A student cognitive diagnosis method based on data forgetting

Abstract

The application discloses a student cognitive diagnosis method based on data forgetting, first acquires the cognitive state information of students through model pre-training, and carries out data partitioning on the students according to similarity, and then independently carries out model training on each partition. When student score prediction is carried out, the similarity between the student and other students in each partition is calculated, and then the prediction weight of each partition model is adjusted, and finally the score prediction result is aggregated and output. After the sensitive data is deleted, the forgetting framework can quickly retrain the partition data, and the training time is significantly reduced. The forgetting framework is constructed by the cognitive state of students on knowledge points, so that the model is prevented from being disturbed by dirty data and invalid data, and the cognitive diagnosis is more accurately completed. Through timely removal of obsolete data, the current cognitive state of students can be more accurately diagnosed.

Description

Technical Field

[0001] This invention relates to the field of educational big data, and in particular to a student cognitive diagnosis method based on data forgetting. Background Technology

[0002] With the rapid development of information technology, the scale of educational data is constantly expanding, covering various aspects such as students' learning behavior data, exam scores, and feedback. The generation and application of educational big data have brought unprecedented opportunities and challenges to personalized learning. How to effectively manage and utilize educational big data, and securely and efficiently mine the information contained within it, has become a key focus in the development of educational models. Protecting student user privacy data requires not only preventing data leakage but also considering the issue of forgotten privacy data to prevent incomplete deletion of privacy data parameters during model training.

[0003] Cognitive diagnostic technology, as an advanced educational assessment tool, helps educators accurately identify students' learning weaknesses and provide targeted feedback through detailed analysis of their knowledge acquisition and cognitive abilities during the learning process. In contrast, traditional examination and assessment methods struggle to comprehensively and meticulously reflect students' cognitive states, often posing the challenge for teachers to promptly understand students' learning progress and effectively address their weaknesses. Applying cognitive diagnostic technology to student learning assessment not only provides a deeper understanding of students' mastery of specific knowledge points but also offers a scientific basis for personalized teaching and learning path design.

[0004] With the increasing number of student users, cognitive diagnostic models are being used more and more widely in education systems, but this has also brought significant information security risks. A large amount of students' personal information and learning behavior data are centrally stored and processed, making the system face a greater risk of data leakage. Traditional data protection measures are difficult to cope with the complex scenarios of large-scale data storage and model parameter updates, which means that while cognitive diagnostic models provide accurate educational services, potential security vulnerabilities may endanger students' privacy and data security. While updating and iterating on users' privacy data, existing cognitive diagnostic models have failed to delete model parameters related to privacy data, facing the risk of privacy leakage. In addition, cognitive models face the following major challenges when performing forgetting on privacy data: (1) The impact of data deletion on model accuracy: Cognitive diagnostic models rely on a large amount of historical data to build and optimize model structures. Once some data is deleted, the model's predictive ability may decrease, leading to a decrease in overall performance; (2) The efficiency of model retraining: After performing data forgetting, the model may need to be retrained or fine-tuned to adapt to the new state after data deletion. Especially when dealing with large-scale data, this process can be very time-consuming; (3) Difficulty in completely deleting data: In complex cognitive diagnostic models, data often exists in multiple forms, such as direct input, derived features, model parameters, etc. To completely delete a user's data, it may be necessary to operate at multiple levels of the model, which makes it difficult to guarantee the integrity of data forgetting. How to ensure that all privacy data related to users is completely cleared is the core issue in achieving data forgetting; (4) Fulfillment of compliance requirements: Under different regions and legal frameworks, the requirements for data forgetting may be different, and cognitive diagnostic models need to comply with the provisions of these laws and regulations. How to take into account different compliance requirements in the technical implementation and ensure the legality of data forgetting is an important challenge.

[0005] Against this backdrop, machine forgetting learning, as an emerging technological approach, is gradually becoming an important means of solving privacy protection and data management problems. Machine forgetting learning aims to effectively remove specific data records from a model to prevent these records from affecting the model's training and prediction. This technology is significant in protecting user privacy and complying with data protection regulations (such as GDPR). For example, when students request the deletion of their personal data, how to efficiently perform data forgetting while effectively protecting user privacy and meeting increasingly stringent data regulations, without causing a significant decline in model performance after forgetting, has become a pressing problem to be solved. Summary of the Invention

[0006] The purpose of this invention is to provide a student cognitive diagnosis method based on data forgetting, which protects privacy while ensuring the accuracy and stability of the prediction model, and provides strong technical support for personalized learning for students.

[0007] The technical solution adopted in this invention is:

[0008] A student cognitive diagnostic method based on data forgetting includes the following steps:

[0009] S1, model the student's answer information using the selected pre-training algorithm to obtain the student's current cognitive state, and process the student's current cognitive state into a knowledge point vector that reflects the student's mastery of knowledge points;

[0010] Knowledge point vector this The students' proficiency with each knowledge point was recorded and then stored for use in subsequent steps;

[0011] S2, Select N Using seed users as the center of a subset of data, student knowledge point vectors are calculated. this The similarity of cognitive states between other users and the central user is used to divide user data into categories. N Sub-datasets.

[0012] S3. Use the selected student cognitive diagnostic algorithm to train the model, train an independent sub-model for each subset of data, and ensure that each sub-model learns the data in its corresponding subset of data independently.

[0013] S4. The outputs of each sub-model are aggregated using preset weights to generate a preliminary score prediction model.

[0014] S5, based on the data forgetting request, delete the relevant data in the sub-model containing privacy data and the corresponding sub-dataset;

[0015] S6, retrain the sub-model modified due to privacy data deletion using the selected student cognitive diagnostic algorithm, and then re-aggregate the sub-models according to the current sub-model weights;

[0016] S7. When predicting the score of a specified student, the pre-training algorithm is first used to obtain the latest cognitive status of the specified student and generate the corresponding student knowledge point vector.

[0017] S8, check if there are relevant answer records for the specified student in the training dataset; if yes, proceed to S9; otherwise, proceed directly to S10;

[0018] S9: Locate the subset of data containing the specified student, set the corresponding sub-model weight to 1, and set the weights of other sub-models to 0; aggregate the predicted scores of each sub-model based on the current sub-model weights.

[0019] S10, calculate the cognitive similarity between the specified student and the central user of each sub-model, and dynamically adjust the weights of each sub-model based on the cognitive similarity;

[0020] S11, each sub-model predicts the score of the specified student, and then dynamically adjusts the predicted score according to the weight of the sub-model to obtain the final predicted score of the specified student.

[0021] Furthermore, in step S5, a data forgetting request is generated when privacy data needs to be deleted, in order to determine the location of the sub-model and sub-dataset where the privacy data is located.

[0022] Furthermore, in step S1, the pre-training algorithm is used. algorithm 1 Obtain the student's cognitive state and generate a sequence of length [length missing]. K Pre-training student knowledge point vector this ; in the knowledge point vector K Each element corresponds to a student's... K The student's proficiency in each knowledge point; the value of each element represents the student's mastery of that knowledge point, thus reflecting the student's overall cognitive state.

[0023] Furthermore, step S2 specifically includes the following steps:

[0024] S2-1, firstly, utilize the student knowledge point vectors generated during pre-training. this The Pearson correlation coefficient between each remaining user and each center user was calculated to measure the cognitive similarity among users.

[0025] S2-2, Based on the calculated cognitive similarity, sort the remaining users in descending order of their similarity to users of each center;

[0026] S2-3, add user information to the corresponding subsets sequentially until the size of the subsets reaches the preset maximum size. S max And so on, until completion. N The dataset is divided into several subsets. This ensures that each subset contains user data with the highest cognitive similarity to the central user, thereby improving the training effect and prediction accuracy of the sub-models.

[0027] Furthermore, in steps S1 and S7, a pre-training algorithm is used on the training dataset. SThe purpose of model pre-training is to generate pre-trained student knowledge point vectors. this This allows for more accurate dataset partitioning in subsequent steps. A cognitive diagnostic algorithm is used. algorithm 2 Sub-models are trained on various subsets of datasets to generate score prediction models. Employing different cognitive diagnostic algorithms at different stages optimizes computational resource usage, especially on large-scale datasets, making the entire training process more efficient. Furthermore, it reduces the likelihood of model overfitting and dependence on specific algorithms.

[0028] Furthermore, step S5 specifically includes the following steps:

[0029] S5-1 First, identify the user data that needs to be forgotten, including users' personal information, answer records or other sensitive data, and accurately locate the data in the model using a unique identifier (such as user ID or record ID);

[0030] S5-2, determine the subset and sub-model where the forgotten user data is located, and generate the corresponding forgetting request;

[0031] Furthermore, the forget request should include key information such as the data identifier to be deleted, the subset containing the data, and the relevant sub-model.

[0032] Furthermore, to ensure the traceability of the data deletion process, the generated forgetting requests are recorded in a log for future review and verification.

[0033] Furthermore, step S10 specifically includes the following steps:

[0034] S10-1 First, based on the calculated similarity, a non-linear function is used to map the similarity to weights, ensuring that sub-models with higher similarity are assigned greater weights;

[0035] S10-2, normalize the initially assigned weights to ensure that the sum of all weights is 1.

[0036] Furthermore, a regularization term is introduced to prevent certain sub-models from having excessively high or low weights, thereby avoiding impacting overall prediction performance. In each new prediction task, the weights are dynamically adjusted based on the latest student cognitive state and historical feedback data to continuously optimize prediction results.

[0037] Furthermore, in step S11, each sub-model performs a weighted average of the predicted scores of the specified students according to the adjusted weights to generate the final predicted score.

[0038] This invention employs the above technical solutions, enabling cognitive diagnostic models to rapidly perform data forgetting or model updates while ensuring user privacy and security, better adapting to the high update frequency of educational big data. This method, based on pre-training and sub-model partitioning strategies, and dynamically adjusting sub-model weights, not only effectively alleviates the performance degradation caused by data deletion but also significantly reduces the cost of model data updates. As students' cognitive states change and historical data accumulates, this invention can continuously optimize prediction results. Beyond utility, this invention ensures the thoroughness of privacy data deletion by generating forgetting requests and precisely locating the data to be deleted. Furthermore, the introduced dynamic weight adjustment and retraining mechanism guarantees model performance after privacy data deletion. Through similarity-based user data partitioning and dynamic adjustment mechanisms, this invention protects privacy while ensuring the accuracy and stability of the prediction model, providing strong technical support for personalized student learning. Attached Figure Description

[0039] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments;

[0040] Figure 1 A structural diagram of data forgetting methods for the cognitive diagnostic process of students;

[0041] Figure 2 A flowchart illustrating the data forgetting factor score prediction process for student cognitive diagnostics.

[0042] Figure 3 This is a flowchart illustrating the forgetting process of a data forgetting method for student cognitive diagnostics. Implementation

[0043] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0044] like Figure 1 to Figure 3 As shown in the figure, the present invention discloses a student cognitive diagnosis method based on data forgetting, comprising the following steps:

[0045] S1, model the student's answer information using the selected pre-training algorithm to obtain the student's current cognitive state, and process the student's current cognitive state into a knowledge point vector that reflects the student's mastery of knowledge points;

[0046] Knowledge point vector this The students' proficiency with each knowledge point was recorded and then stored for use in subsequent steps;

[0047] S2, Select NUsing seed users as the center of a subset of data, student knowledge point vectors are calculated. this The similarity of cognitive states between other users and the central user is used to divide user data into categories. N Sub-datasets.

[0048] S3. Use the selected student cognitive diagnostic algorithm to train the model, train an independent sub-model for each subset of data, and ensure that each sub-model learns the data in its corresponding subset of data independently.

[0049] S4. The outputs of each sub-model are aggregated using preset weights to generate a preliminary score prediction model.

[0050] S5, based on the data forgetting request, delete the relevant data in the sub-model containing privacy data and the corresponding sub-dataset;

[0051] S6, retrain the sub-model modified due to privacy data deletion using the selected student cognitive diagnostic algorithm, and then re-aggregate the sub-models according to the current sub-model weights;

[0052] S7. When predicting the score of a specified student, the pre-training algorithm is first used to obtain the latest cognitive status of the specified student and generate the corresponding student knowledge point vector.

[0053] S8, check if there are relevant answer records for the specified student in the training dataset; if yes, proceed to S9; otherwise, proceed directly to S10;

[0054] S9: Locate the subset of data containing the specified student, set the corresponding sub-model weight to 1, and set the weights of other sub-models to 0; aggregate the predicted scores of each sub-model based on the current sub-model weights.

[0055] S10, calculate the cognitive similarity between the specified student and the central user of each sub-model, and dynamically adjust the weights of each sub-model based on the cognitive similarity;

[0056] S11, each sub-model predicts the score of the specified student, and then dynamically adjusts the predicted score according to the weight of the sub-model to obtain the final predicted score of the specified student.

[0057] Furthermore, in step S5, a data forgetting request is generated when privacy data needs to be deleted, in order to determine the location of the sub-model and sub-dataset where the privacy data is located.

[0058] Furthermore, in step S1, the pre-training algorithm is used. algorithm 1 Obtain the student's cognitive state and generate a sequence of length [length missing]. K Pre-training student knowledge point vector this ; in the knowledge point vectorK Each element corresponds to a student's... K The student's proficiency in each knowledge point; the value of each element represents the student's mastery of that knowledge point, thus reflecting the student's overall cognitive state.

[0059] Furthermore, step S2 specifically includes the following steps:

[0060] S2-1, firstly, utilize the student knowledge point vectors generated during pre-training. this The Pearson correlation coefficient between each remaining user and each center user was calculated to measure the cognitive similarity among users.

[0061] S2-2, Based on the calculated cognitive similarity, sort the remaining users in descending order of their similarity to users of each center;

[0062] S2-3, add user information to the corresponding subsets sequentially until the size of the subsets reaches the preset maximum size. S max And so on, until completion. N The dataset is divided into several subsets. This ensures that each subset contains user data with the highest cognitive similarity to the central user, thereby improving the training effect and prediction accuracy of the sub-models.

[0063] Furthermore, in steps S1 and S7, a pre-training algorithm is used on the training dataset. S The purpose of model pre-training is to generate pre-trained student knowledge point vectors. this This allows for more accurate dataset partitioning in subsequent steps. A cognitive diagnostic algorithm is used. algorithm 2 Sub-models are trained on various subsets of datasets to generate score prediction models. Employing different cognitive diagnostic algorithms at different stages optimizes computational resource usage, especially on large-scale datasets, making the entire training process more efficient. Furthermore, it reduces the likelihood of model overfitting and dependence on specific algorithms.

[0064] Furthermore, step S5 specifically includes the following steps:

[0065] S5-1 First, identify the user data that needs to be forgotten, including users' personal information, answer records or other sensitive data, and accurately locate the data in the model using a unique identifier (such as user ID or record ID);

[0066] S5-2, determine the subset and sub-model where the forgotten user data is located, and generate the corresponding forgetting request;

[0067] Furthermore, the forget request should include key information such as the data identifier to be deleted, the subset containing the data, and the relevant sub-model.

[0068] Furthermore, to ensure the traceability of the data deletion process, the generated forgetting requests are recorded in a log for future review and verification.

[0069] Furthermore, step S10 specifically includes the following steps:

[0070] S10-1 First, based on the calculated similarity, a non-linear function is used to map the similarity to weights, ensuring that sub-models with higher similarity are assigned greater weights;

[0071] S10-2, normalize the initially assigned weights to ensure that the sum of all weights is 1.

[0072] Furthermore, a regularization term is introduced to prevent certain sub-models from having excessively high or low weights, thereby avoiding impacting overall prediction performance. In each new prediction task, the weights are dynamically adjusted based on the latest student cognitive state and historical feedback data to continuously optimize prediction results.

[0073] Furthermore, in step S11, each sub-model performs a weighted average of the predicted scores of the specified students according to the adjusted weights to generate the final predicted score.

[0074] The specific principles of this invention will be explained in detail below:

[0075] like Figure 1 As shown, most student cognitive diagnostic models are divided into four main parts according to the process: pre-training, data partitioning based on cognitive similarity, sub-model training, and score prediction aggregation. 1. Pre-training. Objective: To generate vectors reflecting students' mastery of knowledge points. this Process: The selected pre-training algorithm models the student's answer information to obtain the student's current cognitive state, which is then processed into a vector reflecting the student's mastery of knowledge points. this These vectors record students' proficiency levels for various knowledge points and are then stored for use in subsequent steps; 2. Data segmentation based on cognitive similarity. Objective: To divide user data into several subsets. Process: Determine the number of sub-models. N and the maximum capacity of each subset of data S max Select N Using seed users as the center of a subset of data, student knowledge point vectors are calculated. this The similarity of cognitive states between other users and these central users is used to divide user data into categories. N3. Sub-model training section. Objective: To train an independent sub-model for each subset of datasets. Process: Using a student cognitive diagnostic algorithm. algorithm 2 4. Score Prediction Aggregation. Objective: To generate a student score prediction model. Process: After completing the sub-model training, the outputs of each sub-model are aggregated using preset weights to generate a preliminary score prediction model. This model combines the predictive capabilities of all sub-models and improves the overall prediction performance by dynamically adjusting the weights.

[0076] For the score prediction process of data forgetting methods for student cognitive diagnostic processes, see [link to relevant documentation]. Figure 2 The flowchart shown is shown.

[0077] 1. Student status update. This is achieved through a pre-trained algorithm. algorithm 1 Get the specified student u The latest cognitive state, and transform it into a vector of student knowledge points. this u The length of this vector is K Each element represents a student's level of mastery of a specific knowledge point. 2. Record checking and weight adjustment. The system checks students'... u Does the training dataset contain historical answer records? If so, directly find the subset of data containing that student, set the weight of the corresponding sub-model to 1, and set the weights of other sub-models to 0. This maximizes the use of students' historical data for prediction and reduces interference from irrelevant data. If no historical records exist, calculate the student's answer history. u The similarity between the cognitive state of the students and the central users of each sub-model is mapped to weights using a non-linear mapping function. This ensures that students with high similarity receive higher weights, thus influencing the final prediction results. Dynamic weight updates help the model adjust its prediction strategy based on the latest cognitive state. 3. Score Prediction and Aggregation. Each sub-model independently predicts student scores based on its current weights. u The student's score is generated by weighting the predicted scores of each sub-model (using regularization to prevent excessive influence from a single sub-model). u The final prediction score is calculated. This process ensures the robustness and accuracy of the prediction results.

[0078] For data forgetting methods aimed at students' cognitive diagnostic processes, see [link to relevant documentation]. Figure 3 The flowchart shown is shown.

[0079] 1. Forgot Request Generation. When it's necessary to delete a user's private data, a forget request is generated first. A unique identifier (such as a user ID or record ID) is used to accurately locate the data to be deleted. This data may include sensitive information such as personal information and quiz records. The forget request should record the identifier of the data to be deleted, the subset containing the data, and the relevant sub-model information. This information will be used in subsequent steps to accurately delete the data. 2. Data Location and Deletion. Based on the forget request, the subset containing the private data and the sub-model are located. Subsequently, a specially designed deletion algorithm is used to ensure that this data is completely deleted, avoiding any potential recovery possibilities. To ensure the traceability of the entire data forgetting process, the system logs each generated forget request. This helps in future reviews of the completeness and accuracy of data deletion. 3. Sub-model Retraining and Re-aggregation. Since data deletion results in an incomplete training dataset for the sub-models, the affected sub-models need to be retrained. During retraining, the same training dataset as the initial training is used. algorithm 2 This ensures the new model can adapt to changes in data. The retrained sub-models will be aggregated again based on their current weights. In this way, the system can maintain the overall model's predictive power and performance stability while removing privacy-sensitive data.

[0080] This invention integrates machine forgetting learning technology into a cognitive diagnostic model, enabling the complete deletion of user data and model parameters. This enhances privacy protection while ensuring the system complies with relevant laws and regulations. Furthermore, the technology can dynamically adjust data distribution and remove outdated or invalid data, further improving the accuracy and efficiency of the diagnostic model. More importantly, the forgetting mechanism optimizes the system's storage and computing resources, allowing it to maintain high efficiency even when processing large amounts of data. Applying this technology, the diagnostic model can not only handle scenarios with frequently updated student data but also track students' cognitive states in more real-time, expanding its application scenarios in the field of educational technology.

[0081] Obviously, the described embodiments are only a part of the embodiments of this application, not all of them. Without conflict, the embodiments and features in the embodiments of this application can be combined with each other. The components of the embodiments of this application described and illustrated herein can generally be arranged and designed in various different configurations. Therefore, the detailed description of the embodiments of this application is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

Claims

1. A student cognitive diagnosis method based on data forgetting, characterized in that: It includes the following steps: S1, model the student's answer information using the selected pre-training algorithm to obtain the student's current cognitive state, and process the student's current cognitive state into a knowledge point vector that reflects the student's mastery of knowledge points; Knowledge point vector stu The students' proficiency with each knowledge point was recorded and then stored for use in subsequent steps; In step S1, the pre-training algorithm is used. algorithm 1 Obtain the student's cognitive state and generate a sequence of length [length missing]. K Pre-training student knowledge point vector stu ; in the knowledge point vector K Each element corresponds to a student's... K The proficiency level of each knowledge point; the value of each element represents the student's mastery of that knowledge point, thus reflecting the student's overall cognitive state; S2, Select N Using seed users as the center of a subset of data, student knowledge point vectors are calculated. stu The similarity of cognitive states between other users and the central user is used to divide user data into categories. N Each subset of datasets; Step S2 specifically includes the following steps: S2-1, firstly, utilize the student knowledge point vectors generated during pre-training. stu The Pearson correlation coefficient between each remaining user and each center user was calculated to measure the cognitive similarity among users. S2-2, Based on the calculated cognitive similarity, sort the remaining users in descending order of their similarity to users of each center; S2-3, add user information to the corresponding subsets sequentially until the size of the subsets reaches the preset maximum size. S max And so on, until completion. N The partitioning of the dataset into subsets; S3. Use the selected student cognitive diagnostic algorithm to train the model, train an independent sub-model for each subset of data, and ensure that each sub-model learns the data in its corresponding subset of data independently. S4. The outputs of each sub-model are aggregated using preset weights to generate a preliminary score prediction model. S5, based on the data forgetting request, delete the relevant data in the sub-model containing privacy data and the corresponding sub-dataset; S6, retrain the sub-model modified due to privacy data deletion using the selected student cognitive diagnostic algorithm, and then re-aggregate the sub-models according to the current sub-model weights; S7. When predicting the score of a specified student, the pre-training algorithm is first used to obtain the latest cognitive status of the specified student and generate the corresponding student knowledge point vector. S8, check if there are relevant answer records for the specified student in the training dataset; if yes, proceed to S9; otherwise, proceed directly to S10; S9: Locate the subset of data containing the specified student, set the corresponding sub-model weight to 1, and set the weights of other sub-models to 0; aggregate the predicted scores of each sub-model based on the current sub-model weights. S10, calculate the cognitive similarity between the specified student and the central user of each sub-model, and dynamically adjust the weights of each sub-model based on the cognitive similarity; S11, each sub-model predicts the score of the specified student, and then dynamically adjusts the predicted score according to the weight of the sub-model to obtain the final predicted score of the specified student.

2. The student cognitive diagnosis method based on data forgetting according to claim 1, characterized in that: In step S5, a data forgetting request is generated when privacy data needs to be deleted, in order to determine the location of the sub-model and sub-dataset where the privacy data is located.

3. The student cognitive diagnosis method based on data forgetting according to claim 2, characterized in that: Step S5 specifically includes the following steps: S5-1, First, identify the user data that needs to be forgotten, including users' personal information, answer records or other sensitive data, and accurately locate the data in the model using a unique identifier; S5-2, determine the subset and sub-model where the forgotten user data is located, and generate the corresponding forgetting request.

4. The student cognitive diagnosis method based on data forgetting according to claim 2, characterized in that: The forget request includes key information such as the data identifier to be deleted, the subset containing the data, and the related sub-model.

5. A student cognitive diagnosis method based on data forgetting according to claim 2, characterized in that: The generated forget requests will be logged for future review and verification.

6. The student cognitive diagnosis method based on data forgetting according to claim 1, characterized in that: Step S10 specifically includes the following steps: S10-1 First, based on the calculated similarity, a non-linear function is used to map the similarity to weights, ensuring that sub-models with higher similarity are assigned greater weights; S10-2, normalize the initially assigned weights to ensure that the sum of all weights is 1.

7. The student cognitive diagnosis method based on data forgetting according to claim 1, characterized in that: In step S11, each sub-model calculates a weighted average of the predicted scores of the specified students based on the adjusted weights to generate the final predicted score.

Images