A student social visual analysis method based on hierarchical hypergraph

Abstract

The application relates to an educational field and relates to a student social visual analysis method based on a hierarchical hypergraph. Student social relationships are closely related to their physical and mental health, academic performance and the like, but research on visual analysis of student campus social relationships is less, and in-depth analysis of high-order community relationships of the students is less, and the complex social state of mutual influence between the students cannot be expressed. Therefore, the application provides a student social visual analysis method based on a hierarchical hypergraph, a social relationship network is constructed according to the space-time co-occurrence characteristics of the students, Louvain is used for community division and community fusion of the student social relationships. Through different view analysis, community member distribution and behavior time distribution, individual social member distribution and behavior time distribution and correlation between individual behavior characteristics and performance are explored. The application carries out case analysis research and expert evaluation based on student basic information, campus behavior data and academic performance, and verifies the ease of use and effectiveness of the application.

Description

Technical Field

[0001] This invention relates to hierarchical hypergraph technology and visualization technology, utilizing students' basic information, campus behavior data, and academic performance to conduct visual analysis of students' social relationships and behavioral characteristics within their communities. This technology can be widely applied by education administrators and other relevant personnel to objectively and intuitively analyze and study the correlation between students' campus social relationships and their physical and mental health, academic performance, and team building. Background Technology

[0002] In the field of education, students' social relationships are closely related to their physical and mental health, academic performance, and so on. Therefore, a harmonious social atmosphere helps to improve the quality of education and teaching. For administrators and teachers of ordinary universities, they are concerned about the actual social relationships of students on campus. They usually use methods such as observation and interviews to understand the social relationships among students and use statistical techniques for analysis. However, this method cannot objectively, conveniently, and deeply analyze students' campus social relationships.

[0003] With the construction of smart campuses, various behavioral data generated by students on campus are recorded and stored, such as dining behavior data, shopping behavior data, library borrowing data, and building entry and exit behavior data. This behavioral data with spatiotemporal information helps us analyze students' campus social relationships. However, there is a limited amount of research in this area, mainly analyzing one-on-one first-order social relationships among students. It does not delve into the higher-order community relationships formed by several closely connected students, failing to express the complex social states of mutual influence among students. Therefore, it is necessary to conduct research on students' community relationships.

[0004] In addition, there is very little research specifically on the visual analysis of students' campus social relationships. People can only rely on general tools to conduct the analysis, which greatly affects the efficiency of the analysis. Summary of the Invention

[0005] To address the aforementioned problems, this invention proposes a student social visualization analysis method based on a hierarchical hypergraph. This method utilizes students' basic information, campus behavior data, and academic performance to analyze students' social relationships in an intuitive, interactive, and multi-perspective manner. The method centers on a hierarchical community view of the social network, supplemented by a community social relationship analysis view, an individual social relationship analysis view, and a behavioral feature view, supporting users in flexibly analyzing students' social relationships. This invention has the following innovations:

[0006] 1. A visualization method based on hierarchical hypergraph to represent students' high-order social relationships is proposed, which breaks through the traditional one-to-one social relationship representation. It can discover student communities from the complex campus social relationship network and understand the integration process of different communities. At the same time, a student social visualization analysis system is developed.

[0007] 2. A method is proposed to display students' behavioral characteristics using a behavioral feature matrix view, which facilitates understanding the behavioral characteristics of individual students and analyzing the correlation between behavioral characteristics and academic performance.

[0008] This invention uses multi-source behavioral data of students on campus to objectively calculate the strength of social relationships among students, and uses interactive links of community hierarchy view, community social relationship analysis view, personal social relationship analysis view and behavioral feature matrix view to assist education administrators in intuitively and conveniently to deeply analyze high-level community relationships among students and the process of community integration, and to support the work of team building, individual social anomaly detection and analysis of factors affecting academic performance.

[0009] During the testing process, the developers designed a questionnaire using a Likert scale and invited 15 faculty and students from Beijing University of Technology to complete the assigned tasks and fill out the questionnaire using the system. The feedback from faculty and students indicated that no prior knowledge of statistics was required; the system provided an intuitive and interactive way to analyze the social relationships of students in a class. In particular, it allowed for in-depth understanding of the composition and characteristics of the class community, evaluation of each student's social relationships, exploration of the correlation between social relationships and academic performance, and early warning systems for students with abnormal social relationships. This provides valuable information for both administrators and students. Attached Figure Description

[0010] Figure 1 This is the initial interface of the visualization system of the present invention.

[0011] Figure 2 This is a hierarchical community view of social relationships in the visualization system of this invention.

[0012] Figure 3(a) is a social relationship diagram of the two-layer community integration in the visualization system of the present invention.

[0013] Figure 3(b) is a social relationship diagram of the three-layer community integration in the visualization system of the present invention.

[0014] Figure 4 This is a community social relationship analysis view in the visualization system of the present invention.

[0015] Figure 5 This is a view for analyzing personal social relationships in the visualization system of this invention.

[0016] Figure 6 This is a view of individual student behavior characteristics in the visualization system of this invention. Detailed Implementation

[0017] Step 1: Import student behavior data from different information management systems

[0018] The invention utilizes tools to aggregate student behavior data stored in different information management systems within the school into its database, including dining behavior data, shopping behavior data, building entry and exit data, library borrowing data, and gateway billing data.

[0019] Step 2: Select the student group and time range to be queried and analyzed in the visual system.

[0020] In the search box on the system's initial interface, select [Department / School], [Major], [Class], [Academic Year], or [Semester] to search.

[0021] Step 3: Building Social Relationships

[0022] The hierarchical community view of social networks displays a social relationship graph formed from retrieved student data, such as... Figure 2 As shown, this part utilizes the force-guided layout network graph of ZoomCharts. Each node in the graph represents a student, node color indicates dormitory information (nodes of the same color belong to the same dormitory), and node outline style indicates gender information (smooth circular outline nodes represent males, petal-shaped outline nodes represent females). Node size is calculated using betweenness centrality to represent the student's importance in the social network. The thickness of the lines connecting nodes represents the strength of the social relationship between students; thicker lines indicate a stronger social relationship. Social relationship strength and betweenness centrality calculations are required when constructing social relationships.

[0023] (1) Calculation of social relationship strength

[0024] Based on students' behavioral data, a co-occurrence tensor containing three dimensions—location, date, and time—is constructed for each pair of students to accurately record the number of times they co-occur in different times and spaces. Then, based on this tensor, two co-occurrence features—co-occurrence spatiotemporal diversity and weighted co-occurrence frequency—are extracted. Subsequently, a linear regression method is used to predict the strength of social relationships between students and to construct a student social relationship network.

[0025] 1) First, establish a spatiotemporal activity sequence for each student {<s,l,d,t>}, where s∈{s1,s2,...,s M Let} represent the student set, l∈{l1,l2,...,l...} N Let d ∈ {d1, d2, ..., d} represent the set of activity locations. R} represents the event date, t∈{t1,t2,...,t K} indicates the event time.

[0026] 2) Calculate the spatiotemporal diversity of co-occurrence based on Ruili entropy. Let c represent the Ruili entropy of the co-occurrence event of students i and j in three-dimensional spacetime. ij,l,d,tThis represents students i and j in three-dimensional spacetime.<l,d,t> Co-occurrence frequency, This represents the total co-occurrence frequency of students i and j. q is set to 0.1 to suppress the impact of frequent encounters between the two students in hotspot spatiotemporal regions on co-occurrence diversity.

[0027]

[0028] 3) Calculate the weighted co-occurrence frequency. First, use Shannon entropy to measure the spatiotemporal region.<l,d,t> The popularity of P is shown in Formula 2). s,l,d,t Representing spacetime<l,d,t> The probability of a location being visited by student s; hotspot areas are visited by many students; entropy value H. l,d,t The entropy value is relatively high in non-hotspot areas, while it is relatively low in non-hotspot areas. Based on the entropy value of the spatiotemporal region, the frequency of student co-occurrence is weighted, as shown in Formula 3), where... The weights represent the number of times a region becomes active, with less weight for hotspot areas and more weight for non-hotspot areas. This effectively adjusts the contribution of co-occurrence frequency to social intensity.

[0029]

[0030]

[0031] 4) Calculate the social strength value. Based on the two features of spatiotemporal co-occurrence diversity and weighted co-occurrence frequency, the social strength value among students is calculated using linear regression, as shown in Formula 4). Here, α, β, and ε are calculated from the students' binary social relationship labels (friends and classmates) using Katz scores. α and β represent the weights of the two features, respectively, and ε represents the offset. For a student group, once the social strength between any two students is calculated, a weighted undirected graph G = (S, E, W) representing the social relationships of the group can be constructed, where S represents the set of students, E represents the set of social relationship edges between students, and W represents the edge weight w. ij gather.

[0032] w ij =α*D ij +β*F ij +ε 4)

[0033] (2) Calculation of betweenness centrality. σ st σ represents the number of shortest paths from node s to node t. st (v) represents the number of nodes v in the shortest path from node s to node t, and V represents the set of student nodes. Students with high betweenness centrality can act as bridges, helping to enhance the cohesion of the class or group.

[0034]

[0035] Step 4: Community Division and Community Integration

[0036] In the social relationship graph, the color of the block at the bottom of the node represents community information, and students with the same block color belong to the same community.

[0037] (1) Community segmentation. The Louvain algorithm based on modularity is used to segment the community composition in the student social network and the hierarchical integration between communities to form larger communities.

[0038] 1) Modularity calculation. As shown in Formula 6, This represents the sum of the weights of all edges connected to student i, and similarly, k represents the sum of the weights of all edges connected to student i. j c represents the sum of the weights of all edges connected to student j; i Let c represent the community where student i resides, and similarly, c represents the community where student i resides. j This represents the community where student j resides; when students i and j belong to the same community, the function δ(c i ,c j The value of ) is 1 if it is equal to 1 otherwise it is 0. The value of Q represents the sum of the weights of all edges in the entire student social network. The modularity Q is located in the interval [-1, 1]. The larger the value, the better the quality of the community division.

[0039]

[0040] 2) Community Detection. For a social network with N students, the community detection process is as follows: First, each student is considered an independent community; initially, the number of communities is the same as the number of students. Then, for each student i, the modularity gain ΔQ resulting from moving them to the community C of each neighbor node is calculated sequentially, as shown in Equation 7). in ∑ represents the sum of the weights of all edges within community C. tot k represents the sum of the weights of all edges connected to nodes in community C. i,in This represents the sum of the weights of the edges connecting node i to nodes in community C. Finally, all gains are compared. If the largest gain is greater than zero, student i is moved to the community corresponding to the largest gain. Otherwise, student i remains stationary. The above operation is repeated for all student nodes until all nodes remain stationary, and the community division is completed.

[0041]

[0042] (2) Community Merging. Based on the detected student community structure, all nodes within the same community are merged into a new node. The sum of the weights of the edges between the two communities is used as the weight of the edges between the new node, while the sum of the weights of the edges within the community is assigned to the new node's self-loop edges, thus generating a new social network. Community detection is performed again on the new social network to obtain the second-layer community structure. The community detection and community merging operations are iteratively performed until the modularity of the entire network no longer changes. Figure 3(a) , 3(b) As shown.

[0043] Step 5: Analyze and view the distribution of community members and the distribution of their behavior over time in the community social relationship analysis view, and compare the differences between communities.

[0044] The community social relationship analysis view uses Echarts' polar coordinate stacked plot and polar coordinate scatter plot to display the distribution of student community members and the temporal distribution of various behaviors, respectively. This provides a visual understanding of the characteristics of a single community and allows for comparison of differences between multiple communities. Figure 4 As shown.

[0045] The polar coordinate stacked graph displays the distribution of community members from four dimensions: job title, gender, dormitory, and academic performance. Users can understand the distribution of each community member in each dimension, such as how many boys and girls there are; how many dormitories they live in and how many people are in each dormitory; how many students have excellent academic performance and how many students have poor academic performance; whether there are class leaders, etc., thereby measuring the diversity of the community, analyzing the reasons for the formation of the community, and whether the community is conducive to the growth of students.

[0046] The polar scatter plot displays the temporal distribution of community members' behaviors. The radius coordinates represent the academic calendar dates, with the innermost point representing the first day of the semester, and so on, with the outermost point representing the last day. The angle coordinates represent time. Each point represents a single behavior record, with different colors representing different behaviors, including dining, shopping, and showering. Users can understand the temporal distribution of various community members' behaviors and determine the frequency and regularity of certain behaviors.

[0047] Step 6: In the personal social relationship analysis view, analyze and view the distribution of each student's social members and the distribution of their behavioral time, and compare the differences between individuals.

[0048] The individual social relationship analysis view and the community social relationship analysis view use the same visualization components, namely, a stacked polar coordinate graph to display the distribution of a student's social members from multiple dimensions, measuring the diversity, activity level, and degree of isolation in their social relationships. The difference is that the social members in this view are no longer limited to the student's local community but extend to the entire social network, and dynamically change with the social strength threshold in the hierarchical community view. Simultaneously, a polar coordinate scatter plot is used to display the temporal distribution of the student's various behaviors, thereby determining characteristics such as the regularity of the student's lifestyle. Figure 5 As shown.

[0049] Step 7: Analyze and view students' behavioral characteristics in the behavioral characteristics view, compare the differences in behavioral characteristics between different individuals, and explore the correlation between behavioral characteristics and performance.

[0050] Behavioral feature views can effectively evaluate and compare the behavioral patterns of all students within a group, and are displayed using a behavioral feature matrix diagram, such as... Figure 6 As shown, the columns and rows of the matrix represent students and behavioral characteristics, respectively. Each column represents a student, with the column name indicating the student's ID number. The column name background color matches the color of the student's community, allowing users to intuitively understand the behavioral characteristics of students in each community. The bar above the column name represents the student's GPA. Each row represents a behavioral characteristic, such as breakfast frequency or internet usage. Matrix cells represent the values ​​of student behavioral characteristics; when the user moves the mouse over a cell, the specific characteristic value is displayed. Due to the differences in the magnitude of different characteristics and the significant differences in values ​​for the same characteristic among different students, a linear normalization method is used to convert the value range of each characteristic to [0,1], and then the values ​​are mapped to corresponding colors, with darker colors indicating higher characteristic values. This view supports interactive sorting of students by academic performance or behavioral characteristics to explore the correlation between behavioral characteristics and academic performance. Through this view, users can clearly understand each student's community, academic performance, behavioral characteristics, and differences among students.

Claims

1. A student social visualization analysis method based on hierarchical hypergraphs, characterized in that... Includes the following steps: Step 1: Import student behavior data from different information management systems Tools are used to aggregate student behavior data stored in different information management systems of the school into a database, including dining behavior data, shopping behavior data, building entry and exit data, library borrowing data, and gateway billing data; Step 2: Select the student group and time range to be queried and analyzed in the visual system. In the search box on the system's initial interface, select [Department (College)], [Major], [Class], [Academic Year], or [Semester] to search; Step 3: Building Social Relationships The hierarchical community view of the social network displays a social relationship graph formed from retrieved student data. This part is implemented using the force-guided layout network graph of Zoomcharts. Each node in the graph represents a student, the node color indicates dormitory information (nodes of the same color represent students in the same dormitory), and the node outline style indicates gender information (smooth circular outline nodes represent males, and petal-shaped outline nodes represent females). The node size is calculated using betweenness centrality to represent the importance of a student in the social network. The thickness of the lines connecting nodes represents the strength of the social relationship between students; thicker lines indicate a stronger social relationship between two students. Social relationship strength and betweenness centrality calculations are required when constructing social relationships. (1) Calculation of social relationship strength Based on students' behavioral data, a co-occurrence tensor containing three dimensions—location, date, and time—is constructed for each pair of students to accurately record the number of times they co-occur in different times and spaces. Then, based on this tensor, two co-occurrence features—co-occurrence spatiotemporal diversity and weighted co-occurrence frequency—are extracted. Subsequently, a linear regression method is used to predict the strength of social relationships between students and to construct a student social relationship network. 1) First, establish a spatiotemporal activity sequence for each student. ,in Represents a set of students. Indicates the meeting point for the activity. Indicates the date of the event. Indicates the event time; 2) Calculation of co-occurrence spatiotemporal diversity based on Ruili entropy; The Ruili entropy represents the co-occurrence event of students i and j in three-dimensional spacetime. This represents students i and j in three-dimensional spacetime. Co-occurrence frequency, This represents the total co-occurrence frequency of students i and j; q is set to 0.1 to suppress the impact of frequent encounters between the two students in hotspot spatiotemporal regions on co-occurrence diversity. 1) 3) Calculate the weighted co-occurrence frequency; first, use Shannon entropy to measure the spatiotemporal region. The popularity of [the product / service] is shown in Formula 2). Representing spacetime The probability of being visited by student s, hotspot areas being visited by many students, and entropy. The entropy values ​​are relatively high in non-hotspot areas, while they are relatively low in non-hotspot areas. Furthermore, based on the entropy values ​​of spatiotemporal regions, the frequency of student co-occurrence is weighted, as shown in Formula 3), where... The weights are indicated by the fact that hot spots have smaller weights and non-hot spots have larger weights, which can effectively adjust the contribution of co-occurrence frequency to social intensity. 2) 3) 4) Calculate the social strength value; based on the two characteristics of spatiotemporal co-occurrence diversity and weighted co-occurrence frequency, the linear regression method is used to calculate the social strength value among students, as shown in formula 4), where, , and The value was calculated using Katz scores from the student's binary social relationship labels, namely friends and ordinary classmates. , These represent the weights of the two features, This represents the offset; for a student group, after calculating the social strength between any two students, a weighted undirected graph representing the social relationships of the group is constructed. , Represents a set of students. This represents a set of social relationships among students. Represents edge weight gather; 4) (2) Calculation of betweenness centrality; This represents the number of shortest paths from node s to node t. This represents the number of nodes v that are passed through in the shortest path from node s to node t, where V represents the set of student nodes. Step 4: Community Division and Community Integration In the social relationship graph, the color of the block at the bottom of the node represents community information, and students with the same block color belong to the same community; (1) Community segmentation; using the Louvain algorithm based on modularity to segment the community composition in the student social network and the hierarchical integration between communities to form a larger community; 1) Modularity calculation; as shown in Formula 6, This represents the sum of the weights of all edges connected to student i, and similarly... This represents the sum of the weights of all edges connected to student j; This represents the community where student i resides, and similarly... Let represent the community where student j resides; when students i and j belong to the same community, the function... The value is 1 if it is equal to 1 otherwise it is 0. The modularity represents the sum of the weights of all edges in the entire student social network. The value is located in the interval [-1, 1], and the larger the value, the better the quality of the community division; 6） 2) Community detection; For a social network containing N students, the community detection process is as follows: First, each student is regarded as an independent community, that is, in the initial stage, the number of communities is the same as the number of students; Then, for each student i, the modularity gain ΔQ brought about by moving it to the community C of each neighbor node is calculated in turn, and the calculation process is shown in Formula 7). This represents the sum of the weights of all edges within community C. This represents the sum of the weights of all edges connected to nodes in community C. This represents the sum of the weights of the edges connecting node i to nodes in community C. Finally, all gains are compared. When the largest gain is greater than zero, student i is moved to the community corresponding to the largest gain. Otherwise, student i remains stationary. The above operation is repeated for all student nodes until all nodes remain stationary and the community division is complete. 7) (2) Community merging; Based on the detected student community structure, all nodes in the same community are merged into a new node. The sum of the weights of the edges between the two communities is used as the weight of the edges between the new node, while the sum of the weights of the edges within the community is assigned to the self-loop edges of the new node, thereby generating a new social network. On the new social network, community detection is performed again to obtain the second-layer community structure. The community detection and community merging operations are performed iteratively until the modularity of the entire network no longer changes. Step 5: Analyze and view the distribution of community members and the distribution of their behavior over time in the community social relationship analysis view, and compare the differences between communities. The community social relationship analysis view uses Echarts' polar coordinate stacked chart and polar coordinate scatter plot to show the distribution of student community members and the time distribution of various behaviors, which can intuitively understand the characteristics of a single community and compare the differences between multiple communities. The polar coordinate stacked graph displays the distribution of community members from four dimensions: job title, gender, dormitory, and academic performance. Users can understand the distribution of each community member in each dimension, including how many boys and girls there are; how many dormitories they live in and how many people are in each dormitory; how many students have excellent academic performance and how many students have poor academic performance; whether there are class leaders, and thus measure the diversity of the community, analyze the reasons for the formation of the community, and whether the community is conducive to the growth of students. The polar scatter plot shows the temporal distribution of community members' behaviors. The radius coordinates in the plot represent the school calendar dates, with the innermost point representing the first day of the semester, and so on, with the outermost point representing the last day of the semester; the angle coordinates represent time; each point represents a behavior record, and different colors represent different behaviors, including dining, shopping, and showering; users can understand the temporal distribution of various behaviors of community members and judge the frequency and regularity of a certain behavior; Step 6: In the personal social relationship analysis view, analyze and view the distribution of each student's social members and the distribution of their behavioral time, and compare the differences between individuals. The individual social relationship analysis view and the community social relationship analysis view use the same visualization components, namely, using a polar coordinate stacked graph to display the distribution of a student's social members from multiple dimensions, measuring the diversity, activity level, and isolation level of the student's social relationships. The difference is that the social members in this view are no longer limited to the local community, but extend to the entire relationship network, and will dynamically change as the social intensity threshold in the hierarchical community view changes. At the same time, a polar coordinate scatter plot is used to display the temporal distribution of the student's various behaviors, thereby judging whether the student's life is regular. Step 7: Analyze and view students' behavioral characteristics in the behavioral characteristics view, compare the differences in behavioral characteristics between different individuals, and explore the correlation between behavioral characteristics and academic performance; The Behavioral Characteristics View effectively evaluates and compares the behavioral patterns of all students within a group, presented using a behavioral characteristic matrix. The columns and rows of the matrix represent students and behavioral characteristics, respectively. Each column represents a student, with the column name indicated by their student ID. The column name background color matches the color of the student's community, allowing users to intuitively understand the behavioral characteristics of students in each community. The bars above the column names represent the students' average GPA. Each row represents a behavioral characteristic, including breakfast frequency and internet usage. Matrix cells represent the values ​​of student behavioral characteristics; when the user moves the mouse over a cell, the specific characteristic value is displayed. A linear normalization method is used to convert the value range of each characteristic to [0,1], and then the values ​​are mapped to corresponding colors, with darker colors indicating higher characteristic values. This view supports interactive sorting of students by academic performance or behavioral characteristics to explore the correlation between behavioral characteristics and academic performance. Through this view, users can clearly understand each student's community, academic performance, behavioral characteristics, and differences among students.

Images