Learning condition analysis method and system for labor education management information system

Abstract

The invention relates to a learning condition analysis method and system oriented to a labor education management information system, and belongs to the field of education management. The method comprises the following steps: data collection and preprocessing: acquiring basic information, course completion conditions and learning behavior data of students, and preprocessing the data to obtain normalized representation of features and target variables; feature engineering: processing the normalized representation of the features to respectively obtain feature matrixes of low feature dimensions of the main features; model construction: constructing a learning condition analysis model oriented to the labor education management information system, wherein the model comprises a gradient boosting decision tree GBDT model and a long short-term memory network LSTM model; and prediction result generation: inputting the low-feature-dimension feature matrix of the main features into the model to obtain a final prediction target result. According to the method, data quality and feature richness can be ensured, static and time sequence features are effectively fused, and prediction accuracy and robustness are improved.

Description

Technical Field

[0001] The present invention belongs to the field of education management, and particularly relates to a learning situation analysis method and system for a labor education management information system. Background Art

[0002] With the rapid development of educational informatization, school management information systems are increasingly widely used in education management. With the rapid development of big data, artificial intelligence, and machine learning technologies, educational data analysis has become an important means to improve the level of education management. By systematically analyzing students' learning behavior data, course completion status, and other information, accurate prediction of students' learning progress can be achieved, thereby providing scientific decision-making support for teachers and managers. However, the existing learning situation analysis methods have the following problems when applied to labor education management information systems: First, most methods only utilize the original learning behavior data, lacking the modeling of students' knowledge states and cognitive characteristics, and it is difficult to comprehensively describe students' learning characteristics; second, the existing model fusion methods usually adopt a fixed weight strategy and cannot dynamically adjust the prediction strategy according to the characteristics of different students, resulting in limited personalized prediction ability; in addition, the learning framework with a single prediction target often ignores the correlation between multi-dimensional indicators such as learning progress and learning efficiency, restricting the generalization ability of the model. In terms of specific implementation, the existing technologies mostly adopt simple feature engineering and traditional machine learning algorithms, failing to fully utilize the advantages of deep learning in modeling time series data, especially lacking in capturing students' long-term learning behavior patterns. To address the above problems, the present invention proposes a learning situation analysis method for a labor education management information system. This method enriches the feature representation through knowledge feature enhancement and cognitive state modeling, adopts an adaptive weighted fusion mechanism to achieve personalized prediction, and introduces a multi-task learning framework to simultaneously predict learning progress and learning efficiency, effectively improving the prediction accuracy and the generalization ability of the model. Summary of the Invention

[0003] In order to solve the above problems, the present invention provides a learning situation analysis method and system for a labor education management information system.

[0004] To achieve the above object, the present invention is realized through the following technical solutions: The present invention provides a learning situation analysis method for a labor education management information system, including the following steps: S1. Data collection and preprocessing: Obtain students' basic information, course completion status, and learning behavior data, and preprocess the data to obtain the normalized representations of features in the static feature set and the time series feature set, as well as the target variable; S2. Feature Engineering: Process the normalized representations of features in the static feature set and the time series feature set to obtain a feature matrix with low feature dimensions for the main features of the static feature set and a feature matrix with low feature dimensions for the main features of the time series feature set respectively; S3. Model Construction: Construct a learning situation analysis model for the labor education management information system, and the model includes a Gradient Boosting Decision Tree (GBDT) model and a Long Short-Term Memory (LSTM) model; S4. Prediction Result Generation: Input the feature matrix with low feature dimensions for the main features of the static feature set and the feature matrix with low feature dimensions for the main features of the time series feature set into the learning situation analysis model for the labor education management information system to obtain the final prediction target result.

[0005] Further, the data collection in step S1 specifically includes: Extract the basic information of students and the course completion status from the labor education management information system, including student ID : the student ID of the student, name : the name of the student, grade : the grade where the student is located, course completion status : the number of class hours completed by the student in the course , labor education credits : the labor education credits obtained by the student, completion time : the time when the student completed the course; Collect the learning behavior data of students as auxiliary features, including login frequency : the average number of weekly logins of the student within a period of time, homework completion rate : the homework completion rate of the student, with a value range of , class participation : the participation score of the student in class, with a value range of , number of posts in the discussion area : the number of posts made by the student in the discussion area, online learning duration : the total duration of the student's online learning.

[0006] Further, the data cleaning in step S1 specifically includes: Use the mean filling method to fill the missing data in the feature set of the feature set including: course completion status feature , labor education credits feature , completion time feature , login frequency feature , assignment completion rate feature , class participation feature , number of posts in the discussion area feature , online learning duration feature , for the course completion status feature The formula for filling in the missing data in the feature is as follows: , where is the number of samples with non-missing values in the feature , represents the filled value. Similarly, the missing data of other features in the feature set is obtained; the three-sigma method is used to detect outliers in the feature set. For the course completion status feature , the outlier satisfies the following formula: , where represents the outlier in the course completion status feature , represents the mean of the course completion status of all students feature , represents the standard deviation of the course completion status of all students feature ; the truncation method is used to replace the outlier with upper and lower limits. The formula is as follows: , Similarly, the outliers of other features in the feature set are replaced.

[0007] Furthermore, the data conversion in step S1 specifically includes: Construct a static feature set , including: total login frequency feature : the total number of logins of the student during the entire learning period, assignment completion rate Feature , total class participation Feature , total number of posts in the discussion area Feature : Student Total number of posts in the discussion area, total online learning duration Feature : Student Total online learning duration during the entire learning period; construct a time series feature set , including: login frequency time series Feature : Student At time Number of logins, homework completion rate time series Feature : Student At time Homework completion rate, class participation time series Feature : Student At time Class participation score, number of posts in the discussion area time series Feature : Student At time Number of posts in the discussion area, online learning duration time series Feature : Student At time Online learning duration; Indicates the total number of time steps.

[0008] Furthermore, in step S1, feature normalization and construction of the target variable: Feature normalization: Use the standardization method to transform the features in the static feature set and the time series feature set into a distribution with a mean of 0 and a standard deviation of 1, obtaining the normalized representation of each feature. The feature normalization formula for the total login frequency feature is expressed as follows: , where, represents the normalized representation of the total login frequency feature of student , represents the mean of the total login frequency feature in the static feature set , represents the static feature set Total login frequency Standard deviation of the feature; Similarly, the static feature set is obtained And the time series feature set Normalized representation of other features in Construct the target variable: learning progress : Student At time step The learning progress of, the formula is as follows: , Among them, Is the student The labor education credits completed by the end of time step t, Is the total labor education credits required for the course, where the time step Represents the time point of the learning activity, divided in weeks.

[0009] Furthermore, step S2 specifically includes: Feature selection: Calculate the total login frequency in the normalized static feature set Among the Feature And the learning progress The Pearson correlation coefficient between : Among them, Represents the normalized total login frequency Feature The mean value of, Represents the learning progress The mean value of, similarly, the static feature set is obtained And the time series feature set The Pearson correlation coefficient between other features in and the learning progress Select the features corresponding to the correlation coefficients with absolute values greater than or equal to 0.5 as the main features. The main features include the main feature set of the static feature set And the time series feature set The main feature set in ; ; Dimensionality reduction of the features in the main feature set of the static feature set : Construct the feature matrix , where Is the number of students, Is the number of features. Perform data centering on the feature matrix to obtain the processed feature matrix , and use the covariance matrix for the processed feature matrix ​Perform eigenvalue decomposition to obtain eigenvalues and eigenvectors, and select the eigenvectors corresponding to the first largest eigenvalues to form a projection matrix , and then obtain 's feature matrix with low feature dimensions. The formula is as follows: , , , where, represents the feature mean vector, , represents 's feature matrix with low feature dimensions; similarly, obtain 's feature matrix with low feature dimensions .

[0010] Furthermore, step S4 specifically includes: Input the 's feature matrix with low feature dimensions into the GBDT model to obtain the GBDT model prediction result , input the 's feature matrix with low feature dimensions into the LSTM model to obtain the prediction result of the LSTM model , the prediction result of the GBDT model and the prediction result of the LSTM model are weighted and averaged to obtain the predicted target result , and the formula is as follows: , where, is the adaptive weight parameter of student at time step , and is obtained through the following formula: , where, represents the sigmoid activation function, represents the first learnable parameter, represents the second learnable parameter; represents the feature vector of student at time step , , where, represents the learning regularity factor, which measures the stability of the student's learning behavior; represents the progress change factor, which reflects the change trend of the recent learning progress; Denote the historical prediction error factor, which considers the historical prediction accuracy of the two models for this student, and is expressed by the formula as follows: , , , where, denotes the login frequency of student at time step , denotes the time window size considered for progress change, denotes a constant to avoid a zero denominator.

[0011] Furthermore, in step S4, a knowledge state tracking mechanism is introduced in the input layer and memory cell update process of the LSTM model, and is expressed by the formula as follows: , , where, denotes the hidden state of student at time step , denotes the input feature at the current time step, denotes the learning activity representation, denotes the learning activity encoding function of the multi-layer perceptron MLP structure.

[0012] Furthermore, the model is trained and optimized by the total loss of the model: During the model training process, the loss function optimized by the GBDT model is: , where, denotes the true value of the learning progress of student at time , denotes the mean squared error, is the regularization term, is the th tree in the model; The LSTM model adopts a multi-task learning framework and simultaneously predicts the main task "learning progress" and the auxiliary task "learning efficiency", and the loss function is expressed as follows: , , , where, is the prediction loss of learning progress, is the prediction loss of learning efficiency, is a weight coefficient used to balance the importance of the two tasks; represents a student at time The learning efficiency, defined as the amount of learning tasks completed per unit time, is calculated as follows: where is the labor education credits completed by student at time ; is the cumulative learning duration of student up to time ; The total loss of the model is: .

[0013] The present invention provides a learning situation analysis system for a labor education management information system, which executes the learning situation analysis method for the labor education management information system, including: Data collection and preprocessing module: used to obtain the basic information of students, course completion status, and learning behavior data, preprocess the data, and obtain the normalized representations of features in the static feature set and the time series feature set, as well as the target variable; Feature engineering module: used to process the normalized representations of features in the static feature set and the time series feature set, and respectively obtain the feature matrices with low feature dimensions of the main features of the static feature set and the feature matrices with low feature dimensions of the main features of the time series feature set; Model construction module: used to construct a learning situation analysis model for the labor education management information system, and the model includes a gradient boosting decision tree GBDT model and a long short-term memory network LSTM model; Prediction result generation module: used to input the feature matrices with low feature dimensions of the main features of the static feature set and the feature matrices with low feature dimensions of the main features of the time series feature set into the learning situation analysis model for the labor education management information system to obtain the final predicted target result.

[0014] The advantages of the present invention are: The present invention breaks through the limitation of the prior art that only uses the original learning behavior data through the knowledge feature enhancement and cognitive state modeling mechanisms. By constructing a knowledge association matrix and a knowledge mastery vector, and combining cognitive state features such as the overall progress, learning difficulty, and learning speed of students, it realizes a comprehensive representation of the learning characteristics of students. At the same time, the adaptive weighted fusion mechanism dynamically adjusts the model weights according to three key factors: learning regularity, progress change, and historical prediction error, providing personalized predictions for students with different learning modes and significantly improving the adaptability of the prediction. In addition, the multi-task learning framework overcomes the problem of insufficient information utilization caused by a single prediction target by simultaneously predicting the learning progress and learning efficiency, not only enhancing the generalization ability of the model but also improving the processing ability of sparse data. The synergistic effect of these designs can more accurately predict the learning progress of students, provide data-driven decision support for education managers, optimize the allocation of teaching resources, and ultimately promote the learning effect of students in the labor education curriculum. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention, and do not constitute a limitation to the present invention.

[0016] Figure 1 is a flowchart of the steps of the method of the present invention; Figure 2 is the error distribution of different models of the method of the present invention in learning progress prediction. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0018] Embodiment 1 In this embodiment, as Figure 1 shown, the present invention provides a learning situation analysis method for a labor education management information system, and the specific steps include: S1. Data collection and preprocessing: Obtain the basic information of students, course completion status, and learning behavior data, and preprocess the data to obtain the normalized representations of the features in the static feature set and the time series feature set, as well as the target variables. Specifically, the data collection includes: Extract the basic information of students and course completion status from the labor education management information system, including student numbers : students Student ID number and name : Student Name and grade : Student Grade and course completion status : Student In the course Number of class hours completed and labor education credits : Student Obtained labor education credits and completion time : Student Completion of the course Time; collect students' learning behavior data as auxiliary features, including login frequency : Student Average number of logins per week and homework completion rate within a period of time : Student Homework completion rate, with a value range of Class participation : Student Participation score in class, with a value range of Number of posts in the discussion area : Student Number of posts in the discussion area and online learning duration : Student Total online learning duration.

[0019] Specifically, data cleaning includes: Use the mean filling method to fill the missing data in the feature set The feature set Includes: Course completion status Feature Labor education credits Feature Completion time Feature Login frequency Feature Homework completion rate Feature Class participation Feature Number of posts in the discussion area Feature Online learning duration Feature , For the course completion status Feature The formula for filling the missing data in the feature is as follows: , Where as a feature the number of samples with non-missing values in denotes the imputed value. Similarly, the feature set is obtained the missing data of other features in ; The three-sigma method is used to detect outliers in the feature set, and the course completion status feature the outliers of satisfy the following formula: , where denotes the course completion status feature the outliers in denotes the course completion status of all students feature the mean of denotes the course completion status of all students feature the standard deviation of ; The truncation method is used to replace the outliers with upper and lower limits, and the formula is as follows: , Similarly, the outliers of other features in the feature set are replaced.

[0020] Specifically, the data transformation includes: Construct a static feature set , including: total login frequency feature : the student the total number of logins during the entire learning period, assignment completion rate feature total class participation feature total number of posts in the discussion area feature : the student the total number of posts in the discussion area, total online learning duration feature : the student the total online learning duration during the entire learning period; Construct a time series feature set , including: login frequency time series feature : the student at time the number of logins, assignment completion rate time series feature : the student at time the assignment completion rate, class participation time series feature : Student 's class participation score and the time series of the number of posts in the discussion area features : Student 's number of posts in the discussion area and the time series of the online learning duration at time features : Student 's online learning duration at time ; represents the total number of time steps.

[0021] Specifically, feature normalization and construction of the target variable: Feature normalization: The features in the static feature set and the time series feature set are transformed into a distribution with a mean of 0 and a standard deviation of 1 using the standardization method to obtain the normalized representation of each feature. The total login frequency The feature normalization formula for the feature is expressed as follows: , where represents the total login frequency of student 's normalized representation, represents the mean of the total login frequency feature in the static feature set , represents the standard deviation of the total login frequency feature in the static feature set ; Similarly, the normalized representations of other features in the static feature set and the time series feature set are obtained; Construction of the target variable: Learning progress : Student 's learning progress at time step , and the formula is expressed as follows: , where is the labor education credits completed by student up to time step t, is the total labor education credits required for the course, where time step represents the time point of the learning activity, divided in weeks.

[0022] ​S2. Feature Engineering: Process the normalized representations of features in the static feature set and the time series feature set to obtain a low-dimensional feature matrix of the main features of the static feature set and a low-dimensional feature matrix of the main features of the time series feature set respectively; Specifically, feature selection: Calculate the total login frequency in the normalized static feature set feature and the Pearson correlation coefficient with the learning progress : wherein, represents the normalized total login frequency feature mean, represents the mean of the learning progress Similarly, obtain the Pearson correlation coefficients between other features in the static feature set and the time series feature set and the learning progress Select the features corresponding to the correlation coefficients with absolute values greater than or equal to 0.5 as the main features. The main features include the main feature set of the static feature set and the main feature set in the time series feature set ; Reduce the dimensionality of the features in the main feature set of the static feature set : Construct a feature matrix where is the number of students, is the number of features. Perform data centering on the feature matrix to obtain the processed feature matrix . Use the covariance matrix to perform eigenvalue decomposition on the processed feature matrix to obtain eigenvalues and eigenvectors. Select the eigenvectors corresponding to the first largest eigenvalues to form a projection matrix . Furthermore, obtain low-dimensional feature matrix, and the formula is as follows: , , , wherein, represents the feature mean vector, , represents low-dimensional feature matrix; Similarly, obtain Feature matrix with low feature dimension .

[0023] S3. Model construction: Construct a learning situation analysis model for the labor education management information system. The model includes a Gradient Boosting Decision Tree (GBDT) model and a Long Short-Term Memory (LSTM) model; S4. Prediction result generation: The feature matrix with low feature dimension of the main features of the static feature set and the feature matrix with low feature dimension of the main features of the time series feature set are input into the learning situation analysis model for the labor education management information system to obtain the final predicted target result.

[0024] Specifically, the feature matrix with low feature dimension is input into the GBDT model to obtain the GBDT model prediction result , and the feature matrix with low feature dimension is input into the LSTM model to obtain the prediction result of the LSTM model . The prediction result of the GBDT model and the prediction result of the LSTM model are weighted and averaged to obtain the predicted target result , where is the adaptive weight parameter of the student at time step and is obtained through the following formula: , where represents the sigmoid activation function, represents the first learnable parameter, represents the second learnable parameter; represents the feature vector of the student at time step , , where represents the learning regularity factor, which measures the stability of the student's learning behavior; represents the progress change factor, which reflects the change trend of the recent learning progress; represents the historical prediction error factor, which considers the historical prediction accuracy of the two models for this student. The formula is expressed as follows: , , , wherein, represents the login frequency of the student at time step , represents the time window size considered for progress change, represents a constant to avoid a zero denominator.

[0025] Specifically, a knowledge state tracking mechanism is introduced in the input layer and the memory cell update process of the LSTM model, and the formula is as follows: , , wherein, represents the hidden state of the student at time step , represents the input feature at the current time step, represents the learning activity representation, represents the learning activity encoding function of the multi-layer perceptron (MLP) structure, which is used to map the hidden state of the LSTM to the vector space representing the current learning activity and knowledge state of the student.

[0026] Specifically, the model is trained and optimized through the total loss of the model: During the model training process, the loss function optimized by the GBDT model is: , wherein, represents the true value of the learning progress of the student at time , represents the mean squared error, is the regularization term, is the th tree in the model; The LSTM model adopts a multi-task learning framework and simultaneously predicts the main task "learning progress" and the auxiliary task "learning efficiency", and the loss function is as follows: , , , wherein, is the learning progress prediction loss, is the learning efficiency prediction loss, is the weight coefficient, which is used to balance the importance of the two tasks; represents the learning efficiency of the student at time , which is defined as the amount of learning tasks completed per unit time, and the calculation formula is as follows: Among them, is the labor education credits completed by the student at time the completed labor education credits; is the student as of time the cumulative learning duration; The total loss of the model is: .

[0027] Embodiment 2 This embodiment provides a learning situation analysis system for a labor education management information system, which executes the learning situation analysis method for a labor education management information system described in Embodiment 1, including: Data collection and preprocessing module: used to obtain the basic information of students, course completion status, and learning behavior data, preprocess the data, and obtain the normalized representations of features and target variables in the static feature set and the time series feature set; Feature engineering module: used to process the normalized representations of features in the static feature set and the time series feature set, and respectively obtain the low-dimensional feature matrices of the main features of the static feature set and the low-dimensional feature matrices of the main features of the time series feature set; Model construction module: used to construct a learning situation analysis model for a labor education management information system, and the model includes a gradient boosting decision tree GBDT model and a long short-term memory network LSTM model; Prediction result generation module: used to input the low-dimensional feature matrices of the main features of the static feature set and the low-dimensional feature matrices of the main features of the time series feature set into the learning situation analysis model for a labor education management information system to obtain the final predicted target result.

[0028] Embodiment 3 In this embodiment, in order to verify the effectiveness of the learning situation analysis method for a labor education management information system proposed by the present invention, we designed a simulation experiment for the system. The experiment uses a real student learning dataset collected from a labor education management information system of a certain university, and the dataset includes the basic information of students, course completion status, and multi-dimensional learning behavior data. The experiment first preprocesses and performs feature engineering on the data, and then randomly divides the dataset into a training set and a test set. In terms of model construction, traditional linear regression, random forest, a single gradient boosting decision tree (GBDT), a single long short-term memory network (LSTM), etc. are selected as baseline methods for comparison with the method of the present invention.

[0029] Table 1 Experimental comparison between the method of the present invention and the prior art In this embodiment, as shown in Table 1, we used the root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R²) as evaluation indicators. The experimental results show that the method of the present invention is superior to traditional methods in terms of prediction accuracy and robustness, fully verifying its application value in the labor education management information system.

[0030] Figure 2 The error distribution of different models in learning progress prediction was compared in the form of a box plot. The horizontal axis represents different models, namely the linear regression model, GBDT model, LSTM model, and the method of the present invention; the vertical axis represents the error range of learning progress prediction. The experimental results show that the method of the present invention is superior to other models in learning progress prediction, showing advantages such as smaller errors, more concentrated distribution, and fewer outliers, indicating its significant advantages in terms of accuracy and robustness.

[0031] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, for those skilled in the art, they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A learning situation analysis method for a labor education management information system, characterized in that: The following steps are involved: S1. Data collection and preprocessing: Obtain students’ basic information, course completion status, and learning behavior data, preprocess the data, and obtain the normalized representation of features in the static feature set and time series feature set as well as the target variable; S2. Feature Engineering: Process the normalized representations of the features in the static feature set and the time series feature set to obtain the feature matrices of low feature dimensions of the main features of the static feature set and the feature matrices of low feature dimensions of the main features of the time series feature set, respectively; S3. Model construction: construct a learning situation analysis model for the labor education management information system, the model includes a gradient boosting decision tree GBDT model and a long short-term memory network LSTM model; S4. Generation of prediction results: The feature matrix of low feature dimensions of the main features of the static feature set and the feature matrix of low feature dimensions of the main features of the time series feature set are input into the learning situation analysis model for the labor education management information system to obtain the final prediction target results.

2. The learning situation analysis method for the labor education management information system according to claim 1 is characterized in that: Step S1 data collection specifically includes: Extract students’ basic information and course completion status, including student ID, from the labor education management information system :student Student ID and name :student Name, grade :student Grade level and course completion status :student In Course Number of completed class hours and labor education credits :student Labor education credits earned and completion time :student Complete the course time; collect students’ learning behavior data as auxiliary features, including login frequency :student Average weekly logins and job completion rates over a period of time :student The job completion rate ranges from , Classroom Participation :student The participation score in the class ranges from , Number of posts in discussion forum :student Number of posts in the discussion forum and online learning time :student The total duration of online learning.

3. The learning situation analysis method for the labor education management information system according to claim 2 is characterized in that: Step S1 data cleaning specifically includes: Use mean imputation to fill in the feature set The missing data of the feature set Including: Course completion feature , Labor Education Credits feature , Completion time feature , Login frequency feature , homework completion rate feature , Classroom Participation feature , Number of posts in discussion forum feature , online learning duration feature , on course completion feature The formula for filling missing data in features is as follows: ， in, Features The number of samples with non-missing values ​​in , Represents the filling value. Similarly, we get the feature set Missing data for other features in the feature set; triple standard deviation method is used to detect outliers in the feature set, course completion feature The outlier value satisfies the following formula: ， in, Indicates course completion status feature The outliers in Indicates the course completion status of all students feature The mean of Indicates the course completion status of all students feature The standard deviation of the value is 0.0000. The outliers are replaced with upper and lower limits by truncation method. The formula is as follows: ， Similarly, the feature set The outliers of other features in are replaced.

4. The learning situation analysis method for the labor education management information system according to claim 3 is characterized in that: Step S1 data conversion specifically includes: Constructing static feature sets , including: Total login frequency feature :student Total login times and homework completion rate during the entire learning period feature , total class participation feature , Total number of discussion forum posts feature :student Total number of posts in the discussion area and total online learning time feature :student The total online learning time during the entire learning period; construct a time series feature set , including: login frequency time series feature :student In time Login times and job completion rate time series feature :student In time Time series of homework completion rate and class participation feature :student In time Time series of class participation scores and number of discussion forum posts feature :student In time The number of posts in the discussion forum and the time series of online learning feature :student In time Length of online learning; Represents the total number of time steps.

5. The learning situation analysis method for the labor education management information system according to claim 4 is characterized in that: In step S1, feature normalization and target variable construction are performed: Feature normalization: Use standardization methods to normalize static feature sets and time series feature set The features in are converted to a distribution with a mean of 0 and a standard deviation of 1, and the normalized representation of each feature is obtained. The total login frequency The feature normalization formula of the feature is expressed as follows: ， in, Indicates students Total login frequency The normalized representation of the features, Represents a static feature set Login frequency The mean of the feature, Represents a static feature set Login frequency The standard deviation of the feature; similarly, the static feature set is obtained and time series feature set Normalized representation of other features in; Constructing the target variable: learning progress :student At time step The learning progress is expressed as follows: ， in, For Students The labor education credits completed up to time step t, Total labor education credits required for the course, including the time step Indicates the time point of learning activities, divided into weeks.

6. The learning situation analysis method for the labor education management information system according to claim 5 is characterized in that: Step S2 specifically includes: Feature selection: Calculate the normalized static feature set Total login frequency in feature and learning progress Pearson correlation coefficient : in, Represents the normalized total login frequency feature The mean of Indicates learning progress The mean of and time series feature set Other features and learning progress The Pearson correlation coefficient between them is selected, and the feature corresponding to the correlation coefficient with an absolute value greater than or equal to 0.5 is selected as the main feature. The main feature includes a static feature set The main feature set and time series feature set Main feature set ; For static feature sets The main feature set Reduce the dimension of the features in: Construct the feature matrix ,in is the number of students, is the feature number, the feature matrix is ​​centrally processed to obtain the processed feature matrix , the processed feature matrix is ​​converted into the covariance matrix Perform eigenvalue decomposition to obtain eigenvalues ​​and eigenvectors, and select The eigenvectors corresponding to the largest eigenvalues ​​form the projection matrix , and then get The low-dimensional feature matrix of , the formula is as follows: ， ， ， in, represents the feature mean vector, ,express The low-dimensional feature matrix of The low-dimensional feature matrix .

7. The learning situation analysis method for the labor education management information system according to claim 6 is characterized in that: Step S4 specifically includes: Will The low-dimensional feature matrix Input into the GBDT model to get the GBDT model prediction results ,Will The low-dimensional feature matrix Input into the LSTM model to get the prediction result of the LSTM model , the prediction results of the GBDT model And the prediction results of the LSTM model Perform weighted averaging to obtain the predicted target result , the formula is as follows: ， in, For Students At time step The adaptive weight parameter is obtained by the following formula: ， in, represents the sigmoid activation function, represents the first learnable parameter, represents the second learnable parameter; Indicates students At time step The characteristic vector of ,in, It represents the learning regularity factor, which measures the stability of students’ learning behavior; Indicates the progress change factor, reflecting the recent trend of learning progress; Represents the historical prediction error factor. Considering the historical prediction accuracy of the two models on this student, the formula is as follows: ， ， ， in, Indicates students At time step Login frequency, Indicates the time window size considered for progress changes, Represents a constant that prevents the denominator from being zero.

8. The learning situation analysis method for the labor education management information system according to claim 7 is characterized in that: Step S4, introduce the knowledge state tracking mechanism in the LSTM model input layer and memory unit update process, the formula is as follows: ， ， in, Indicates students At time step The hidden state of represents the input features of the current time step, Represents learning activities, Represents the learning activity encoding function of the multi-layer perceptron MLP structure.

9. The learning situation analysis method for the labor education management information system according to claim 8 is characterized in that: The model is trained and optimized through the total loss of the model: During the model training process, the loss function optimized by the GBDT model is: ， in, Indicates students In time The true value of the learning progress, represents the mean square error, is the regularization term, For the model The LSTM model adopts a multi-task learning framework to simultaneously predict the main task "learning progress" and the auxiliary task "learning efficiency". The loss function is expressed as follows: ， ， ， in, Predict the loss for learning progress, Prediction loss for learning efficiency, is the weight coefficient, used to balance the importance of the two tasks; Indicates students In time The learning efficiency is defined as the amount of learning tasks completed per unit time, and the calculation formula is as follows: in, For Students In time Completed labor education credits, For Students Deadline Cumulative study time; The total loss of the model is: .

10. A learning situation analysis system for a labor education management information system, executing the learning situation analysis method for a labor education management information system as claimed in claim 1, characterized in that: include: Data collection and preprocessing module: used to obtain students' basic information, course completion status, and learning behavior data, preprocess the data, and obtain the normalized representation of the features in the static feature set and time series feature set as well as the target variable; Feature engineering module: used to process the normalized representation of features in the static feature set and the time series feature set, and obtain the feature matrix of low feature dimensions of the main features of the static feature set and the feature matrix of low feature dimensions of the main features of the time series feature set respectively; Model building module: used to build a learning situation analysis model for the labor education management information system, the model includes a gradient boosting decision tree GBDT model and a long short-term memory network LSTM model; Prediction result generation module: used to input the feature matrix of low feature dimension of the main features of the static feature set and the feature matrix of low feature dimension of the main features of the time series feature set into the learning situation analysis model for the labor education management information system to obtain the final prediction target result.

Images