Prediction application based on adaptive algorithm in student portrait data

Abstract

The present invention discloses a prediction application based on an adaptive algorithm in student portrait data, which belongs to the technical field of personalized learning in education, including S1, collecting multi-dimensional data of students, and performing data cleaning and standardization preprocessing; S2, using a hierarchical clustering chameleon algorithm to perform cluster analysis on student data to form different student group portraits, and establishing a linear regression dynamic model based on these portraits to describe the student's behavior and grades, and using the recursive least squares method to identify the model parameters; S3, using the AGPC model to perform multi-step prediction, and according to the prediction results, using optimization technology to calculate the control input to achieve the expected learning effect. At the same time, through an adaptive mechanism, the model parameters and control strategies are dynamically adjusted according to actual feedback information to adapt to the changes of students; S4, applying the calculated control input to the actual teaching process to adjust the teaching content, difficulty or method, thereby optimizing teaching decisions and improving teaching quality.

Description

Technical Field

[0001] The present invention belongs to the technical field of personalized learning in education, and specifically refers to the prediction application based on adaptive algorithms in student portrait data. Background Art

[0002] In the field of education, there are currently a variety of relevant technical means for analyzing student learning situations and optimizing teaching. The most commonly used model is based on simple statistical analysis. By collecting and organizing basic data such as students' grades and attendance rates, and then calculating statistical indicators such as the mean and standard deviation of each data, we can roughly understand the overall learning status of the student group and the relative position of individuals in the group. In addition, some applications based on traditional machine learning algorithms have also been introduced into student portrait data analysis, and have achieved certain results.

[0003] However, there are still certain defects and the data is not fully utilized: relying solely on simple statistical indicators, such as mean and standard deviation, can only obtain very superficial and general information, and there is insufficient exploration of the uniqueness of individual students and the intrinsic correlation between various data. For example, although a student's grades are around the class average, his or her performance in different subjects may vary greatly. It is difficult to accurately understand this imbalance between subjects through simple statistical analysis, and it is impossible to formulate targeted strategies for improving individual subjects for him or her. Lack of predictive ability: Traditional methods mainly summarize the data that has occurred, and cannot effectively predict students' future learning outcomes, behavioral trends, etc. It is difficult for teachers to know in advance whether a student's learning status will improve or decline in the future, so they cannot take corresponding teaching intervention measures in time. Poor model adaptability: Traditional machine learning models have a relatively fixed structure. Once the data characteristics or the situation of the student group changes, the model often cannot adapt well to the new situation. For example, when a school adds a new feature course, the original decision tree model built based on the previous course data may not be able to accurately classify and analyze the students participating in the new course, because the new data characteristics and relationships are not fully considered in the original model, and the response to dynamic changes is insufficient: Traditional machine learning algorithms lack timely and effective response mechanisms when facing the dynamic changes of individual students in the learning process. Students' learning status is not static and may change significantly in a short period of time due to various factors such as family environment and their own psychological state. However, existing applications based on these algorithms find it difficult to capture these changes in real time and adjust analysis and intervention strategies accordingly, resulting in a disconnect between teaching strategies and students' actual needs. In summary, existing technologies are obviously insufficient in making full use of student portrait data to accurately predict students' learning outcomes and behavioral trends and dynamically optimize teaching accordingly, and cannot meet the urgent needs of today's education field for personalized and precise teaching. Therefore, a predictive application based on adaptive algorithms in student portrait data is proposed. Summary of the invention

[0004] The purpose of the present invention is to provide a predictive application based on an adaptive algorithm in student portrait data to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solution: based on the prediction application of the adaptive algorithm in the student portrait data, the method comprises the following steps:

[0006] S1. Data collection and preprocessing: Collect students' multi-dimensional data, including but not limited to grades, attendance, homework submission, participation in activities, etc. Clean the data, remove invalid and erroneous data records, handle missing values, and standardize the data to ensure consistency and comparability;

[0007] S2. Student portrait construction: Use the Chameleon algorithm to perform cluster analysis on student data to form different student group portraits. The algorithm calculates the similarity between data points and groups similar students to form student groups with common characteristics. When determining the number of clusters, the elbow rule or silhouette coefficient can be used to evaluate the clustering effect;

[0008] S3. Model building: Establish a dynamic model (linear regression) to describe students’ behavior and performance. The model includes input variables: attendance rate, homework submission rate, etc., and output variables: semester performance. The model predicts students’ learning performance through linear relationships, while considering the impact of time series.

[0009] S4. Parameter identification: Use recursive least squares (RLS) to identify the model parameters and estimate the model parameters, including students’ responsiveness to different teaching methods, learning rate, etc.;

[0010] S5. Prediction: Using the established AGPC model and the identified parameters, multi-step prediction is performed to determine the performance trend of students in the future. The prediction includes not only the prediction of grades, but also the prediction of activities and behaviors that students may participate in.

[0011] S6. Optimization: Based on the prediction results, use optimization techniques to calculate the control inputs to achieve the desired learning effect by minimizing the prediction error, evaluating the prediction effect, and adjusting the control inputs to optimize the learning effect;

[0012] S7, Adaptive mechanism: According to actual feedback information, dynamically adjust model parameters and control strategies to adapt to changes in students. The adaptive mechanism includes real-time monitoring of students' learning progress and behavioral changes, and timely adjustment of teaching strategies and resource allocation;

[0013] S8. Control implementation: Apply the calculated control input to the actual teaching process to adjust the teaching content, difficulty or method. Control implementation may include adjusting the course difficulty, increasing tutoring time, changing teaching methods, etc. to ensure that teaching activities match the individual needs of students.

[0014] The multi-dimensional data includes but is not limited to grades, attendance rate, homework submission, and participation in activities.

[0015] The student portrait construction adopts the Chameleon algorithm for cluster analysis, which does not assume any specific number of clusters, can adapt to the characteristics of the data set to find natural clusters, and uses a dynamic model to measure the similarity and interconnectivity between subclusters, and has the ability to discover clusters of any shape and size;

[0016] The relative interconnectivity RI and relative proximity RC between subclusters include:

[0017]

[0018] Among them, C i represents the number of data points in cluster i, EC(C i ,C j ) is the connection cluster C i and C j The sum of the weights of the edges, EC(C i ) is C i The sum of the edge weights within a cluster, is the connection cluster C i and C j The average weight of the edge, It is C i The average weight of the edges within a cluster.

[0019] Among them, the chameleon algorithm process can be divided into two stages:

[0020] In the first stage, a K-nearest neighbor graph G is constructed from the data set. k , and then use a graph partitioning algorithm to transform G k Divide into a large number of subgraphs, each subgraph represents an initial subcluster;

[0021] The second stage uses an agglomerative hierarchical clustering algorithm to repeatedly merge subclusters to find the true clustering result. The merging process is as follows:

[0022] Given a metric function: FV = RI (C i ,C j )×RC(C i ,C j );

[0023] Visit each cluster, calculate its RI and RC with each adjacent cluster, and calculate the similarity measure value (K) through the metric function formula;

[0024] Traverse each K, if K max >FV, merge the cluster with the cluster corresponding to this value;

[0025] If K max <FV, it indicates that the cluster has been merged, and the cluster list is removed and added to the result cluster;

[0026] Recursively repeat step 2 until the final size of the list of clusters to be merged is empty.

[0027] Among them, the S3 model is established as a linear regression dynamic model, and the fitting function is a cubic polynomial:

[0028] P(x)=a 0 x 3 +a 1 x 2 +a 2 x+a 3 ,

[0029] The Hurwitz matrix constructed from the polynomial coefficients:

[0030]

[0031] According to the fitted function, three sequential principal minors are calculated:

[0032]

[0033] All ordered principal minors are greater than zero, and the function is evaluated at x=1:

[0034] P(1)=a 0 +a 1 +a 2 +a 3 ≠1, so P(x) is not equal to 1 at x=1.

[0035] Among them, the Hurwitz matrix is ​​not equal to 1, and a controlled autoregressive integrated moving average model can be constructed:

[0036] y(k+j)=F j Δu f (k+j-1)+G j y f (k)+E j ε(k+j),

[0037] Among them, Δu f (k+j-1)=Δu(k+j-1) / p(x), indicating the control increment; yf (k) = y(k) / p(x), indicating the output; take j = N 1 ~N 7 ,but:

[0038] Y=F 1 ΔU f +F z ΔU f (kj)+GY f (k)+Eε,

[0039] Among them, Y is the predicted output in the future, ΔU f is the increment vector, F z is the solution of the Diophantine equation, Y f (k) is the past prediction output, and ε is the interference amount.

[0040] The adaptive mechanism dynamically adjusts the model parameters and control strategy according to the feedback information, and the model GPC control rate is:

[0041] ΔU f (k)=(F 1 T F 1 +Γ) -1 F 1 T (Y r -F 2 ΔU f (kj)-GY f (k)),

[0042] Among them, Γ is the weighting matrix, F 1 is the solution of the Diophantine equation; take the first parameter, then:

[0043] ΔU f (k)=[1,…,0](F 1 T F 1 +Γ) -1 F 1 T (Y r -F 2 ΔU f (kj)-GY f (k)),

[0044] The current control rate can be obtained as:

[0045] u(k)=u(k-1)+Δu(k)=u(k-1)+C(Z -1 )Δu f (k).

[0046] The prediction is a multi-step prediction to determine the future performance trend of students, and the prediction effect is evaluated and the control input is calculated by minimizing the prediction error;

[0047] Assume that system models A and B are unknown, and the system model is:

[0048] A(z -1 )y(k)=z -d B(z -1 )u(k)+C(z -1 )ε(k) / Δ,

[0049] but

[0050] in, where θ = [a 1 ,…,a n ] T ,because θ is unmeasurable, so parameter estimates are used instead.

[0051] The parameter estimation uses the recursive least squares method (RLS):

[0052]

[0053] Among them, the control implementation in S8 applies the control input to the actual teaching process to adjust the teaching content, difficulty or method.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] 1. The present invention can automatically adjust the model parameters through the AGPC algorithm to adapt to the dynamic changes of student behavior and learning outcomes, thereby improving the accuracy of prediction and the effectiveness of personalized learning programs;

[0056] 2. The present invention analyzes students' learning data in real time, and the AGPC algorithm can provide timely feedback to help teachers adjust teaching strategies in time to meet students' immediate learning needs;

[0057] 3. The present invention reduces the need for manual adjustment and parameter setting through the AGPC algorithm, simplifies the operation process, and enables educational institutions and teachers to implement personalized learning programs more easily;

[0058] 4. The present invention improves teaching quality and student satisfaction: Through more accurate learning outcome prediction and behavior trend analysis, the AGPC algorithm helps to improve teaching quality and student learning satisfaction, thereby improving the overall educational effect;

[0059] 5. The present invention can optimize the allocation of educational resources according to the prediction results through the AGPC algorithm, ensuring that each student can obtain appropriate support and resources, thereby improving the utilization efficiency of educational resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 It is a schematic diagram of parameter evaluation effect of the predictive application of the adaptive algorithm in student portrait data according to the present invention;

[0061] Figure 2 It is a schematic diagram of the control effect of the predictive application of the adaptive algorithm in student portrait data according to the present invention. DETAILED DESCRIPTION

[0062] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] Example

[0064] See also Figure 1-Figure 2 As shown, the present invention provides a technical solution: comprising the following steps:

[0065] S1. Data collection and preprocessing: Collect students' multi-dimensional data, including but not limited to grades, attendance, homework submission, participation in activities, etc. Clean the data, remove invalid and erroneous data records, handle missing values, and standardize the data to ensure consistency and comparability;

[0066] S2. Student portrait construction: Use the Chameleon algorithm to perform cluster analysis on student data to form different student group portraits. The algorithm calculates the similarity between data points and groups similar students to form student groups with common characteristics. When determining the number of clusters, the elbow rule or silhouette coefficient can be used to evaluate the clustering effect;

[0067] S3. Model building: Build a dynamic model to describe students’ behaviors and grades. The model includes input variables: attendance rate, homework submission rate, etc., and output variables: semester performance. The model predicts students’ learning performance through linear relationships while considering the impact of time series.

[0068] S4. Parameter identification: Use recursive least squares method to identify the model parameters and estimate the model parameters, including students’ responsiveness to different teaching methods, learning rate, etc.;

[0069] S5. Prediction: Using the established AGPC model and the identified parameters, multi-step prediction is performed to determine the performance trend of students in the future. The prediction includes not only the prediction of grades, but also the prediction of activities and behaviors that students may participate in.

[0070] S6. Optimization: Based on the prediction results, use optimization techniques to calculate the control inputs to achieve the desired learning effect by minimizing the prediction error, evaluating the prediction effect, and adjusting the control inputs to optimize the learning effect;

[0071] S7, Adaptive mechanism: According to actual feedback information, dynamically adjust model parameters and control strategies to adapt to changes in students. The adaptive mechanism includes real-time monitoring of students' learning progress and behavioral changes, and timely adjustment of teaching strategies and resource allocation;

[0072] S8. Control implementation: Apply the calculated control input to the actual teaching process to adjust the teaching content, difficulty or method. Control implementation may include adjusting the course difficulty, increasing tutoring time, changing teaching methods, etc. to ensure that teaching activities match the individual needs of students.

[0073] The multi-dimensional data includes but is not limited to grades, attendance rate, homework submission, and participation in activities.

[0074] The student portrait construction adopts the Chameleon algorithm for cluster analysis, which does not assume any specific number of clusters, can adapt to the characteristics of the data set to find natural clusters, and uses a dynamic model to measure the similarity and interconnectivity between subclusters, and has the ability to discover clusters of any shape and size;

[0075] The relative interconnectivity RI and relative proximity RC between subclusters include:

[0076]

[0077]

[0078] Among them, C i represents the number of data points in cluster i, EC(C i ,C j ) is the connection cluster C i and C j The sum of the weights of the edges, EC(C i ) is C i The sum of the edge weights within a cluster, is the connection cluster C i and C j The average weight of the edge, It is C i The average weight of the edges within a cluster.

[0079] Among them, the chameleon algorithm process can be divided into two stages:

[0080] In the first stage, a K-nearest neighbor graph G is constructed from the data set. k , and then use a graph partitioning algorithm to transform G k Divide into a large number of subgraphs, each subgraph represents an initial subcluster;

[0081] The second stage uses an agglomerative hierarchical clustering algorithm to repeatedly merge subclusters to find the true clustering result. The merging process is as follows:

[0082] Given a metric function: FV = RI (C i ,C j )×RC(C i ,C j );

[0083] Visit each cluster, calculate its RI and RC with each adjacent cluster, and calculate the similarity measure value (K) through the metric function formula;

[0084] Traverse each K, if K max >FV, merge the cluster with the cluster corresponding to this value;

[0085] If K max <FV, it indicates that the cluster has been merged, and the cluster list is removed and added to the result cluster;

[0086] Recursively repeat step 2 until the final size of the list of clusters to be merged is empty.

[0087] Among them, the S3 model is established as a linear regression dynamic model, and the fitting function is a cubic polynomial:

[0088] P(x)=a 0 x 3 +a 1 x 2 +a 2 x+a 3 ,

[0089] The Hurwitz matrix constructed from the polynomial coefficients:

[0090]

[0091] According to the fitted function, three sequential principal minors are calculated:

[0092] Δ 1 =|a 0 |>0,

[0093] All ordered principal minors are greater than zero, and the function is evaluated at x=1:

[0094] P(1)=a 0 +a 1 +a 2 +a 3 ≠1, so P(x) is not equal to 1 at x=1.

[0095] Among them, the Hurwitz matrix is ​​not equal to 1, and a controlled autoregressive integrated moving average model can be constructed:

[0096] y(k+j)=F j Δu f (k+j-1)+G j y f (k)+E j ε(k+j),

[0097] Among them, Δu f (k+j-1)=Δu(k+j-1) / p(x), indicating the control increment; y f (k) = y(k) / p(x), indicating the output; take j = N 1 ~N 7 ,but:

[0098] Y=F 1 ΔU f +F z ΔU f (kj)+GY f (k)+Eε,

[0099] Among them, Y is the predicted output in the future, ΔU f is the increment vector, F z is the solution of the Diophantine equation, Y f (k) is the past prediction output, and ε is the interference amount.

[0100] The adaptive mechanism dynamically adjusts the model parameters and control strategy according to the feedback information, and the model GPC control rate is:

[0101] ΔU f (k)=(F 1 T F 1 +Γ) -1 F 1 T (Y r -F 2 ΔU f (kj)-GY f (k)),

[0102] Among them, Γ is the weighting matrix, F 1 is the solution of the Diophantine equation; take the first parameter, then:

[0103] ΔU f (k)=[1,…,0](F 1 T F 1 +Γ) -1 F 1 T (Y r -F 2 ΔU f (kj)-GY f (k)),

[0104] The current control rate can be obtained as:

[0105] u(k)=u(k-1)+Δu(k)=u(k-1)+C(Z -1 )Δu f (k).

[0106] The prediction is a multi-step prediction to determine the future performance trend of students, and the prediction effect is evaluated and the control input is calculated by minimizing the prediction error;

[0107] Assume that system models A and B are unknown, and the system model is:

[0108] A(z -1 )y(k)=z -d B(z -1 )u(k)+C(z -1 )ε(k) / Δ,

[0109] but

[0110] in, where θ = [a 1 ,…,a n ] T ,because θ is unmeasurable, so parameter estimates are used instead.

[0111] The parameter estimation uses the recursive least squares method (RLS):

[0112]

[0113] Among them, the control implementation in S8 applies the control input to the actual teaching process to adjust the teaching content, difficulty or method.

[0114] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

[0115] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. Prediction application based on adaptive algorithm in student portrait data, characterized by: The following steps are involved: S1. Data collection and preprocessing: Collect students' multi-dimensional data, clean the data, remove invalid and erroneous data records, handle missing values, standardize the data, and ensure the consistency and comparability of the data; S2. Student portrait construction: The chameleon algorithm is used to perform cluster analysis on student data to form different student group portraits. The algorithm calculates the similarity between data points and groups similar students to form student groups with common characteristics. When determining the number of clusters, the elbow rule or silhouette coefficient method is used to evaluate the clustering effect; S3. Model building: Build a dynamic model to describe students’ behaviors and performance. The model includes input variables: attendance rate, homework submission rate, and output variables: semester performance. The model predicts students’ learning performance through linear relationships while considering the impact of time series. S4. Parameter identification: Use recursive least squares method to identify the model parameters and estimate the model parameters, including students’ responsiveness to different teaching methods and learning rate; S5. Prediction: Using the established AGPC model and the identified parameters, multi-step prediction is performed to determine the performance trend of students in the future. The prediction includes not only the prediction of grades, but also the prediction of activities and behaviors that students may participate in. S6. Optimization: Based on the prediction results, use optimization techniques to calculate the control inputs to achieve the desired learning effect by minimizing the prediction error, evaluating the prediction effect, and adjusting the control inputs to optimize the learning effect; S7, Adaptive mechanism: According to actual feedback information, dynamically adjust model parameters and control strategies to adapt to changes in students. The adaptive mechanism includes real-time monitoring of students' learning progress and behavioral changes, and timely adjustment of teaching strategies and resource allocation; S8. Control implementation: Control implementation includes at least adjusting the course difficulty, increasing tutoring time, and changing teaching methods to ensure that teaching activities match the individual needs of students.

2. The prediction application based on the adaptive algorithm in student portrait data according to claim 1, characterized in that: The multi-dimensional data includes but is not limited to grades, attendance rate, homework submission, and participation in activities.

3. The prediction application based on the adaptive algorithm in student portrait data according to claim 1 is characterized by: The student portrait construction adopts the Chameleon algorithm for cluster analysis, which does not assume any specific number of clusters, adapts to the characteristics of the data set to find natural clusters, and uses a dynamic model to measure the similarity and interconnectivity between subclusters, and has the ability to discover clusters of arbitrary shapes and sizes; The relative interconnectivity RI and relative proximity RC between subclusters include: Among them, C i represents the number of data points in cluster i, EC(C i ,C j ) is the connection cluster C i and C j The sum of the weights of the edges, EC(C i ) is C i The sum of the edge weights within a cluster, is the connection cluster C i and C j The average weight of the edge, It is C i The average weight of the edges within a cluster.

4. The prediction application based on the adaptive algorithm in student portrait data according to claim 3 is characterized by: The chameleon algorithm process is divided into two stages: In the first stage, a K-nearest neighbor graph G is constructed from the data set. k , and then use a graph partitioning algorithm to transform G k Divide into multiple subgraphs, each subgraph represents an initial subcluster; The second stage uses an agglomerative hierarchical clustering algorithm to repeatedly merge subclusters to find the true clustering result. The merging process is as follows: Given a metric function: FV = RI (C i ,C j )×RC(C i ,C j ); Visit each cluster, calculate the RI and RC of each cluster adjacent to the cluster, and calculate the similarity measure value (K) through the metric function formula; Traverse each K, if K max >FV, merge the cluster with the cluster corresponding to this value; If K max <FV, it indicates that the cluster has been merged, and the cluster list is removed and added to the result cluster; Recursively repeat step 2 until the final size of the list of clusters to be merged is empty.

5. The prediction application based on the adaptive algorithm in student portrait data according to claim 1, characterized in that: In S3, the model is established as a linear regression dynamic model, and the fitting function is a cubic polynomial: P(x)=a0x 3 +a1x 2 +a2x+a3, Hurwitz matrix constructed from polynomial coefficients: According to the fitted function, three sequential principal minors are calculated: All ordered principal minors are greater than zero, and the function is evaluated at x=1: P(1)=a0+a1+a2+a3≠1, Therefore, P(x) is not equal to 1 at x=1.

6. The prediction application based on the adaptive algorithm in student portrait data according to claim 5 is characterized by: The Hurwitz matrix is ​​not equal to 1, so a controlled autoregressive integrated moving average model can be constructed: y(k+j)=F j Δu f (k+j-1)+G j and f (k)+E j ε(k+j), Among them, Δu f (k+j-1)=Δu(k+j-1) / p(x), indicating the control increment; y f (k) = y(k) / p(x), which represents the output; if j = N1~N7, then: Y=F1ΔU f +F z ΔU f (k-j)+GY f (k)+Eε, Among them, Y is the predicted output in the future, ΔU f is the increment vector, F z is the solution of the Diophantine equation, Y f (k) is the past prediction output, and ε is the interference amount.

7. The prediction application based on the adaptive algorithm in student portrait data according to claim 6 is characterized by: The adaptive mechanism dynamically adjusts the model parameters and control strategy according to the feedback information. The model GPC control rate is: Among them, Γ is the weighting matrix, F1 is the solution of the Diophantine equation; take the first parameter, then: The current control rate can be obtained as: u(k)=u(k-1)+Δu(k)=u(k-1)+C(Z -1 )Δu f (k)。 8. The prediction application based on the adaptive algorithm in student portrait data according to claim 6 is characterized by: The multi-step prediction determines the future performance trend of the student and evaluates the prediction effect and calculates the control input by minimizing the prediction error; Assume that system models A and B are unknown, and the system model is: A(z -1 )y(k)=z -d B(z -1 )u(k)+C(z -1 )ε(k) / Δ, but in, where θ=[a1,…,a n ] T ,because θ is unmeasurable, so parameter estimates are used instead.

9. The prediction application based on the adaptive algorithm in student portrait data according to claim 8, characterized in that: The parameter estimation uses the recursive least squares method:

10. The prediction application based on the adaptive algorithm in student portrait data according to claim 1, characterized in that: The control implementation in S8 applies the control input to the actual teaching process to adjust the teaching content, difficulty or method.

Images