Teaching data management system and method based on cloud service
By preprocessing and weighted distance calculation of student learning behavior data, combined with cluster-inside-outside feature value analysis, and dynamically adjusting the K value, the problem of setting the K value in student group segmentation is solved, and more accurate student group segmentation and educational resource optimization are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 山东外事职业大学
- Filing Date
- 2026-01-14
- Publication Date
- 2026-04-17
AI Technical Summary
Existing technologies make it difficult to set a suitable K value in student group segmentation, ignoring the flexibility and variability of students' learning behaviors and personalities, and failing to reflect students' true learning characteristics.
By preprocessing student learning behavior data, using weighted distance calculation and cluster analysis, and combining cluster-inside/outside feature values, density, and separation degree, the clustering results are comprehensively evaluated, and the K value is dynamically adjusted to obtain the optimal clustering results.
Ensure that the student group segmentation conforms to the actual data structure, and that the clustering results reflect students' learning behavior patterns, so as to provide a more scientific allocation of educational resources.
Smart Images
Figure CN121880977A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of teaching data processing technology, specifically to a cloud-based teaching data management system and method. Background Technology
[0002] Teaching data refers to various data used in an educational environment to support the teaching and learning process. The main purpose of teaching data management is to track each student's learning progress, interests, and characteristics, provide students with personalized teaching plans, fully unleash students' potential, and optimize the allocation of educational resources. Therefore, the management of student learning behavior data is extremely important in teaching data. This includes student learning behavior data collected through online learning platforms, student management systems, etc., such as learning time, participation, homework scores, and exam scores. By analyzing and identifying student learning behaviors using the collected data, we can gain a comprehensive understanding of students' learning characteristics and segment students into groups based on these characteristics. Teaching administrators can then adjust teaching methods and strategies based on these characteristics, optimize teaching management and the allocation of educational resources, and provide more personalized learning support tailored to students' learning characteristics.
[0003] K-means clustering algorithm can quickly divide student behavior data into different groups, helping administrators to quickly identify students' behavioral characteristics and improve management efficiency. Administrators can develop personalized management strategies for different student groups, improving the targeting and effectiveness of management; at the same time, the algorithm is simple in principle, easy to understand and implement, and suitable for widespread application in school management systems.
[0004] However, in practical applications, it is often difficult to set a suitable K value. Usually, the number of student groups (such as excellent, good, and average) is predetermined by human experience, or determined according to the elbow rule. Such methods ignore the flexibility and variability of students' learning behavior, personality, and acceptance of different course content, and cannot guarantee that the division of student groups can reflect the students' true learning characteristics. Summary of the Invention
[0005] To address the aforementioned technical problems, the present invention aims to provide a cloud-based teaching data management system and method, the specific technical solution of which is as follows:
[0006] In a first aspect, one embodiment of the present invention provides a cloud service-based teaching data management method, the method comprising:
[0007] The learning behavior data of each student is preprocessed to obtain the data points corresponding to each student; the distance between any two data points is calculated using a weighted method and denoted as the weighted distance.
[0008] The data points are clustered once based on weighted distance to obtain different clusters; the average silhouette coefficient feature value of each cluster is obtained.
[0009] The density of data point distribution in a cluster is obtained by calculating the weighted distance from each data point within a cluster to the centroid of the cluster and the weighted distance between each data point and other data points.
[0010] The inter-cluster separation degree of this clustering is obtained by weighting the distance between the centroid of each cluster and the centroids of other clusters; the clustering result evaluation index is obtained by using the average silhouette coefficient feature value of each cluster after this clustering, the density of data point distribution, and the inter-cluster separation degree of this clustering.
[0011] The data points were clustered using different K values, and the clustering result evaluation index was obtained for each cluster. The clustering result with the largest evaluation index was taken as the basis for dividing the student groups.
[0012] Preferably, the learning behavior data of each student is preprocessed to obtain data points corresponding to each student, including:
[0013] The learning behavior data of each student is preprocessed, mainly including handling missing values, data standardization, and dimensionality reduction. Principal component analysis algorithm is used to reduce the dimensionality of the learning behavior data of each student after handling missing values and data standardization, to obtain the first principal component and the second principal component for each student. Scatter plots are drawn based on the first principal component and the second principal component, with the X-axis representing the first principal component and the Y-axis representing the second principal component. Each point in the scatter plot is a data point corresponding to each student.
[0014] Preferably, the distance between any two data points is calculated using a weighted method, denoted as the weighted distance, which includes:
[0015] Obtain the absolute value of the difference between the x-coordinates of two data values, denoted as the x-coordinate difference; obtain the absolute value of the difference between the y-coordinates of two data values, denoted as the y-coordinate difference; the x-coordinate difference divided by the sum of the x-coordinate difference and the y-coordinate difference is the x-coordinate difference weight; the y-coordinate difference divided by the sum of the x-coordinate difference and the y-coordinate difference is the y-coordinate difference weight; calculate the distance between two data points using the x-coordinate difference weight and the y-coordinate difference weight based on the Euclidean distance calculation method, denoted as the weighted distance.
[0016] Preferably, obtaining the average silhouette coefficient feature value for each cluster includes:
[0017] Add the contour coefficient of each data point in a cluster to a first preset value and multiply it by a second preset value to obtain the normalized contour coefficient of each data point; average the normalized contour coefficients of all data points to obtain the average contour coefficient feature value of the cluster.
[0018] Preferably, the density of data point distribution in a cluster is obtained based on the weighted distances from each data point within a cluster to the centroid of the cluster and the weighted distances between each data point and other data points, including:
[0019] The first mapping result is obtained by negatively correlating the mean weighted distances from all data points within a cluster to the centroid of the cluster using an exponential function with the base of the natural constant. The weighted distances between a data point within a cluster and other data points form a weighted distance sequence corresponding to that data point. The standard deviation of the weighted distances in the distance sequence formed by the weighted distance sequences corresponding to all data points is obtained, and the standard deviation is negatively correlated using an exponential function with the base of the natural constant to obtain the second mapping result. The mean of the first mapping result and the second mapping result is the density of the data point distribution in that cluster.
[0020] Preferably, the inter-cluster separation degree of this clustering is obtained based on the weighted distance between the centroid of each cluster and the centroids of other clusters, including:
[0021] Calculate the weighted distances between the centroids of each cluster and the centroids of other clusters, and average them to obtain the average value. Then, use the hyperbolic tangent function to map the average value to obtain the degree of separation between clusters.
[0022] Preferably, the evaluation metrics for the clustering results of this clustering are obtained, including:
[0023] The mean of the average silhouette coefficient feature values of all clusters obtained in this clustering is calculated and denoted as the first mean. The mean of the data point distribution density of all clusters obtained in this clustering is calculated and denoted as the second mean. The mean of the first mean and the second mean are calculated and denoted as the third mean. The clustering result evaluation index of this clustering is obtained by weighted summing of the third mean and the inter-cluster separation degree of this clustering.
[0024] Preferably, different K values are used to cluster the data points and obtain the clustering result evaluation index for each clustering, including:
[0025] The K value is incremented from the initial value to obtain different K values. Clustering is performed using different K values, and the evaluation index of the clustering results for each clustering is obtained. The increment step size is a preset step size.
[0026] Secondly, the present invention also provides a cloud-based teaching data management system, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of a cloud-based teaching data management method.
[0027] The embodiments of this invention have at least the following beneficial effects: This application obtains data points for each student by preprocessing student learning behavior data, which facilitates subsequent analysis. Simultaneously, it calculates the distance between every two data points using a weighted method to obtain a weighted distance, providing a foundation for subsequent cluster analysis and making the subsequent cluster analysis more closely reflect student learning behavior. Next, it analyzes the silhouette coefficient, the distribution of data points within clusters, and the distribution between clusters to obtain the average silhouette coefficient characteristic value, the density of data point distribution within clusters, and the degree of separation between clusters. These three factors are then combined to obtain the evaluation index of the clustering results. Different K values are then set, and multiple clustering operations are performed to obtain the evaluation index of the clustering results for different clustering operations. Finally, the clustering result with the best quality is found, and this is used as the basis for dividing the student group. This method utilizes the flexibility of student learning characteristics and behavioral patterns, ensuring that the division of student groups always conforms to the characteristics of the data, ensuring that the clustering results reflect the true data structure, and merging students with more similar learning behaviors and patterns into the same group, providing a more scientific allocation of educational resources. Attached Figure Description
[0028] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0029] Figure 1 A flowchart illustrating a cloud-based teaching data management method provided in this embodiment of the invention. Detailed Implementation
[0030] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a cloud-based teaching data management system and method proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.
[0031] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.
[0032] The following description, in conjunction with the accompanying drawings, details a specific solution for a cloud-based teaching data management system and method provided by the present invention.
[0033] Example 1:
[0034] The main application scenario of this invention is: by acquiring students' learning behavior data, classifying students' learning behaviors, and then formulating corresponding teaching strategies in a targeted manner.
[0035] Please see Figure 1 The diagram illustrates a flowchart of a cloud-based teaching data management method according to an embodiment of the present invention, which includes the following steps:
[0036] Step S1: Preprocess the learning behavior data of each student to obtain the data points corresponding to each student; calculate the distance between every two data points using a weighted method, and record it as the weighted distance.
[0037] First, it is necessary to collect students' learning behavior data. This data should include at least learning activity data, learning duration, learning progress, and academic performance, but is not limited to these data.
[0038] Learning behavior data for each student is obtained through learning management systems, online course platforms, or school-specific software or platforms that record student learning behavior. This data primarily includes learning activity data (e.g., login frequency, the frequency of accessing the learning platform or system; learning duration, the duration of each login session; learning progress, the student's progress in completing courses, etc.) and performance data (e.g., scores for each assignment submitted, and scores for each exam). This learning behavior data includes various types of data, some of which require quantification. For example, learning progress can be quantified as a percentage; this technique is existing and will not be elaborated upon here.
[0039] Furthermore, it is necessary to preprocess the learning behavior data of each student, mainly including handling missing values, data standardization, and dimensionality reduction.
[0040] In some cases, missing values might be imputed due to a student's absence from an activity or the lack of data recording. Median imputation is used to handle missing values. Furthermore, the range of values for different dimensions varies, and since K-means is based on distance calculations, data standardization is necessary first. This yields standardized learning behavior data for each student after removing missing values. Further dimensionality reduction is then required.
[0041] Student learning behavior data is diverse and highly dimensional, and direct clustering may lead to the "curse of dimensionality." Furthermore, directly clustering high-dimensional data results in difficult-to-interpret clustering results, making it hard for the system to visualize the results and for educational administrators to understand their meaning. Therefore, PCA (Programmable Analytical Processing) is used to reduce the dimensionality of multidimensional learning behavior data to address the curse of dimensionality and improve computational efficiency.
[0042] Reducing the data to two dimensions preserves the main information, avoids the curse of dimensionality, and prevents oversimplification. Two dimensions provide a balance, and they are also the easiest dimension to visualize using visualization tools, making it easier for managers to understand intuitively.
[0043] Therefore, the Principal Component Analysis (PCA) algorithm is used to reduce the dimensionality of each student's learning behavior data after processing missing values and data standardization, so as to obtain the first principal component and the second principal component corresponding to each student. A scatter plot is then drawn based on the first principal component and the second principal component, with the X-axis representing the first principal component (PC1) and the Y-axis representing the second principal component (PC2). Thus, each point in the scatter plot is the data point corresponding to each student.
[0044] In this scheme, since PCA was used for dimensionality reduction first, the contributions of different principal components in the data points (PC1, PC2) are uneven in the two-dimensional space. In order to better reflect the differences in the actual learning behavior of students, weighted Euclidean algorithm is used to calculate the distance between data points.
[0045] Specifically, the absolute value of the difference between the x-coordinates of two data values is obtained and denoted as the x-coordinate difference; the absolute value of the difference between the y-coordinates of two data values is obtained and denoted as the y-coordinate difference; the x-coordinate difference divided by the sum of the x-coordinate difference and the y-coordinate difference is the x-coordinate difference weight; the y-coordinate difference divided by the sum of the x-coordinate difference and the y-coordinate difference is the y-coordinate difference weight; the distance between two data points is calculated using the x-coordinate difference weight and the y-coordinate difference weight based on the Euclidean distance calculation method, and denoted as the weighted distance.
[0046] The specific formula for calculating the weighted distance is as follows:
[0047] ,
[0048] ,
[0049] in, Representing data points ( , ) and data points ( , The weighted Euclidean distance between them, also known as the weighted distance, is the distance between them. , The table represents the difference weights on the horizontal and vertical axes. Traditional Euclidean algorithm calculations assume that all data points have the same principal component contribution. However, in this scenario, the differences in learning behaviors among students are considered, and the difference is based on the projected difference of each principal component. (difference on the horizontal axis) The weights are set based on the difference in the ordinate. Principal components with larger projection differences usually correspond to more important directions of change in the original data. It is hoped that principal components with larger projection differences will dominate the distance calculation, increase the discrimination between data points, and better reflect the differences between data points.
[0050] Furthermore, in this scheme, since PCA was used for dimensionality reduction first, the contributions of different principal components in the data points (PC1, PC2) in the two-dimensional space are uneven. That is, the learning behavior data of different students have strong individual differences. For example, PC1 may represent the student's level of learning engagement (such as class time and homework completion), while PC2 may represent the student's learning methods (such as self-study and interaction with classmates). The proportions of PC1 and PC2 in the original data differ for different students. If the traditional Euclidean distance calculation is used, the differences in students' data in different dimensions may be ignored. In order to reflect the different characteristics of different students in each dimension, the Euclidean distance calculation is assigned weights by calculating the projection difference between two-dimensional data points, thereby obtaining a weighted distance, which makes the distance calculation between data points in this application more in line with the actual scenario.
[0051] This allows us to obtain the distance between every two data points. It should be noted that, since the data points obtained from each student's learning behavior data in this application are all obtained through dimensionality reduction using principal component analysis, the distance calculations involved in this invention are all performed using the weighted Euclidean distance calculation method, that is, the obtained distances are all weighted distances.
[0052] Step S2: Perform a clustering operation on the data points based on the weighted distance to obtain different clusters; obtain the average silhouette coefficient feature value of each cluster.
[0053] Since it is necessary to evaluate the clustering effect, an initial K value needs to be set. Preferably, in this invention, the initial K value is set to 2. A clustering operation is performed using the initial K value to obtain clustering results, which include multiple different clusters. Then, the clustering results are analyzed.
[0054] For clustering results, high-quality clustering results have relatively uniform data distribution within clusters, concentrated near the centroid, and high separation between clusters with obvious separation intervals, making them less prone to confusion. Therefore, the silhouette coefficient is used to evaluate the quality of clusters. The silhouette coefficient reflects the clustering effect by comparing the distance between data points within the same cluster with the Euclidean distance of the data point in the nearest cluster. Then, the density of data point distribution within clusters and the degree of separation between clusters are obtained to supplement the information not provided by the average silhouette coefficient feature value, thus comprehensively obtaining the evaluation index of the current clustering results.
[0055] Furthermore, the average profile coefficient feature value of the entire cluster is obtained by using the profile coefficient of each data point within a cluster. Specifically, the profile coefficient of each data point within a cluster is added to a first preset value and multiplied by a second preset value to obtain the normalized profile coefficient of each data point; the average of the normalized profile coefficients of all data points is then used to obtain the average profile coefficient feature value of the cluster.
[0056] The specific calculation formula is as follows:
[0057] ,
[0058] in, This represents the average silhouette coefficient eigenvalue of the t-th cluster after clustering with the initial K value. It reflects the consistency of data points within the cluster. A higher eigenvalue indicates better separation between data points within the cluster and other clusters, resulting in better clustering performance. . This represents the silhouette coefficient of the i-th data point within the t-th cluster. This represents the total number of data points within the t-th cluster. Due to the silhouette coefficient... The range of values is (-1, 1). To facilitate subsequent calculations and better explain the clustering effect, we will use... The normalized silhouette coefficients have their range rounded to 0 to 1, thus... The range of its value is from 0 to 1. The closer the value is to 1, the better the clustering effect. Additionally, when calculating the silhouette coefficient for each data point, it is necessary to calculate the distance between each data point; this scheme uses weighted distance as the distance between data points.
[0059] This allows us to obtain the average silhouette coefficient feature value of each cluster after this clustering, thus reflecting the clustering effect.
[0060] Step S3: Obtain the data point density of a cluster based on the weighted distance from each data point within a cluster to the centroid of the cluster and the weighted distance between each data point and other data points.
[0061] In step S2, the average silhouette coefficient feature value of each cluster is obtained. The average silhouette coefficient feature value is a relative indicator, indicating whether a data point in a cluster is more suitable in its own cluster than in the nearest cluster. However, when classifying student learning behavior data, it is desirable for the learning behavior data of students within the same cluster to have higher consistency, so as to reflect the high similarity of learning and imitation. Therefore, it is necessary to perform an absolute analysis of the data within the cluster. By calculating the average distance from the data point to the centroid and the standard deviation of the distance between each data point, the density of the data within the cluster is analyzed. The higher the density, the more compact the distribution of the data within the current cluster and the stronger the similarity of the students' learning behavior patterns.
[0062] Although the silhouette coefficient can assess the overall quality of a cluster, it does not provide a very effective analysis of the distribution of data points within the cluster. To improve the reliability of the assessment, this step will analyze the distribution of data points within the cluster.
[0063] The density of data point distribution within a cluster is obtained by calculating the weighted distances from each data point within the cluster to the centroid of the cluster and the weighted distances between each data point and other data points. Specifically, a first mapping result is obtained by negatively correlating the mean weighted distances from all data points within a cluster to the centroid of the cluster using an exponential function with the natural constant as the base. The weighted distances between a data point within a cluster and other data points form a weighted distance sequence corresponding to that data point. The standard deviation of the weighted distances in the distance sequence formed by the weighted distance sequences corresponding to all data points is obtained, and a second mapping result is obtained by negatively correlating the standard deviations using an exponential function with the natural constant as the base. The mean of the first and second mapping results is the density of data point distribution within the cluster.
[0064] The specific calculation formula is as follows:
[0065] ,
[0066] in, This indicates the density of data points in the t-th cluster after this clustering, and exp(-) represents the inverse proportional normalization function, which restricts the output to (0,1). This represents the i-th data point in the t-th cluster. With the centroid of the t-th cluster The weighted distance between them, and the centroid coordinates of the clusters are known after K-means clustering. This represents the average distance from all data points within the cluster to the centroid λ. A smaller average distance indicates a more tightly clustered data point group, meaning these data points are more similar and have higher consistency. This is the first mapping result. The larger the value, the higher the consistency.
[0067] Meanwhile, to analyze the uniformity of data point distribution within the cluster and ensure that it is not affected by extreme distributions, the weighted distance between each data point in the cluster and other data points in the cluster is calculated. Here, the weighted distance between each data point and other data points is obtained, and then a weighted distance sequence corresponding to each data point is formed. Further, the standard deviation of the weighted distance in the distance sequence composed of all the weighted distance sequences corresponding to all data points is obtained. This leads to the second mapping result. The smaller the standard deviation, the larger the second mapping result, indicating that the distance difference between data points within the cluster fluctuates less and the data points within the cluster are more evenly distributed. The closer the value is to 1, the more clustered the data points are within the centroid, indicating higher consistency, smaller fluctuations in distance between data points, more uniform distribution, and better clustering quality. This yields the density of data point distribution in each cluster after clustering using the initial K value; a higher density indicates better clustering performance.
[0068] Step S4: Obtain the inter-cluster separation degree of this clustering based on the weighted distance between the centroid of each cluster and the centroids of other clusters; obtain the clustering result evaluation index of this clustering by using the average silhouette coefficient feature value of each cluster after this clustering, the density of data point distribution, and the inter-cluster separation degree of this clustering.
[0069] Steps S2 and S3 obtained the average silhouette coefficient feature value and data point distribution density of each cluster after clustering, but did not effectively analyze the distribution between clusters. In order to improve the reliability of the evaluation, this step will analyze the distribution between clusters to supplement the information not provided by the average silhouette coefficient feature value and data point distribution density, and evaluate the clustering effect by combining these three factors.
[0070] By calculating the distance between the centroids of each cluster and taking their mean, we can analyze whether there are clear boundaries between the clusters after clustering, so that each cluster can capture an independent subset well, and truly better distinguish student learning behavior data.
[0071] The degree of inter-cluster separation in this clustering is obtained by calculating the weighted distance between the centroid of each cluster and the centroids of other clusters. Specifically, the weighted distance between the centroid of each cluster and the centroids of other clusters is calculated and averaged to obtain the average value. The degree of inter-cluster separation is obtained by mapping the average value using the hyperbolic tangent function.
[0072] The specific calculation formula is as follows:
[0073] ,
[0074] Where L represents the degree of separation between clusters in this clustering. The greater the degree of separation between clusters, the clearer the boundaries between clusters, indicating that each cluster captures an independent subset of the data well and will not be confused with other clusters, resulting in a better clustering effect; t represents the total number of clusters and also the total number of centroids; tanh() represents the hyperbolic tangent function, which restricts the output result to (0,1). L represents the weighted distance between the centroid of the k-th cluster and the centroid of the j-th cluster. t*(t-1) represents the total number of combinations between clusters, and t represents the total number of clusters. Since any cluster C has a distance correspondence with t-1 clusters, the mean distance between the centroids of all clusters is used to reflect the degree of separation between clusters. The closer L is to 1, the clearer the boundaries between clusters and the better the quality of the clustering results.
[0075] After clustering is performed using the initial K value, the quality of the current clustering is comprehensively evaluated by obtaining the average silhouette coefficient feature value of each cluster, the density of data point distribution within the cluster, and the degree of separation between clusters. Therefore, the evaluation index of the current clustering result is obtained by comprehensively obtaining the above three indicators.
[0076] The clustering results evaluation index is obtained by using the average silhouette coefficient eigenvalues of each cluster after the clustering, the density of data point distribution, and the degree of inter-cluster separation. The mean of the average silhouette coefficient eigenvalues of all clusters obtained in this clustering is calculated and denoted as the first mean; the mean of the density of data point distribution of all clusters obtained in this clustering is calculated and denoted as the second mean; the mean of the first and second means is calculated and denoted as the third mean; the clustering results evaluation index is obtained by weighted summing the third mean and the degree of inter-cluster separation.
[0077] The specific calculation formula is as follows:
[0078] ,
[0079] in, This indicates the clustering result evaluation metric for clustering performed with the initial K value. This represents the mean of the silhouette coefficient eigenvalues of all clusters obtained in this clustering, which is also the first mean. The mean value represents the density of data point distribution across all clusters obtained in this clustering, also known as the second mean value. L represents the degree of separation between clusters in this clustering. and The weights represent the weights in the weighted summation, which can be set according to different scenarios and needs. If more attention is paid to the distribution of data points within the cluster, then... If you set the settings higher and are more focused on the overall separation and clearer categorization, you can... The initial setting is too high; this solution is a compromise, with the following setting as a reference. Thus, we can obtain the clustering result evaluation index after clustering with the initial K value. The higher the value, the better the quality of the current clustering result, the better the density of data points within the cluster, and the clearer the boundaries between clusters.
[0080] Step S5: Cluster the data points using different K values and obtain the clustering result evaluation index for each clustering; take the clustering result with the largest clustering result evaluation index as the basis for dividing the student groups.
[0081] To find the optimal clustering result, different K values need to be set to cluster the data points. Specifically, the initial K value is incremented to obtain different K values, and different K values are used for clustering. The clustering result evaluation index of each clustering is obtained, and the clustering result with the largest evaluation index is taken as the basis for dividing the student groups. The increment step size is a preset step size. This invention finds the preset step size of 1.
[0082] By analyzing the optimal clustering results, student groups can be effectively segmented, grouping students with similar learning behaviors together, achieving efficient management of teaching data, adjusting different teaching methods and strategies according to different student groups, and optimizing the allocation of educational resources.
[0083] Example 2:
[0084] This embodiment provides a cloud-based teaching data management system, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When executed by the processor, the computer program implements the steps of a cloud-based teaching data management method. Since Embodiment 1 has already described a cloud-based teaching data management method in detail, it will not be elaborated upon further here.
[0085] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.
[0086] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.
[0087] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A cloud-based teaching data management method, characterized in that, The method includes: The learning behavior data of each student is preprocessed to obtain the data points corresponding to each student; the distance between any two data points is calculated using a weighted method and denoted as the weighted distance. The data points are clustered once based on weighted distance to obtain different clusters; the average silhouette coefficient feature value of each cluster is obtained. The density of data point distribution in a cluster is obtained by calculating the weighted distance from each data point within a cluster to the centroid of the cluster and the weighted distance between each data point and other data points. The inter-cluster separation degree of this clustering is obtained by weighting the distance between the centroid of each cluster and the centroids of other clusters; the clustering result evaluation index is obtained by using the average silhouette coefficient feature value of each cluster after this clustering, the density of data point distribution, and the inter-cluster separation degree of this clustering. The data points were clustered using different K values, and the clustering result evaluation index was obtained for each cluster. The clustering result with the largest evaluation index was taken as the basis for dividing the student groups.
2. The teaching data management method based on cloud services according to claim 1, characterized in that, The process of preprocessing each student's learning behavior data to obtain data points for each student includes: The learning behavior data of each student is preprocessed, mainly including handling missing values, data standardization, and dimensionality reduction. Principal component analysis algorithm is used to reduce the dimensionality of the learning behavior data of each student after handling missing values and data standardization, to obtain the first principal component and the second principal component for each student. Scatter plots are drawn based on the first principal component and the second principal component, with the X-axis representing the first principal component and the Y-axis representing the second principal component. Each point in the scatter plot is a data point corresponding to each student.
3. The teaching data management method based on cloud services according to claim 1, characterized in that, The method of calculating the distance between every two data points using a weighted approach, denoted as the weighted distance, includes: Obtain the absolute value of the difference between the x-coordinates of two data values, denoted as the x-coordinate difference; obtain the absolute value of the difference between the y-coordinates of two data values, denoted as the y-coordinate difference; the x-coordinate difference divided by the sum of the x-coordinate difference and the y-coordinate difference is the x-coordinate difference weight; the y-coordinate difference divided by the sum of the x-coordinate difference and the y-coordinate difference is the y-coordinate difference weight; calculate the distance between two data points using the x-coordinate difference weight and the y-coordinate difference weight based on the Euclidean distance calculation method, denoted as the weighted distance.
4. The teaching data management method based on cloud services according to claim 1, characterized in that, The process of obtaining the average silhouette coefficient feature value for each cluster includes: Add the contour coefficient of each data point in a cluster to a first preset value and multiply it by a second preset value to obtain the normalized contour coefficient of each data point; average the normalized contour coefficients of all data points to obtain the average contour coefficient feature value of the cluster.
5. The teaching data management method based on cloud services according to claim 1, characterized in that, The process of obtaining the data point density of a cluster based on the weighted distances from each data point within a cluster to the centroid of the cluster and the weighted distances between each data point and other data points includes: The first mapping result is obtained by negatively correlating the mean weighted distances from all data points in a cluster to the centroid of the cluster using an exponential function with the base of the natural constant. The weighted distances between a data point in a cluster and other data points form a weighted distance sequence corresponding to that data point. The standard deviation of the weighted distances in the distance sequence formed by the weighted distance sequences corresponding to all data points is obtained, and the standard deviation is negatively correlated using an exponential function with the base of the natural constant to obtain the second mapping result. The mean of the first mapping result and the second mapping result is the density of the data point distribution in the cluster.
6. The teaching data management method based on cloud services according to claim 1, characterized in that, The step of obtaining the inter-cluster separation degree of this clustering based on the weighted distance between the centroid of each cluster and the centroids of other clusters includes: Calculate the weighted distances between the centroids of each cluster and the centroids of other clusters, and average them to obtain the average value. Then, use the hyperbolic tangent function to map the average value to obtain the degree of separation between clusters.
7. The teaching data management method based on cloud services according to claim 1, characterized in that, The evaluation metrics for obtaining the clustering results of this clustering include: The mean of the average silhouette coefficient feature values of all clusters obtained in this clustering is calculated and denoted as the first mean. The mean of the data point distribution density of all clusters obtained in this clustering is calculated and denoted as the second mean. The mean of the first mean and the second mean are calculated and denoted as the third mean. The clustering result evaluation index of this clustering is obtained by weighted summing of the third mean and the inter-cluster separation degree of this clustering.
8. The teaching data management method based on cloud services according to claim 1, characterized in that, The method of clustering data points using different K values and obtaining clustering result evaluation metrics for each cluster includes: The K value is incremented from the initial value to obtain different K values. Clustering is performed using different K values, and the evaluation index of the clustering results for each clustering is obtained. The increment step size is a preset step size.
9. A cloud-based teaching data management system, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the computer program is executed by the processor, it implements the steps of a cloud service-based teaching data management method as described in any one of claims 1-8.