Methods, systems, and media for generating educational reports based on learning behavior data

Abstract

This application discloses a method, system, and medium for generating educational reports based on learning behavior data, relating to the field of educational technology. The method includes the following steps: acquiring learning behavior data of each user in an online education platform; calculating the learning enthusiasm of each user based on their interest in the learning content; determining the degree of learning effectiveness for each user based on their homework completion indicators and their learning enthusiasm; and clustering the users and the learning behavior data based on the degree of learning effectiveness to obtain multiple final clusters, and generating educational reports corresponding to the final clusters. This application improves the accuracy of the generated educational reports.

Description

Technical Field

[0001] This application relates to the field of educational technology, and in particular to a method, system, and medium for generating educational reports based on learning behavior data. Background Technology

[0002] Currently, in the development of modern educational technology, by analyzing students' behavioral data during the learning process, necessary educational reports can be generated, and teaching methods can be improved based on these reports.

[0003] In existing educational report generation processes, analysis is often based solely on basic learning behavior data (such as learning time and homework completion indicators), and the resulting learning behavior data is then clustered. This approach considers only a single data dimension and fails to effectively extract students' true learning behavior characteristics, leading to poor accuracy in cluster analysis results and consequently, low accuracy in the generated educational reports. Summary of the Invention

[0004] The main purpose of this application is to provide a method, system, and medium for generating educational reports based on learning behavior data, aiming to solve the technical problem in related technologies that simply cluster the obtained learning behavior data, which leads to poor accuracy of cluster analysis results and thus low accuracy of educational report generation results.

[0005] To achieve the above objectives, embodiments of this application provide a method for generating educational reports based on learning behavior data, including:

[0006] Obtain learning behavior data from each user on the online education platform;

[0007] Based on the level of interest each user has in the learning content from the learning behavior data, the learning enthusiasm of each user is calculated.

[0008] The degree to which each user's learning effectiveness is demonstrated is determined based on each user's homework completion metrics and learning enthusiasm.

[0009] Based on the degree of learning effectiveness, the data of each user and their learning behavior are clustered to obtain multiple final clusters, and an educational report corresponding to each final cluster is generated.

[0010] In one possible implementation of this application, before calculating each user's learning motivation based on their level of interest in the learning content from the learning behavior data, the method further includes:

[0011] Extract the set of courses watched by each user from the learning behavior data;

[0012] Determine the intersection and union of course sets between any two users;

[0013] Based on the course intersection and course union, the similarity of learning content between any two users is calculated.

[0014] Based on the similarity of the learning content, the degree of interest of each user in the learning content is calculated.

[0015] In one possible implementation of this application, the degree of interest of each user in the learned content is calculated based on the similarity of the learned content, including:

[0016] Based on the similarity of the learning content, the clustering distance is determined, and each user is clustered according to the clustering distance to obtain multiple first clusters;

[0017] For any first cluster, extract the first duration of the courses watched by each user, and the second duration required to watch the courses without repeating them;

[0018] Based on the first and second durations, the degree of interest of each user in the learning content is determined.

[0019] In one possible implementation of this application, the learning motivation of each user is calculated based on the degree of interest each user has in the learning content from the learning behavior data, including:

[0020] Statistics were compiled on each user's assignment completion rate across all viewed courses, as well as the number of comments posted across all viewed courses.

[0021] The after-class knowledge consolidation factor for each user was calculated based on the homework completion rate and the number of comments sent.

[0022] For any given user, determine the number of interactions between the current user's comments and other users, and calculate the interaction effect value between the current user and other users based on the number of interactions;

[0023] Based on the interaction effect value, the after-class knowledge consolidation factor was adjusted to obtain the after-class learning performance factor for each user.

[0024] The learning motivation of each user is calculated based on their level of interest and post-class learning performance.

[0025] In one possible implementation of this application, after calculating the learning motivation of each user based on the level of interest and after-class learning performance factors, the method further includes:

[0026] Determine the maximum and minimum values ​​of learning motivation among all users in the current user's cluster;

[0027] For any given user, the relative learning motivation of each user is calculated based on the maximum value, minimum value, and learning motivation level.

[0028] Based on each user's assignment completion metrics and learning engagement, the degree to which each user demonstrates their learning effectiveness is determined, including:

[0029] Based on each user's assignment completion metrics and relative learning enthusiasm, the degree of learning effectiveness demonstrated by each user is determined.

[0030] In one possible implementation of this application, the degree of learning effectiveness of each user is determined based on their task completion metrics and relative learning enthusiasm, including:

[0031] Based on each user's homework completion metrics, the learning effectiveness value of each user for each chapter of the course is calculated. Homework completion metrics include homework completion rate, homework completion time, and homework grade.

[0032] Based on the relative level of learning enthusiasm and the learning effect value, the comprehensive learning efficiency of each user is calculated.

[0033] The level of understanding of knowledge points is calculated based on the degree of correlation between the homework questions in each chapter.

[0034] The learning outcomes of each user are determined based on their overall learning efficiency and understanding of the knowledge points.

[0035] In one possible implementation of this application, the degree of relevance includes potential relevance. Based on the degree of relevance between the homework assignments for each chapter, a knowledge point comprehension value is calculated, including:

[0036] The homework questions for each chapter are processed using a pre-defined association rule mining algorithm to determine the potential associations between the homework questions.

[0037] For any user, based on potential correlations, the completed questions of the current user are clustered to obtain multiple question clusters and the membership degree of each completed question to each question cluster;

[0038] Construct a sequence of completed questions based on question clusters;

[0039] Based on the score and set score of each completed question in the sequence of completed questions, the relative score of each completed question is calculated.

[0040] For any question cluster, based on the membership degree and the completion time of each completed question for the current user, a linear fit is performed on the relative score to obtain the knowledge point comprehension value of the current question cluster.

[0041] In one possible implementation of this application, the learning performance level of each user is obtained based on the comprehensive learning efficiency and the degree of understanding of knowledge points, including:

[0042] For any question cluster, determine the first mean among the membership degrees of each completed question in the current question cluster;

[0043] Based on the first mean, the weight values ​​of the knowledge point comprehension levels are calculated.

[0044] Based on the weight values ​​and the degree of understanding of knowledge points, the overall learning efficiency is adjusted to obtain the degree of learning effect of each user under active learning.

[0045] This application also provides an educational report generation system based on learning behavior data, which includes:

[0046] The acquisition module is used to acquire learning behavior data of each user in the online education platform.

[0047] The calculation module calculates each user's learning motivation based on their interest in the learning content from the learning behavior data.

[0048] The module determines the extent to which each user's learning outcomes are demonstrated based on their homework completion metrics and learning enthusiasm.

[0049] The processing module clusters the data of each user and their learning behavior based on the degree of learning effectiveness, resulting in multiple final clusters and generating educational reports corresponding to the final clusters.

[0050] To achieve the above objectives, a storage medium is also provided, on which an educational report generation program based on learning behavior data is stored. When the educational report generation program based on learning behavior data is executed by a processor, it implements the steps of any of the above-described educational report generation methods based on learning behavior data.

[0051] This application provides a method, system, and medium for generating educational reports based on learning behavior data. Compared to related technologies that merely cluster the obtained learning behavior data, which leads to poor accuracy in cluster analysis results and consequently low accuracy in generated educational reports, this application acquires learning behavior data from various users on an online education platform. Based on each user's interest in the learning content, the application calculates their learning motivation. Then, based on the user's homework completion indicators and learning motivation, it determines the degree of learning effectiveness demonstrated by the user's active learning. Furthermore, based on the degree of learning effectiveness, the application clusters different users and their corresponding learning behavior data, generating educational reports from the resulting multiple final clusters. This approach combines the motivation and learning effectiveness of different users to perform precise clustering, thereby improving the accuracy of the generated educational reports. Attached Figure Description

[0052] Figure 1 This is a flowchart illustrating the first embodiment of the educational report generation method based on learning behavior data according to this application.

[0053] Figure 2 This is a schematic diagram of the online education platform involved in the educational report generation method based on learning behavior data in this application;

[0054] Figure 3 This is a schematic diagram of the growth curve involved in the educational report generation method based on learning behavior data in this application;

[0055] Figure 4 This is a schematic diagram of the educational report involved in the educational report generation method based on learning behavior data in this application;

[0056] Figure 5 This is a flowchart illustrating the second embodiment of the educational report generation method based on learning behavior data in this application;

[0057] Figure 6 This is a schematic diagram of the overall execution process involved in the educational report generation method based on learning behavior data in this application;

[0058] Figure 7 This is a schematic diagram of the device structure of the hardware operating environment involved in the embodiments of this application. Detailed Implementation

[0059] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0060] This application provides a method for generating educational reports based on learning behavior data. In the first embodiment of this method, refer to... Figure 1 The methods include:

[0061] Step S10: Obtain learning behavior data of each user in the online education platform;

[0062] Step S20: Calculate the learning motivation of each user based on their level of interest in the learning content from the learning behavior data;

[0063] Step S30: Based on each user's homework completion metrics and learning enthusiasm, determine the degree of learning effectiveness demonstration for each user;

[0064] Step S40: Based on the degree of learning effect, cluster the data of each user and learning behavior to obtain multiple final clusters, and generate the corresponding education report for each final cluster.

[0065] This embodiment aims to improve the accuracy of the generated educational reports.

[0066] The specific steps are as follows:

[0067] Step S10: Obtain learning behavior data of each user in the online education platform.

[0068] As an example, the method for generating educational reports based on learning behavior data can be applied to an educational report generation device based on learning behavior data. The educational report generation device based on learning behavior data belongs to an educational report generation system based on learning behavior data, which in turn belongs to an educational report generation equipment based on learning behavior data.

[0069] As an example, users can be students or other people learning courses on the platform. Learning behavior data includes courses watched by users, viewing time, completion status of homework and completion time, etc., without specific limitations.

[0070] As an example, online education platforms such as Figure 2 As shown, on this platform, teachers are responsible for classroom management, course materials, and post-class assessments, while students use the platform for pre-class learning, in-class discussions, and post-class evaluations.

[0071] As an example, when collecting learning behavior data from online education platforms, one can collect the total viewing time and unique viewing time of each course for each user within the past three months. Each course includes several chapter teaching videos. Then, one can collect the completion status of homework and interaction status in the discussion forum after each chapter teaching video.

[0072] For example, if a user watches a course for 5 minutes and then watches the same content again for 3 minutes, the total time is 8 minutes, with the non-repeated time being 5 minutes. Homework completion includes completion rate, grade, and completion time. Each question in the homework is assigned a completion rate percentage and a score, with the sum of all completion rates set at 100%. The final completion rate is the sum of the completion rates of all completed questions. The final grade is the ratio of the sum of the scores for completed questions to the total homework score (i.e., completion rate and grade are normalized and have a unified dimension). The time each user takes to complete each question is recorded (based on the last submission of that question). Each homework assignment has a set completion time and each user's actual completion time. Discussion forum interaction includes: the number of comments posted by users, the number of replies, likes, and reports under each comment.

[0073] Step S20: Calculate the learning motivation of each user based on their level of interest in the learning content from the learning behavior data.

[0074] As an example, during the learning process on an online education platform, due to differences in users' needs, learning stages, interests, and age groups, the learning content will vary from user to user, and users will also have different levels of interest in the learning content. When the level of interest varies, the users' learning enthusiasm will also vary.

[0075] As an example, different learning content has different requirements for the learning process. For instance, learning mathematics may focus on cultivating logical thinking and problem-solving skills, while learning Chinese emphasizes the comprehensive application of skills such as listening, speaking, reading, and writing. Therefore, the focus and methods in the learning process will also be different. Thus, using data of the same dimension to compare and analyze users of different learning content will lead to a large error in the analysis results. Therefore, in this embodiment, we first perform preliminary classification based on the learning content of different users, and then conduct comparative analysis among users with similar learning content to determine the degree of interest in the learning content and the user's enthusiasm for learning.

[0076] Before step S20, which generates an education report based on learning behavior data, steps A1 to A4 are also included:

[0077] Step A1: Extract the set of courses watched by each user from the learning behavior data.

[0078] As an example, a course collection can be a collection of courses that a user has watched, such as language arts courses, math courses, etc.

[0079] Step A2: Determine the course intersection and course union between any two users' course sets.

[0080] As an example, two users may watch different courses because they have different learning focuses. The course intersection is the set of the same course types watched by the two users, while the course union is the sum of the types of courses watched by the two users.

[0081] Step A3: Calculate the similarity of learning content between any two users based on the course intersection and course union.

[0082] As an example, the closer the intersection of courses is to the union of courses, the higher the similarity of the learning content between the two users. Therefore, the similarity P of the learning content can be calculated as follows:

[0083]

[0084] Where S1 is the number of course types in the course intersection, and S2 is the number of course types in the course union. The closer the result is to 1, the more consistent the types of courses the two users are studying; L1i is the total time one user spends watching the i-th course in the intersection, L2i is the total time another user spends watching the i-th course in the intersection, L1 is the sum of the total time one user spends watching all courses, and L2 is the sum of the total time another user spends watching all courses. Therefore... The closer it is to 1, and The closer a value is to 1, the more time each user spends studying courses in the overlap area within their respective study timeframes. Therefore, using... As The adjustment value is used to obtain the similarity of the learning content between any two users, where norm is a linear normalization function.

[0085] Step A4: Based on the similarity of the learning content, calculate the degree of interest of each user in the learning content.

[0086] As an example, the higher the similarity of the learning content, the more interested the two users are. First, classify the users, and then calculate the degree of interest of users with similar learning content in the content they are learning.

[0087] Step A4, which calculates each user's level of interest in the learned content based on the similarity of the learned content, includes:

[0088] Based on the similarity of the learned content, the clustering distance is determined, and each user is clustered using the clustering distance to obtain multiple first clusters.

[0089] As an example, the clustering distance is obtained by subtracting the difference in the similarity of the learning content between any two users from 1. The K-means clustering algorithm (a well-known technique) is used to cluster all users, resulting in multiple first clusters, in which the learning content of users in each first cluster is similar.

[0090] As an example, when users have high learning motivation and a great interest in the learning content, they often repeat the learning content multiple times. For example, they preview the material before class, relearn the weak points or the content they did not understand after class, and systematically review and consolidate their knowledge. In addition, their homework completion rate is often high and they interact frequently in the discussion forum.

[0091] For any first cluster, extract the first duration of the courses watched by each user, and the second duration required to watch the courses without repeating them.

[0092] As an example, the first duration is the total time a user spends watching all the courses, including the time a user repeatedly watches some videos. The second duration is the total time a user needs to watch all the courses, excluding the repeated video viewing periods.

[0093] Based on the first and second durations, the degree of interest of each user in the learning content is determined.

[0094] As an example, the ratio between the first duration and the second duration can be denoted as the current user's level of interest in the learning content. Similarly, the level of interest of other users in the learning content can be calculated.

[0095] Step S30: Based on each user's homework completion metrics and learning enthusiasm, determine the degree of learning effectiveness of each user.

[0096] As an example, homework completion metrics could include homework rate, homework grade, and the time spent completing the homework.

[0097] As an example, the degree of learning effectiveness is used to represent the user's learning effect. There will be a degree value. The higher the value of the degree of learning effectiveness, the better the user's learning effect.

[0098] As an example, the learning effect of a user is calculated based on the user's learning enthusiasm and homework completion indicators. By taking into account the user's learning attitude during the learning process and the completion status of the homework, a more accurate representation of the learning effect can be obtained.

[0099] Step S40: Based on the degree of learning effect, cluster the data of each user and learning behavior to obtain multiple final clusters, and generate the corresponding education report for each final cluster.

[0100] As an example, users are clustered based on the degree of learning effectiveness, resulting in multiple updated clusters. The learning behavior data of all users in the updated clusters are then used to construct the final cluster.

[0101] As an example, the way to update the construction of clusters could be:

[0102] Will The updated clustering distance is defined as the distance between any two users, where P is the similarity of the learning content between the two users, Q1 and Q2 are the degree of learning effect of the two users under active learning, and || is the absolute value function, that is, the more similar the learning content of the users (the larger P is) and the more similar the degree of learning effect under active learning (the smaller |Q1-Q2| is), the shorter the updated clustering distance. The K-means clustering algorithm (a well-known technique) is used to cluster all users to obtain multiple updated clusters.

[0103] As an example, the method for generating the educational report corresponding to the final cluster could be:

[0104] 1. Assign a cluster label to each user in the final cluster, and save the clustering results (cluster labels) as a data file, such as CSV format.

[0105] 2. Write an R script to read the integrated data file. Further process the data in the R script, such as calculating the average learning time and job completion rate for each cluster.

[0106] 3. Text Description: Using Natural Language Generation (NLG) technology, the analysis results are converted into easy-to-understand text descriptions and then into chart files. Chart Drawing: Using R's plotting libraries (such as ggplot2), visualization charts such as bar charts and line charts are generated to intuitively display the clustering results and related indicators.

[0107] 4. Integrate the generated text and charts into a single report document, such as PDF or HTML format, and output the final educational report. These are all common operations. The user's growth curve is as follows: Figure 3 As shown in the diagram, the educational report is illustrated below. Figure 4 As shown.

[0108] This application provides a method for generating educational reports based on learning behavior data. Compared with related technologies that merely cluster the obtained learning behavior data, which leads to poor accuracy in cluster analysis results and thus low accuracy in generating educational reports, this application obtains learning behavior data of each user in an online education platform, calculates each user's learning enthusiasm based on their interest in the learning content, and then determines the degree of learning effectiveness under active learning based on the user's homework completion indicators and learning enthusiasm. Furthermore, based on the degree of learning effectiveness, different users and their corresponding learning behavior data are clustered, and multiple final clusters are generated to generate educational reports. This method combines the enthusiasm and learning effectiveness of different users to accurately cluster each user, thereby improving the accuracy of the generated educational reports.

[0109] Furthermore, referring to Figure 5 Based on the first embodiment of this application, another embodiment of this application is provided. In this embodiment, step S20, which calculates the learning motivation of each user based on the degree of interest of each user in the learning behavior data towards the content being learned, includes:

[0110] Step S21: Calculate the homework completion rate of each user across all watched courses, and the number of comments sent across all watched courses.

[0111] Step S22: Calculate the after-class knowledge consolidation factor for each user based on the homework completion rate and the number of comments sent.

[0112] As an example, the post-class knowledge consolidation factor is used to represent a user's mastery of the course. The higher the homework completion rate and the more frequent the comments in the discussion area, the better the consolidation effect of post-class knowledge.

[0113] As an example, by statistically analyzing the average homework completion rate of users across all watched courses, we obtain the first average (for users who have watched the chapter's instructional video but haven't done the homework, the homework completion rate is 0). Then, we normalize the number of comments sent by users and add one to obtain the second average corresponding to the number of comments sent (that is, for users who have watched the chapter's instructional video but haven't posted a comment, we set the second average to 1; for users who have posted comments, the second average is between 1 and 2). The sum of the first and second averages is recorded as the post-class knowledge consolidation factor for the current user. Similarly, the post-class knowledge consolidation factors for other users can be calculated.

[0114] Step S23: For any user, determine the number of interactions between the current user's comments and other users, and calculate the interaction effect value between the current user and other users based on the number of interactions.

[0115] As an example, different comments in the after-class discussion area elicit varying levels of discussion from other users. Comments that provide insightful and in-depth analysis of the lecture content will encourage active discussion, replies, and likes from other users, allowing them to further solidify their understanding of the knowledge points discussed and expand their knowledge. In other words, the comment prompts users to repeatedly learn the course content within the comment area. Conversely, comments that provide inaccurate or completely incorrect analysis of the lecture content, or comments unrelated to the course (such as advertisements or insulting remarks), often fail to elicit replies, discussions, or likes from other users. They are often ignored or even reported, meaning these comments do not encourage users to repeatedly learn the course content within the comment area. Therefore, further analysis of the after-class discussion area interaction is needed to adjust the after-class knowledge consolidation factors.

[0116] As an example, the number of interactions can be the number of replies, likes, and reports.

[0117] As an example, the interaction effect value represents the effect of each comment posted by the current user on the discussion among other users. Taking user A as an example, the interaction effect value F can be calculated as follows:

[0118]

[0119] Where norm is a linear normalization function, G1 is the normalized value of the number of replies to each comment sent by user A, G2 is the normalized value of the number of likes to each comment sent by user A, and G3 is the normalized value of the number of reports to each comment sent by user A. The addition of 1 to the denominator is to prevent the denominator from being 0. avg is an average value function, meaning that the more replies and likes and the fewer reports, the stronger the comment's effectiveness. Therefore, the avg function is used to calculate the average effectiveness of all comments sent by user A, which serves as the effect of each comment sent by user A on triggering discussion among other users, i.e., the interaction effect value.

[0120] Among them, the min-max normalization method was used to normalize the number of replies, likes, and reports under all comments.

[0121] Step S24: Based on the interaction effect value, adjust the after-class knowledge consolidation factor to obtain the after-class learning performance factor for each user.

[0122] As an example, the post-class learning performance factor represents a user's performance after watching a course. The post-class knowledge consolidation factor is adjusted using the interaction effect value (the effect of each comment triggering discussion among other users). The normalized value of the product of the interaction effect value and the user's post-class knowledge consolidation factor (processed using the norm linear normalization function) is denoted as the post-class learning performance factor of user A.

[0123] Step S25: Calculate the learning motivation level of each user based on their level of interest and post-class learning performance factors.

[0124] As an example, the product of a user's after-school learning performance factor and the user's interest in the learning content can be denoted as the user's learning motivation. Similarly, the learning motivation of each user can be calculated.

[0125] After step S25, which calculates each user's learning motivation based on interest level and after-class learning performance factors, the following steps are also included:

[0126] Determine the maximum and minimum values ​​of learning motivation among all users in the current user's cluster.

[0127] As an example, since different learning content has different requirements for the learning process, if we directly use learning motivation to compare and analyze users with different learning content, the error will be large. Therefore, we first determine the relative learning motivation of each user based on the comparative analysis between users with the same learning content.

[0128] As an example, the current user belongs to the first cluster. In the current first cluster, the relative learning enthusiasm of each user in the cluster is calculated.

[0129] For any given user, the relative learning motivation level is calculated based on the maximum value, minimum value, and learning motivation level.

[0130] As an example, taking user A as an example, the relative learning motivation T can be calculated as follows:

[0131]

[0132] Where t represents the learning motivation of user A, and tmin and tmax represent the minimum and maximum learning motivation values ​​among all users in the cluster to which user A belongs, respectively.

[0133] Based on each user's assignment completion metrics and learning engagement, the degree to which each user demonstrates their learning effectiveness is determined, including:

[0134] Based on each user's assignment completion metrics and relative learning enthusiasm, the degree of learning effectiveness demonstrated by each user is determined.

[0135] As an example, after adjusting for learning motivation, the degree of learning effectiveness for each user is determined based on their homework completion metrics and relative learning motivation.

[0136] Among these, the degree of learning effectiveness for each user is determined based on their homework completion metrics and relative learning enthusiasm, including:

[0137] Based on each user's homework completion metrics, the learning effectiveness value for each chapter of the course is calculated. The homework completion metrics include homework completion rate, homework completion time, and homework grade.

[0138] As an example, due to differences in individual learning abilities, greater enthusiasm for learning does not necessarily lead to better learning outcomes. As the course progresses and the difficulty of the learning content gradually increases, users with weaker learning abilities, even with significant effort (repeatedly watching course videos, diligently completing assignments, and actively participating in discussions), experience minimal or even regressive improvement in learning outcomes. Therefore, further analysis is needed to determine whether the learning outcomes, based on active learning, meet expectations. The completion scores and time of users' assignments can directly reflect their learning efficiency.

[0139] As an example, for any user B, the formula for calculating the learning effectiveness value H of each chapter from the completion status of homework for all watched chapters in all courses that user B has watched is:

[0140]

[0141] Where M is the completion rate of user B's homework for each chapter, N is the homework score of user B for each chapter, U is the design completion time of the homework for each chapter, and U1 is the actual completion time of user B's homework for each chapter.

[0142] The higher the completion rate of each chapter's homework (the closer to 100%), the higher the grade (the closer to 1), and the shorter the time spent on it. The closer the value is to 1), the better the user's learning effect in that chapter. For those who have watched the chapter's teaching videos but have not done the homework, the learning effect value for that chapter is set to 0.

[0143] Based on relative learning motivation and learning effectiveness, the overall learning efficiency of each user is calculated.

[0144] As an example, the average learning effectiveness value of all chapters and lessons currently viewed by the user is calculated. The product of this average learning effectiveness value and the user's relative learning enthusiasm is recorded as the user's overall learning efficiency. Similarly, the overall learning efficiency of other users can be calculated.

[0145] The level of understanding of knowledge points is calculated based on the degree of correlation between the homework questions in each chapter.

[0146] As an example, the analysis of overall learning efficiency did not consider the change in learning efficiency over time. That is, there may be situations where the learning efficiency is better in the early stages of a course, but decreases in the later stages as the knowledge is gradually accumulated and the learning difficulty gradually increases.

[0147] Therefore, it is necessary to further analyze the changes in learning efficiency over time and adjust the overall learning efficiency accordingly. This requires calculating the user's level of understanding of the knowledge points and adjusting the overall learning efficiency based on the user's level of understanding.

[0148] As an example, the knowledge point comprehension level value represents the trend of the depth of learning and understanding required for the knowledge points in the question. For instance, when a homework question contains core knowledge points, it may also contain other knowledge points. The depth of understanding required for this type of homework question will be different, and this trend value will also vary depending on the time the user spends on the question.

[0149] The degree of relevance includes potential relevance. The steps for calculating the level of understanding of knowledge points based on the degree of relevance between the homework questions in each chapter include:

[0150] By using a pre-defined association rule mining algorithm, the homework questions for each chapter are processed to determine the potential relationships between them.

[0151] As an example, the pre-defined association rule mining algorithm could be the Apriori algorithm (a well-known technique), which processes the homework questions for each chapter of a course to determine the potential associations between the homework questions.

[0152] For example, if homework assignments include questions A, B, and C, the association rules between these assignments can be calculated. For a dataset {A,B,C}, the generated association rules could include A→B,C; B→A,C; A,B→C, etc. Then, the confidence score of each association rule is calculated. The formula for calculating the confidence score is: Confidence (A→B) = Support (A∪B) / Support (A). Support measures the frequency of occurrences in a dataset, while confidence measures the reliability of the association rule. For example, if Support (A∪B) is 0.4 and Support (A) is 0.5, then the confidence score (A→B) is 0.8. This confidence score is used to represent the potential association between homework assignments, and only confidence scores greater than a certain threshold (e.g., 0.6) are retained. Since the specific implementation process of using the Apriori algorithm to calculate the potential association between homework assignments is existing technology, it will not be elaborated here.

[0153] For any given user, based on potential correlations, the completed questions of the current user are clustered to obtain multiple question clusters and the membership degree of each completed question to each question cluster.

[0154] As an example, the potential correlation corresponds to a numerical value. The inverse proportion (i.e., the reciprocal) of the potential correlation between any two completed questions is used as the clustering distance. The fuzzy C-Means (FCM) algorithm (a well-known technique) is used to cluster all completed questions, resulting in multiple question clusters and the membership degree of each completed question to each question cluster.

[0155] As an example, questions in each question cluster share similar knowledge points. Since the same knowledge point can be tested through different question types, it helps users understand and apply it from multiple perspectives. Some comprehensive questions can cover multiple knowledge points, requiring users to master multiple knowledge points simultaneously. Therefore, this solution uses a fuzzy clustering algorithm to divide completed questions into question clusters with similar knowledge points. The fuzzy clustering algorithm allows data points to belong to multiple clusters with different membership degrees. Membership degree represents the degree to which a data point belongs to a certain cluster. Therefore, comprehensive questions containing multiple knowledge points can belong to multiple question clusters, but with a relatively low membership degree.

[0156] As an example, knowledge points are usually organized in a hierarchical structure, from basic concepts to advanced applications. For instance, in mathematics, basic arithmetic is the foundation of algebra, and algebra is the foundation of calculus. When designing questions, the difficulty can be gradually increased according to this hierarchical relationship. Furthermore, some knowledge points require users to master relevant prerequisite knowledge first. For example, before learning trigonometric functions, students need to understand basic geometric concepts. Questions can be designed to review prerequisite knowledge before introducing new knowledge points. Therefore, it is necessary to further analyze the degree of change in user learning feedback as the depth of understanding required for the same knowledge point gradually increases with the increase in learning time.

[0157] Based on the question clusters, construct the sequence of completed questions.

[0158] As an example, in any question cluster, a sequence of completed questions is constructed in chronological order based on the completion time of each completed question for the current user.

[0159] For example, in the b-th question cluster, a sequence of completed questions is constructed according to the completion time of each completed question (completed by user B, and the same applies to other users) in chronological order.

[0160] Based on the score and set score of each completed question in the sequence of completed questions, the relative score of each completed question is calculated.

[0161] As an example, taking user B's completion of questions as an example, in the sequence of completed questions, the ratio of the score of each completed question (completed by user B) to the set score of that question is obtained and recorded as the relative score of each completed question.

[0162] For any question cluster, based on the membership degree and the completion time of each completed question for the current user, a linear fit is performed on the relative score to obtain the knowledge point comprehension value of the current question cluster.

[0163] As an example, in the sequence of completed questions, the membership degree of each completed question to the b-th question cluster is used as the fitting weight, the completion time of each completed question (completed by user B) is used as the horizontal axis, and the relative score of each completed question is used as the vertical axis. Using the weighted least squares method (a known technique), a straight line is fitted to the relative scores of all completed questions, and the slope of the fitted line is obtained. This slope is recorded as the knowledge point comprehension value of the b-th question cluster. Similarly, the knowledge point comprehension values ​​of other question clusters can be obtained.

[0164] As an example, a higher membership degree indicates that the question is more likely to contain only the core knowledge points of the question cluster, while a lower membership degree indicates that the question is more likely to contain knowledge points other than the core knowledge points of the question cluster. Therefore, using membership degree as the fitting weight, if the relative scores of completed questions with similar knowledge points gradually decrease over time, that is, the smaller the slope of the fitted line, it indicates that the depth of learning and understanding of the same knowledge point gradually increases, and users gradually fall behind the learning progress, resulting in a decline in learning effectiveness. Conversely, the larger the slope of the fitted line, the more likely that users are able to keep up with the learning progress and ensure learning effectiveness.

[0165] The learning outcomes of each user are determined based on their overall learning efficiency and understanding of the knowledge points.

[0166] As an example, by using the level of understanding of knowledge points from various types of completed questions, the overall learning efficiency is adjusted to obtain the degree of learning effect demonstrated by the user under active learning conditions.

[0167] Among them, the learning effect of each user is obtained based on the comprehensive learning efficiency and the degree of understanding of knowledge points, including:

[0168] For any question cluster, determine the first mean among the membership degrees of each completed question in the current question cluster.

[0169] As an example, taking user B as an example, each completed question will correspond to a membership degree for the question cluster, and the first mean is the mean membership degree of each completed question.

[0170] Based on the first mean, the weight value of the knowledge point comprehension level is calculated.

[0171] As an example, the weight value corresponds to the weight of the understanding of knowledge points when calculating the degree of learning effectiveness.

[0172] Based on the weight values ​​and the degree of understanding of knowledge points, the overall learning efficiency is adjusted to obtain the degree of learning effect of each user under active learning.

[0173] As an example, taking user B as an example, the calculation method for the degree of learning effectiveness Q can be:

[0174]

[0175] Where D represents user B's overall learning efficiency. The larger the value, the more active user B is in learning and the higher the overall learning efficiency.

[0176] K jThis represents the level of understanding of the knowledge points in the j-th question cluster corresponding to user B. The larger the value, the higher the learning efficiency of user B. This means that user B's learning efficiency does not decrease as the learning depth of the same knowledge point increases, that is, the learning efficiency gradually increases as the learning progresses.

[0177] J represents the number of question clusters corresponding to user B, and V represents... j V is the first mean of the membership degrees of all completed questions belonging to the j-th question cluster corresponding to user B. j The larger the value, the more consistent the knowledge points tested in all completed questions within the j-th question cluster corresponding to user B. Therefore, with... As K j The weighted values ​​are calculated using a weighted average. Adjust D to obtain the degree of learning effect of user B under active learning, and norm() represents the normalization calculation.

[0178] In this embodiment, the user's overall learning efficiency is adjusted by calculating the degree of understanding of knowledge points, thereby obtaining the degree of learning effect of the user under active learning, so as to accurately cluster each user according to the degree of learning effect.

[0179] The overall execution flow diagram of the embodiments of this application is shown below. Figure 6 As shown, the learning behavior data of users is first collected from the online education platform. Then, based on the similarity of the learning content of different users, clusters are formed to determine the relative learning enthusiasm of users in the clusters. Finally, based on the users' overall learning efficiency and the degree of learning effect, all users are clustered to obtain updated clusters. Based on the updated clusters, an education report is generated using a written R script.

[0180] This application embodiment also provides an educational report generation system based on learning behavior data, the educational report generation system based on learning behavior data includes:

[0181] The acquisition module is used to acquire learning behavior data of each user in the online education platform.

[0182] The calculation module calculates each user's learning motivation based on their interest in the learning content from the learning behavior data.

[0183] The module determines the extent to which each user's learning outcomes are demonstrated based on their homework completion metrics and learning enthusiasm.

[0184] The processing module clusters the data of each user and their learning behavior based on the degree of learning effectiveness, resulting in multiple final clusters and generating educational reports corresponding to the final clusters.

[0185] Reference Figure 7 , Figure 7 This is a schematic diagram of the device structure of the hardware operating environment involved in the embodiments of this application.

[0186] like Figure 7 As shown, the educational report generation device based on learning behavior data may include: a processor 1001, a memory 1005, and a communication bus 1002. The communication bus 1002 is used to realize the connection and communication between the processor 1001 and the memory 1005.

[0187] Optionally, the educational report generation device based on learning behavior data may also include a user interface, a network interface, a camera, RF (Radio Frequency) circuitry, sensors, a WiFi module, etc. The user interface may include a display screen and an input submodule such as a keyboard; optional user interfaces may also include standard wired or wireless interfaces. The network interface may include standard wired or wireless interfaces (such as a Wi-Fi interface).

[0188] Those skilled in the art will understand that Figure 7 The structure of the educational report generation device based on learning behavior data shown in the figure does not constitute a limitation on the educational report generation device based on learning behavior data. It may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0189] like Figure 7 As shown, the memory 1005, serving as a storage medium, may include an operating system, a network communication module, and an educational report generation program based on learning behavior data. The operating system is a program that manages and controls the hardware and software resources of the educational report generation device based on learning behavior data, supporting the operation of the educational report generation program based on learning behavior data and other software and / or programs. The network communication module is used to enable communication between the various components within the memory 1005, as well as communication with other hardware and software in the educational report generation system based on learning behavior data.

[0190] exist Figure 7 In the educational report generation device based on learning behavior data shown, the processor 1001 is used to execute the educational report generation program based on learning behavior data stored in the memory 1005 to implement the steps of the educational report generation method based on learning behavior data described above.

[0191] The specific implementation of the educational report generation device based on learning behavior data in this application is basically the same as the embodiments of the educational report generation method based on learning behavior data described above, and will not be repeated here.

[0192] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0193] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0194] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods of the various embodiments of this application.

[0195] The above are merely preferred embodiments of this application and do not limit the scope of this application. Any equivalent structural or procedural transformations made based on the description and drawings of this application, or direct or indirect applications in other related technical fields, are similarly included within the scope of protection of this application.

[0196] 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. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0197] 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.

Claims

1. A method for generating educational reports based on learning behavior data, characterized in that, The method includes: Acquire learning behavior data of each user in the online education platform; determine the clustering distance based on the similarity of learning content in the learning behavior data, and perform clustering processing on each user using the clustering distance to obtain multiple first clusters; Based on the degree of interest of each user in the learning behavior data towards the learning content, the learning enthusiasm of each user is calculated. Based on the task completion indicators and relative learning enthusiasm of each user, the degree of learning performance of each user is determined, wherein the relative learning enthusiasm is obtained based on the learning enthusiasm of users in each of the first clusters; The step of determining the degree of learning effectiveness of each user based on their task completion indicators and relative learning enthusiasm includes: Based on the homework completion indicators of each user, the learning effect value of each user for each chapter of the course is calculated. The homework completion indicators include homework completion rate, homework completion time, and homework score. Based on the relative learning enthusiasm and the learning effect value, the comprehensive learning efficiency of each user is calculated. The level of understanding of knowledge points is calculated based on the degree of correlation between the homework questions in each chapter. The degree of correlation includes potential correlation. The step of calculating the level of understanding of knowledge points based on the degree of correlation between the homework questions of each chapter includes: The homework questions for each chapter are processed using a pre-defined association rule mining algorithm to determine the potential associations between the homework questions. For any of the aforementioned users, based on the potential correlation, the completed questions of the current user are clustered to obtain multiple question clusters and the membership degree of each completed question to each question cluster; Based on the question clusters, construct a sequence of completed questions; Based on the score and set score of each completed question in the completed question sequence, the relative score of each completed question is calculated. For any of the question clusters, based on the membership degree and the completion time of each completed question for the current user, a linear fit is performed on the relative score to obtain the knowledge point comprehension value of the current question cluster. Based on the comprehensive learning efficiency and knowledge point comprehension values, the degree of learning effect demonstration for each user is obtained; Based on the degree of learning effectiveness, the data of each user and the learning behavior are clustered to obtain multiple final clusters, and an educational report corresponding to each final cluster is generated.

2. The method for generating educational reports based on learning behavior data as described in claim 1, characterized in that, Before calculating the learning motivation of each user based on their interest in the learning content from the learning behavior data, the method further includes: Extract the set of courses watched by each user from the learning behavior data; Determine the intersection and union of course sets between any two users; Based on the course intersection and the course union, the similarity of learning content between any two users is calculated. Based on the similarity of the learning content, the degree of interest of each user in the learning content is calculated.

3. The method for generating educational reports based on learning behavior data as described in claim 2, characterized in that, The step of calculating the degree of interest of each user in the learned content based on the similarity of the learned content includes: Based on the similarity of the learned content, the clustering distance is determined, and the users are clustered using the clustering distance to obtain multiple first clusters; For any of the first clusters, extract the first duration of the courses watched by each user, and the second duration required to watch the courses without repeating them; Based on the first duration and the second duration, the degree of interest of each user in the content being learned is determined.

4. The method for generating educational reports based on learning behavior data as described in claim 1, characterized in that, The calculation of each user's learning motivation based on their interest in the learning content from the learning behavior data includes: The statistics show the homework completion rate of each user across all viewed courses, as well as the number of comments sent across all viewed courses. Based on the homework completion rate and the number of comments sent, the after-class knowledge consolidation factor for each user is calculated; For any of the aforementioned users, determine the number of interactions between the comments sent by the current user and other users, and calculate the interaction effect value between the current user and other users based on the number of interactions; Based on the interaction effect value, the after-class knowledge consolidation factor is adjusted to obtain the after-class learning performance factor for each user. Based on the level of interest and the post-class learning performance factor, the learning enthusiasm of each user is calculated.

5. The method for generating educational reports based on learning behavior data as described in claim 4, characterized in that, After calculating the learning enthusiasm of each user based on the level of interest and the after-class learning performance factor, the method further includes: Determine the maximum and minimum values ​​of learning motivation among all users in the current user's cluster; For any of the users, the relative learning motivation of each user is calculated based on the maximum value, the minimum value, and the learning motivation level.

6. The method for generating educational reports based on learning behavior data as described in claim 1, characterized in that, The process of determining the learning outcome level for each user based on the overall learning efficiency and knowledge point comprehension values ​​includes: For any of the aforementioned question clusters, determine the first mean among the membership degrees of each of the completed questions in the current question cluster; Based on the first mean, the weight value of the understanding level of the knowledge point is calculated; Based on the weight values ​​and the knowledge point comprehension values, the overall learning efficiency is adjusted to obtain the degree of learning effect of each user under active learning.

7. An educational report generation system based on learning behavior data, characterized in that, For performing the educational report generation method based on learning behavior data as described in any one of claims 1 to 6, the educational report generation system based on learning behavior data comprises: The acquisition module is used to acquire learning behavior data of each user in the online education platform; The calculation module calculates the learning motivation of each user based on the degree of interest of each user in the learning behavior data regarding the content being learned; The determining module determines the degree of learning effectiveness of each user based on the user's task completion indicators and learning enthusiasm. The processing module performs clustering processing on each user and the learning behavior data based on the degree of learning effect demonstration, obtains multiple final clusters, and generates an education report corresponding to the final clusters.

8. A computer storage medium, characterized in that, The computer storage medium stores an educational report generation program based on learning behavior data, which, when executed by a processor, implements the steps of the educational report generation method based on learning behavior data as described in any one of claims 1 to 6.

Images