagnose the weak knowledge points of students with a higher correct answer rate and the proficient knowledge points of students with a lower correct answer rate. It is suitable for out-of-distribution scenarios, realizes cognitive diagnosis debiasing, and thus more comprehensively supports the field of smart education. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 The present invention is a flow chart of a debiasing method based on adaptive data enhancement for cognitive diagnosis. DETAILED DESCRIPTION

[0058] In this embodiment, a debiased cognitive diagnosis method based on adaptive data enhancement is to use the idea of ​​using a biased cognitive diagnosis model to assist a debiased cognitive diagnosis model, and select the answer records for data enhancement of the debiased cognitive diagnosis model according to the entropy of the sample in the biased cognitive diagnosis model, thereby alleviating the bias problem in cognitive diagnosis. Specifically, Figure 1 As shown, the steps are as follows:

[0059] Step 1: Construct heterogeneous data, including: students’ answer records to exercises And the relationship matrix between exercises and knowledge points ;

[0060] Define the student set as , the exercise set is , the knowledge point set is ;in, Indicates students, Indicates Exercises, Indicates Knowledge points, represents the number of students, represents the number of exercises, represents the number of knowledge points, , , ;

[0061] Record the student's answers to the exercises as ,make Any triple in ,in, Indicates students Answer Exercises The corresponding answer label, if , then it means students Answer Exercises Correct; if , then it means students Answer Exercises mistake;

[0062] The relationship matrix between exercises and knowledge points is recorded as , among which, if , then it means Exercises With Knowledge Points Related; if , then it means Exercises With Knowledge Points Not relevant;

[0063] The first students The set of answered exercises is denoted as ;

[0064] Based on the students' correct answer rate, the correct answer rate exceeds the threshold The students are divided into a group of students with a higher correct answer rate , the correct answer rate does not exceed the threshold The students are divided into a group of students with lower correct answer rates ;

[0065] In the specific implementation, the threshold ; It can also be set to other values ​​according to different application scenarios to meet That's it;

[0066] Will The answer record with the answer label "1" in the corresponding answer record and The answer records with the answer label "0" in the corresponding answer records constitute a typical pattern ;

[0067] Will The answer record with the answer label "0" in the corresponding answer record and The answer records with the answer label "1" in the corresponding answer records constitute an atypical pattern ;

[0068] The cognitive diagnostic model relies too much on students’ accuracy in answering questions to predict performance, with students performing well in typical patterns and performing poorly in atypical patterns.

[0069] Step 2: Define the current number of iterations as , and initialize ;

[0070] Step 3: From Typical Mode In the Deviation data of round iteration , and with The bias cognitive diagnosis model is input together for processing and output The deviation prediction results of the first round of iterations are used to construct the deviation cognitive diagnosis model. Symmetric cross entropy loss for round iterations :

[0071] Step 3.1: From the typical mode Randomly select the Answer record of round iteration , will The first iteration students The one-hot encoding and The first iteration Exercise The unique hot encodings of are input into the bias cognitive diagnosis model, and the embedding layer in the bias cognitive diagnosis model outputs the corresponding The first iteration students Characterization and The first iteration Exercise Characterization ;in, Indicates The first iteration students Answer Exercise The corresponding answer label;

[0072] Step 3.2: From the typical mode Randomly select the answer record of the tth iteration , and satisfies , will The first iteration students The one-hot encoding and The first iteration Exercise The unique hot encodings of are input into the bias cognitive diagnosis model, and the embedding layer in the bias cognitive diagnosis model outputs the corresponding The first iteration students Characterization and The first iteration Exercise Characterization ;in, Indicates The first iteration students Answer Exercise The corresponding answer label; in specific implementation, this method is model-independent, and student and exercise representations can be obtained from commonly used cognitive diagnostic models, such as IRT, MIRT, NCDM and KaNCD.

[0073] Step 3.3: Use formula (1) to generate Deviation triples of round iterations ,like Figure 1 As shown in the selected data enhancement module, we get Deviation data of round iteration :

[0074] (1)

[0075] In formula (1), Show obedience The parameters of the distribution, Indicates Deviation students in round iterations The representation of Indicates Deviation Exercises in Iterations The representation of Representing Deviant Students Deviation Exercises Answer record; in specific implementation, Both parameters of the distribution are set to 1.

[0076] Step 3.4: and They are input into the bias cognitive diagnosis model for processing, and the corresponding output is The first iteration students Answer Exercises The corresponding deviation answer prediction label ;

[0077] Step 3.5: In order to further amplify the bias of the bias cognitive diagnosis model, the bias cognitive diagnosis model is constructed using formula (2) in the first Symmetric cross entropy loss for round iterations :

[0078] (2)

[0079] In formula (2), Indicates The cross entropy loss of the round iteration, Indicates The reverse cross entropy loss of the round iteration is:

[0080] (3)

[0081] (4)

[0082] Step 4: Based on the output of the bias cognitive diagnosis model The prediction results of the first round of iterations are adaptively generated Unbiased data for rounds of iterations , and with The debiased cognitive diagnosis model is input together for processing and the output is The debiased prediction results of the first round of iterations are used to construct the debiased cognitive diagnosis model. Weighted cross entropy loss for each round :

[0083] Step 4.1: In order to achieve more stable debiasing, use formula (5) to calculate the answer record In the Exponential average of iterative cross entropy :

[0084] (5)

[0085] In formula (5), is momentum; Indicates The exponential average of the iterative cross entropy; in the specific implementation, momentum ;momentum It can also be used as a hyperparameter.

[0086] Step 4.2: Use formula (6) to convert the exponential mean Normalize and truncate to get the answer record In the Importance score of the iteration :

[0087] (6)

[0088] In formula (4), Represents the hyperparameter that controls the cutoff threshold;

[0089] Step 4.3: Based on The importance score set of the iteration and answer record , from a multinomial distribution The sampling is obtained Unbiased answer record for each round of iteration ; In the specific implementation, when sampling from the multinomial distribution, the importance score is Normalize it so that the sum of its vector elements is .

[0090] Step 4.4: Unbiased answer record for each round of iteration Randomly select a question record , will The first iteration students The one-hot encoding and The first iteration Exercises The unique hot encodings of are input into the debiased cognitive diagnosis model, and the embedding layer in the debiased cognitive diagnosis model outputs the corresponding The first iteration students Characterization and The first iteration Exercise Characterization ;in, Indicates The first iteration students Answer Exercise The corresponding answer label;

[0091] Step 4.5: Unbiased answer record for each round of iteration Randomly select a question record , and satisfies , will The first iteration students The one-hot encoding and The first iteration Exercise The unique hot encodings of are input into the debiased cognitive diagnosis model, and the embedding layer in the debiased cognitive diagnosis model outputs the corresponding The first iteration students Characterization and The first iteration Exercise Characterization ;in, Indicates The first iteration students answer Exercise The corresponding answer label; in specific implementation, this method is model-independent, and student and exercise representations can be obtained from commonly used cognitive diagnostic models, such as IRT, MIRT, NCDM and KaNCD.

[0092] Step 4.6: Use formula (7) to generate Unbiased triples of round iterations ,like Figure 1 The sampling and mixup modules are shown in Figure 1 to obtain Unbiased data for rounds of iterations :

[0093] (7)

[0094] In formula (7), Show obedience The parameters of the distribution, Indicates Unbiased student in round iteration The representation of Indicates Unbiased exercises in round iterations The representation of Indicates unbiased students Unbiased Exercises Answer record;

[0095] Step 4.7: and They are input into the debiased cognitive diagnosis model for processing, and the corresponding output is The first iteration students Answer Exercises The corresponding debiased answer prediction label ; In the specific implementation, The two parameters of the distribution are set as .

[0096] Step 4.8: Use formula (8) to construct a debiased cognitive diagnosis model in the Weighted cross entropy loss for each round :

[0097] (8)

[0098] In the specific implementation, Unbiased data for rounds of iterations The corresponding importance score From formula (7), we can get that formula (7) not only mixes the answer records, but also mixes the importance scores of the answer records.

[0099] Step 5: Use formula (9) to construct The total objective function of the round iteration :

[0100] (9)

[0101] Step 6: Synchronous training of cognitive diagnosis model:

[0102] Step 6.1: According to formula (2), the symmetric cross entropy loss of the bias cognitive diagnosis model is calculated; according to formula (6), the answer record in the first The importance score of the first round of iteration is calculated, and then the weighted cross entropy loss of the debiased cognitive diagnosis model is calculated according to formula (8); the Adam optimizer is used to perform the first round of iteration on the biased cognitive diagnosis model and the debiased cognitive diagnosis model. round of training and minimize the The loss function of the round iteration , to update the model parameter set and obtain The bias cognitive diagnostic model after the first round of training and Debiased cognitive diagnostic model after rounds of training;

[0103] Step 6.2: Assign to Then return to step 6.1 and execute sequentially until Until convergence, thus obtaining a well-established debiased cognitive diagnostic model;

[0104] Step 6.3: In the model inference stage, use the established debiased cognitive diagnosis model to predict students' ability to answer questions, and obtain the optimal answer prediction label corresponding to the students' answers to the exercises to achieve cognitive diagnosis debiasing.

[0105] In this embodiment, an electronic device includes a memory and a processor, wherein the memory is used to store a program that supports the processor to execute the above method, and the processor is configured to execute the program stored in the memory.

[0106] In this embodiment, a computer-readable storage medium stores a computer program on the computer-readable storage medium, and the computer program executes the steps of the above method when executed by a processor.

[0107] Example:

[0108] In order to evaluate the performance of the method, experiments were conducted on three datasets: ASSIST2009, NIPS2020EC, and PISA2018. First, the datasets were randomly divided into It is divided into training set, validation set and test set. In order to verify the debiasing performance of the model, some typical patterns in the test set are randomly removed, so that the ratio of typical patterns to atypical patterns in the test set is If the model relies too much on the prediction of the correct answer rate, it will not perform well in the test set, which can verify the debiasing performance of the model. Three indicators were used in the test set: Area Under Curve (AUC), Accuracy (ACC) and Root Mean Square Error (RMSE).

[0109] This paper selects four cognitive diagnosis models as the backbone models, namely Item Response Theory (IRT), Multidimensional Item Response Theory (MIRT), Neural Cognitive Diagnosis Model (NCDM) and Knowledge-association based Neural Cognitive Diagnosis (KaNCD), and selects four debiasing methods for effect comparison, namely Reweight, Learning from Failure (LfF), Selected Mixup (SelecMix) and BiasEnsemble (BE). Specifically, Table 1 shows the experimental results. Among them, the best score is shown in bold.

[0110] Table 1 Comparison of diagnostic results between the method of the present invention and the comparative method

[0111]

[0112] It can be seen from Table 1 that on the three data sets, the MixCD method proposed in the present invention is significantly better than the comparison method in terms of AUC, ACC and RMSE.

Claims

1. A debiased cognitive diagnosis method based on adaptive data enhancement, characterized in that: The steps are as follows: Step 1: Construct heterogeneous data, including: students’ answer records to exercises And the relationship matrix between exercises and knowledge points ; Define the student set as , the exercise set is , the knowledge point set is ;in, Indicates students, Indicates Exercises, Indicates Knowledge points, represents the number of students, represents the number of exercises, represents the number of knowledge points, , , ; The student's answer record for the exercise is recorded as ,make Any triple in ,in, Indicates students Answer Exercises The corresponding answer label, if , then it means students Answer Exercises Correct; if , then it means students Answer Exercises mistake; The relationship matrix between exercises and knowledge points is recorded as , among which, if , then it means Exercises With Knowledge Points Related; if , then it means Exercises With Knowledge Points Not relevant; The first students The set of answered exercises is denoted as ; Based on the students' correct answer rate, the correct answer rate exceeds the threshold The students are divided into a group of students with a higher correct answer rate , the correct answer rate does not exceed the threshold The students are divided into a group of students with lower correct answer rates ; Will The answer record with the answer label "1" in the corresponding answer record and The answer records with the answer label "0" in the corresponding answer records form a typical pattern ; Will The answer record with the answer label "0" in the corresponding answer record and The answer records with the answer label "1" in the corresponding answer records constitute an atypical pattern ; Step 2: Define the current number of iterations as , and initialize ; Step 3: From Typical Mode In the Deviation data of round iteration , and The bias cognitive diagnosis model is input together for processing and output The deviation prediction results of the first round of iterations are used to construct the deviation cognitive diagnosis model. Symmetric cross entropy loss for round iterations ; Step 4: Based on the output of the bias cognitive diagnosis model The prediction results of the first round of iterations are adaptively generated Unbiased data for rounds of iterations , and The debiased cognitive diagnosis model is input together for processing and the output is The debiased prediction results of the first round of iterations are used to construct the debiased cognitive diagnosis model. Weighted cross entropy loss for each round ; Step 5: Use formula (9) to construct The total objective function of the round iteration : (9) Step 6: Synchronous training of cognitive diagnosis model: Step 6.1: Use the Adam optimizer to perform the second optimization on the bias cognitive diagnosis model and the debias cognitive diagnosis model. round of training and minimize the The loss function of the round iteration , to update the model parameter set and obtain The bias cognitive diagnostic model after the first round of training and Debiased cognitive diagnostic model after rounds of training; Step 6.2: Assign to Then return to step 6.1 and execute sequentially until Until convergence, thus obtaining a well-established debiased cognitive diagnostic model; Step 6.3: In the model inference stage, use the established debiased cognitive diagnosis model to predict students' ability to answer questions, and obtain the optimal answer prediction label corresponding to the students' answers to the exercises to achieve cognitive diagnosis debiasing.

2. The method for debiasing cognitive diagnosis based on adaptive data enhancement according to claim 1, characterized in that: The step 3 is carried out as follows: Step 3.1: From the typical mode Randomly select the Answer record of round iteration , will The first iteration students The one-hot encoding and The first iteration Exercise The unique hot encodings of are input into the bias cognitive diagnosis model, and the embedding layer in the bias cognitive diagnosis model outputs the corresponding The first iteration students Characterization and The first iteration Exercise Characterization ;in, Indicates The first iteration students Answer Exercise The corresponding answer label; Step 3.2: From the typical mode Randomly select the answer record of the tth iteration , and satisfies , will The first iteration students The one-hot encoding and The first iteration Exercise The unique hot encodings of are input into the bias cognitive diagnosis model, and the embedding layer in the bias cognitive diagnosis model outputs the corresponding The first iteration students Characterization and The first iteration Exercise Characterization ;in, Indicates The first iteration students Answer Exercise The corresponding answer label; Step 3.3: Use formula (1) to generate Deviation triples of round iterations , thus obtaining Deviation data of round iteration : (1) In formula (1), Show obedience The parameters of the distribution, Indicates Deviation students in round iterations The representation of Indicates Deviation Exercises in Iterations The representation of Denotes a student with deviations Deviation Exercises Answer record; Step 3.4: and They are input into the bias cognitive diagnosis model for processing, and the corresponding output is The first iteration students Answer Exercises The corresponding deviation answer prediction label ; Step 3.5: Use formula (2) to construct a bias cognitive diagnosis model in the Symmetric cross entropy loss for round iterations : (2) In formula (2), Indicates The cross entropy loss of the round iteration, Indicates The reverse cross entropy loss of the round iteration is: (3) (4)。 3. The method for debiasing cognitive diagnosis based on adaptive data enhancement according to claim 2, characterized in that: The step 4 is carried out as follows: Step 4.1: Calculate the answer record using formula (5) In the Exponential average of iterative cross entropy : (5) In formula (5), is momentum; Indicates The exponential average of the iterative cross entropy; Step 4.2: Use formula (6) to convert the exponential mean Normalize and truncate to get the answer record In the Importance score of the iteration : (6) In formula (4), Represents the hyperparameter that controls the cutoff threshold; Step 4.3: Based on The importance score set of the round iteration and answer record , from a multinomial distribution The sampling is obtained Unbiased answer record for each round of iteration ; Step 4.4: Unbiased answer record for each round of iteration Randomly select a question record , will The first iteration students The one-hot encoding and The first iteration Exercises The unique hot encodings of are input into the debiased cognitive diagnosis model, and the embedding layer in the debiased cognitive diagnosis model outputs the corresponding The first iteration students Characterization and The first iteration Exercise Characterization ;in, Indicates The first iteration students Answer Exercise The corresponding answer label; Step 4.5: Unbiased answer record for each round of iteration Randomly select a question record , and satisfies , will The first iteration students The one-hot encoding and The first iteration Exercise The unique hot encodings of are input into the debiased cognitive diagnosis model, and the embedding layer in the debiased cognitive diagnosis model outputs the corresponding The first iteration students Characterization and The first iteration Exercise Characterization ;in, Indicates The first iteration students answer Exercise The corresponding answer label; Step 4.6: Use formula (7) to generate Unbiased triples of round iterations , thus obtaining Unbiased data for rounds of iterations : (7) In formula (7), Show obedience The parameters of the distribution, Indicates Unbiased student in round iteration The representation of Indicates Unbiased exercises in round iterations The representation of Indicates unbiased students Unbiased Exercises Answer record; Step 4.7: and They are input into the debiased cognitive diagnosis model for processing, and the corresponding output is The first iteration students Answer Exercises The corresponding debiased answer prediction label ; Step 4.8: Use formula (8) to construct a debiased cognitive diagnosis model in the Weighted cross entropy loss for each round : (8)。 4. An electronic device, comprising a memory and a processor, characterized in that: The memory is used to store a program that supports the processor to execute the debiased cognitive diagnosis method described in any one of claims 1 to 3, and the processor is configured to execute the program stored in the memory.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the debiased cognitive diagnosis method described in any one of claims 1 to 3 are performed.

Images