Multi-scene interaction control method and system in iptv education platform

Abstract

The application relates to the technical field of data processing, and discloses a multi-scene interaction control method and system in an IPTV education platform. The method comprises the following steps: generating a user behavior data set through education behavior Latin hypercube sampling, obtaining a standardized education scene type based on three-dimensional similarity calculation, constructing an education space-time hypergraph to extract a scene feature vector, establishing a resource optimization model with the minimum delay cost as the target, and realizing pre-allocation of network bandwidth and server resources and rapid scene switching scheduling. The application solves the problem that the multi-scene interaction control in the IPTV education platform lacks specialized sampling, modeling and optimization technology. The application improves the response speed of the education scene switching and the intelligent level of the resource allocation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a multi-scene interaction control method and system in an IPTV education platform. BACKGROUND

[0002] With the rapid development of digital education, IPTV education platform as an important remote teaching carrier is widely used. The existing IPTV education platform usually adopts a scene management method based on channel switching, organizes different education contents in the form of channels, and users switch between different education scenes through a remote controller or simple interface operation. The traditional scene control method mainly relies on preset scene templates and fixed resource allocation strategies, provides scene navigation functions through an EPG system, distributes contents using a CDN network, and uses a QoS mechanism to guarantee basic network service quality. These technologies have certain stability and reliability in single-scene applications.

[0003] However, the existing technology has significant deficiencies in handling multi-scene concurrency and dynamic switching. First, the scene switching response delay problem is prominent. The traditional method needs to re-establish network connections and reallocate server resources, resulting in a scene switching time of 3-5 seconds, which seriously affects the continuity of teaching. Second, the resource scheduling algorithm lacks intelligence. The existing system uses a static average allocation strategy and cannot dynamically optimize according to the importance and real-time needs of different education scenes, resulting in key education scenes such as online exams not being given priority. Third, the scene relationship modeling is too simplified. The traditional graph structure can only represent the pairwise connection relationship between scenes and cannot accurately describe the complex teaching logic association between multiple education scenes, resulting in inaccurate scene recommendation and prediction. SUMMARY

[0004] The present application provides a multi-scene interaction control method and system in an IPTV education platform, which solves the problem of lack of specialized sampling, modeling, and optimization technology in multi-scene interaction control in an IPTV education platform. The present application improves the response speed of education scene switching and the intelligence level of resource allocation.

[0005] In a first aspect, the present application provides a multi-scene interaction control method in an IPTV education platform, which includes:

[0006] Step S1: using education behavior Latin hypercube sampling, sampling IPTV user behavior data according to education scene weight factors and learning progress correlation coefficients to generate education scene sampling data sets;

[0007] Step S2: calculating a scene clustering center according to content similarity, time similarity and user similarity of the education scene sampling data set, to obtain a standardized education scene type set;

[0008] Step S3: constructing an education space-time hypergraph with the standardized education scene type set as nodes, wherein a hyperedge connects a scene combination having a teaching logic correlation, to obtain a scene feature vector representation;

[0009] Step S4: establishing a resource optimization model with a target function being a product of scene response delay and education priority, according to the scene feature vector representation, to determine an IPTV scene resource allocation scheme;

[0010] Step S5: pre-allocating network bandwidth and server resources according to the IPTV scene resource allocation scheme, and performing resource scheduling when responding to a user scene switching request.

[0011] In a second aspect, the present application provides an IPTV education platform multi-scene interaction control system, which comprises:

[0012] a sampling module configured to sample IPTV user behavior data according to education scene weight factors and learning progress correlation coefficients by using an education behavior Latin hypercube sampling, to generate an education scene sampling data set;

[0013] a calculation module configured to calculate a scene clustering center according to content similarity, time similarity and user similarity of the education scene sampling data set, to obtain a standardized education scene type set;

[0014] a construction module configured to construct an education space-time hypergraph with the standardized education scene type set as nodes, wherein a hyperedge connects a scene combination having a teaching logic correlation, to obtain a scene feature vector representation;

[0015] an establishment module configured to establish a resource optimization model with a target function being a product of scene response delay and education priority, according to the scene feature vector representation, to determine an IPTV scene resource allocation scheme;

[0016] a response module configured to pre-allocate network bandwidth and server resources according to the IPTV scene resource allocation scheme, and perform resource scheduling when responding to a user scene switching request.

[0017] In a third aspect, an IPTV education platform multi-scene interaction control device is provided, which comprises a memory and at least one processor, the memory storing instructions; the at least one processor invokes the instructions in the memory, so that the IPTV education platform multi-scene interaction control device performs the IPTV education platform multi-scene interaction control method described above.

[0018] In a fourth aspect, a computer readable storage medium is provided, and the computer readable storage medium stores instructions which, when executed on a computer, cause the computer to perform the method for multi-scene interaction control in an IPTV education platform.

[0019] In the technical scheme provided in the present application, the technical problem that traditional random sampling cannot distinguish the importance of education scenes is effectively solved by using the education behavior Latin hypercube sampling technology. The sampling method introduces an education scene weight factor and a learning progress correlation coefficient, can perform hierarchical sampling on user behavior data according to the specific application requirements of the IPTV education platform, ensures that high importance scenes such as live classroom obtain more sampling attention, and at the same time takes into account the user personalized learning progress, and the generated education scene sampling data set more accurately reflects the actual characteristics of the IPTV education application. The three-dimensional clustering algorithm based on content similarity, time similarity and user similarity overcomes the limitation of single dimension of traditional clustering methods, quantifies the similarity difference between the same knowledge points, related knowledge points and different subject scenes, combines time window overlap analysis and user group feature comparison, realizes accurate identification and classification of education scene types, and forms a standardized education scene type set, which lays a solid foundation for subsequent scene relationship modeling. The construction of the education space-time hypergraph structure breaks through the technical limitation that the traditional graph structure can only represent pair connection, connects multiple scene nodes with teaching logic correlation through hyperedges, accurately describes the complex multi-scene correlation relationship in the IPTV education platform, and cooperates with the education scene mask matrix and the hypergraph neural network coding to generate a scene feature vector representation containing multi-dimensional information such as scene content, time, space and correlation.

[0020] The target function in the resource optimization model is designed to minimize the product of the scene response delay and the education priority, which effectively balances the importance and performance requirements of different education scenes. By setting different priority parameters for online exams, live classroom, interactive answering, resource downloading and other scenes, the key education activities are ensured to have priority in resource allocation. The bandwidth constraint condition ensures the feasibility of the resource allocation scheme and the stable operation of the system. The education scene-oriented optimization algorithm EDU-NGO shows significant advantages in the application of IPTV education platform. The algorithm is specially optimized for the specific functional requirements of education scenes. Through the evolution mechanism of the hawk population and the elite retention strategy, it can efficiently search for the optimal solution in the complex multi-constraint optimization space. In particular, the key education scene detection and automatic weight adjustment mechanism enables the algorithm to dynamically respond to real-time changes in education activities, ensuring that key scenes such as online exams always have sufficient resource allocation. The pre-allocated resource scheduling mechanism changes the inefficient mode of reapplying resources for scene switching in traditional IPTV systems by reserving network bandwidth by priority and allocating computing resources by scene type. The scene resource mapping table constructed by the mechanism realizes the direct mapping of scene identifiers to network channels and server nodes. When a user initiates a scene switching request, the pre-allocated resources can be quickly activated through the resource switching instruction, significantly shortening the scene response time and improving the user experience and teaching effect of the IPTV education platform. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor based on these drawings.

[0022] Figure 1 An embodiment of the multi-scene interactive control method in the IPTV education platform in the embodiments of the present application is shown in the figure.

[0023] Figure 2 An embodiment of the multi-scene interactive control system in the IPTV education platform in the embodiments of the present application is shown in the figure.

[0024] Figure 3 The structure schematic diagram of the multi-scene interactive control device in the IPTV education platform in the embodiments of the present application is shown in the figure. DETAILED DESCRIPTION

[0025] The embodiment of the present application provides a multi-scene interaction control method and system in an IPTV education platform. The terms "first", "second", "third", "fourth" and the like (if any) in the specification and claims of the present application and the above-described drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the term "comprising" or "having" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0026] For ease of understanding, the specific flow of the embodiment of the present application is described below. Please refer to Figure 1 One embodiment of the multi-scene interaction control method in the IPTV education platform in the embodiment of the present application includes the following steps.

[0027] Step S1: Latin hypercube sampling of educational behavior is adopted, IPTV user behavior data is sampled according to the education scene weight factor and the learning progress correlation coefficient, and an education scene sampling data set is generated;

[0028] Step S2: the education scene sampling data set is calculated according to the content similarity, time similarity and user similarity to obtain a standard education scene type set;

[0029] Step S3: the standard education scene type set is taken as a node to construct an education space-time hypergraph, wherein the hyperedge connects the scene combination with teaching logic association, and a scene feature vector representation is obtained;

[0030] Step S4: a resource optimization model is established according to the scene feature vector representation, the objective function of which is to minimize the product of scene response delay and education priority, and an IPTV scene resource allocation scheme is determined;

[0031] Step S5: network bandwidth and server resources are pre-allocated according to the IPTV scene resource allocation scheme, and resource scheduling is performed when responding to a user scene switching request.

[0032] It can be understood that the execution subject of the present application can be a multi-scene interaction control system in an IPTV education platform, and can also be a terminal or a server, which is not limited here. The embodiment of the present application takes a server as an execution subject for example.

[0033] Specifically, the education behavior Latin hypercube sampling is an improved multi-dimensional data sampling technology, which is optimized for the user behavior characteristics of the IPTV education platform. The IPTV user behavior data is divided into a five-dimensional hypercube space according to the user ID, timestamp, scene type, network state and device type, forming a user behavior space matrix, then each dimension is uniformly divided into k intervals to ensure equal number of sampling points, generating a uniformly distributed sampling grid. The education scene weight factor alpha and the learning progress correlation coefficient beta are the core parameters of the application, wherein alpha is used to distinguish the importance of different education scenes, the classroom live scene weight is set to 1.0 times of alpha, the interactive answering scene weight is set to 0.8 times of alpha, and the resource download scene weight is set to 0.6 times of alpha, forming a scene weight distribution matrix. The weight of each sampling point is calculated by multiplying the scene priority by alpha and adding the product of the progress correlation and beta, generating an education scene sampling data set.

[0034] The EDU-ISODATA clustering algorithm is used to identify the scene type of the sampling data, which realizes accurate clustering through three similarity calculations. In the content similarity calculation, the similarity value of the same knowledge point scene is set to 1.0, the related knowledge point scene is set to 0.7, and the different subject scenes are set to 0.2, forming a content similarity matrix. The time similarity is determined by the ratio of the overlapping part of the time window to the union of the time window, reflecting the user's usage mode of the education scene in different time periods, generating a time similarity matrix. The user similarity quantitatively considers factors such as grade, subject and learning ability, reflecting the behavior characteristics of different user groups, forming a user similarity matrix. Finally, the three similarity matrices are weighted and fused according to the weights 0.5, 0.3 and 0.2, and the positions of the 8 cluster centers are calculated by the iterative clustering algorithm, which correspond to the standardized education scene types of classroom live, interactive answering, resource download, online examination, group discussion, experimental demonstration, course playback and homework submission.

[0035] An education space-time hypergraph structure is constructed, 8 education scenes are set as hypergraph nodes, and hyperedges connect scene combinations with teaching logic association. The advantage of hypergraph over traditional graph structure is that it can represent high-order relationships between multiple nodes. Classroom live, interactive answering and course playback form the first hyperedge to represent the teaching process, and online examination, experimental demonstration and homework submission form the second hyperedge to represent the evaluation and verification process. Education scene mask matrix processing is a key technology, when two scenes have teaching logic association, the mask value is set to 0, and when there is no association, the mask value is set to negative infinity, forming a scene association mask matrix. The hypergraph neural network encodes through multiple layers of graph convolution, the time dimension convolution extracts the scene time sequence features, and the space dimension convolution extracts the scene geographical distribution features, generating a 128-dimensional scene feature vector representation, which contains the content features, time features, user features and association features of the scene.

[0036] A resource optimization model is established, and an education scene priority parameter is set according to a scene feature vector representation. The online examination scene priority is set to 10, the classroom live broadcast scene is set to 8, the interactive answering scene is set to 6, and the resource download scene is set to 4, so as to form an education priority parameter matrix. The priority parameter matrix is multiplied by the scene response delay data to calculate the delay generation value of each education scene, and a scene delay generation cost matrix is generated. The optimization model is constrained to minimize the sum of all scene delay generation values as the objective function, and the constraint condition is set that the total bandwidth is greater than or equal to the product of the bandwidth demand of each scene and the number of students. The education scene-oriented optimization algorithm EDU-NGO searches for the optimal solution space through the initialization of the eagle population, each eagle individual containing the bandwidth allocation proportion and server allocation strategy parameter of 8 education scenes, through the fitness evaluation, position update and elite reservation mechanism, and outputs the IPTV scene resource allocation scheme after 200 iterations.

[0037] According to the IPTV scene resource allocation scheme, the network bandwidth and server resources are pre-allocated. The network bandwidth pre-allocation reserves the bandwidth demand of each education scene to the corresponding network channel according to the priority, and the server resource pre-allocation allocates the computing resources to the edge node and the cloud server according to the scene type. The constructed resource scheduling mapping table records the network channel identifier and the server node identifier corresponding to each education scene. When receiving a user scene switching request, the pre-allocated resources of the target scene are found according to the mapping relationship, and the corresponding network channel and server node are activated through the resource switching instruction.

[0038] In a specific embodiment, step S1 comprises:

[0039] The IPTV user behavior data is divided into an n-dimensional hypercube space according to the user ID, timestamp, scene type, network state and device type, and a user behavior space matrix is obtained.

[0040] Each dimension of the user behavior space matrix is uniformly divided into k intervals to ensure that the number of sampling points in each interval is equal, and a uniformly distributed sampling grid is obtained.

[0041] Based on the education scene weight factor α, different education scenes in the uniformly distributed sampling grid are assigned different weights, wherein the weight of the classroom live broadcast scene is 1.0 times the weight of α, the weight of the interactive answering scene is 0.8 times the weight of α, and the weight of the resource download scene is 0.6 times the weight of α, and a scene weight distribution matrix is obtained.

[0042] The learning progress correlation coefficient β is combined with the scene weight distribution matrix, and the weight of each sampling point is calculated by multiplying the scene priority by α and adding the product of the progress correlation and β, to obtain an education scene sampling data set.

[0043] Specifically, the IPTV user behavior data is spatially divided into five dimensions. The user ID dimension records the unique identifier of each student, the timestamp dimension records the specific time point of user access to the education scene, the scene type dimension identifies the current education scene category used, the network status dimension reflects the current network connection quality and bandwidth of the user, and the device type dimension distinguishes the terminal device used by the user, such as a set-top box, a tablet, or a mobile phone. The five dimensions form an n-dimensional hypercube space, where n equals 5, and each dimension represents an axis. Each user behavior record forms a data point in this five-dimensional space, and the collection of all data points forms a user behavior space matrix.

[0044] When uniformly dividing the user behavior space matrix, each dimension is equally spaced into k intervals, ensuring that the number of sampling points contained in each interval is equal. In the specific division process, the user ID dimension is divided according to the student number order, the timestamp dimension is divided according to the 24-hour time period, the scene type dimension is divided according to the 8 education scene types, the network status dimension is divided according to the bandwidth range from low to high, and the device type dimension is divided according to different terminal devices. The boundary values of each interval are determined by calculating the minimum and maximum values of the dimension data, and then equally dividing to form a uniformly distributed sampling grid. This grid structure ensures that there is no sampling bias due to uneven distribution of data in a certain dimension during the sampling process.

[0045] The differentiated weight assignment of the education scene weight factor α is based on the importance of different scenes in the IPTV education platform. The classroom live scene, as the core teaching activity, has the highest requirements for real-time and interactivity, so the weight is set to 1.0 times α, indicating that this scene enjoys the highest priority in resource allocation. The interactive answering scene requires real-time interaction between students and teachers, but has relatively high tolerance for network delay, so the weight is set to 0.8 times α. The resource download scene mainly involves the acquisition of teaching materials, with the lowest requirement for real-time performance, so the weight is set to 0.6 times α. In the weight assignment process, each sampling point in the uniformly distributed sampling grid is matched with the corresponding weight coefficient according to its value in the scene type dimension, then the weight coefficient is multiplied by α to obtain the weight value of the sampling point in the scene dimension, and the weight values of all sampling points form a scene weight distribution matrix.

[0046] The combination calculation of the learning progress correlation coefficient β and the scenario weight distribution matrix involves a complex weight fusion process. The learning progress correlation reflects the matching degree of the user's current learning state and a specific educational scenario, which is calculated by analyzing the user's historical learning records, current course progress, and knowledge point mastery. In the weight calculation process of each sampling point, the scenario priority value corresponding to the sampling point is extracted, then the scenario priority is multiplied by α to obtain the scenario weight part, then the progress correlation value of the sampling point is extracted, and the progress correlation is multiplied by β to obtain the progress weight part, and finally the scenario weight part and the progress weight part are added to obtain the weight of the sampling point. This weight fusion mechanism ensures that the sampling process considers both the objective importance of the scenario and the personalized learning needs of the user.

[0047] In a specific embodiment, step S2 comprises:

[0048] Content similarity calculation is performed on the scenario data in the educational scenario sampling data set, where the similarity value of the same knowledge point scenario is 1.0, the similarity value of related knowledge point scenarios is 0.7, and the similarity value of different subject scenarios is 0.2, obtaining a content similarity matrix;

[0049] Time similarity is calculated based on the time window information in the educational scenario sampling data set, and the time similarity value is determined by the ratio of the overlapping part of the time window to the union set of the time window, obtaining a time similarity matrix;

[0050] User similarity is calculated based on the user group characteristics in the educational scenario sampling data set, obtaining a user similarity matrix;

[0051] The content similarity matrix, the time similarity matrix, and the user similarity matrix are weighted and fused with weights of 0.5, 0.3, and 0.2, respectively, and the positions of 8 cluster centers are calculated through an iterative clustering algorithm, obtaining a standardized educational scenario type set.

[0052] Specifically, three similarity calculations are implemented to achieve accurate scene type recognition. Content similarity calculation is based on the correlation degree of knowledge points in educational scenes. The knowledge point label information of each scene is extracted from the educational scene sampling dataset, and then a knowledge point correlation database is established, which records the logical relationship between different knowledge points. When two educational scenes contain exactly the same knowledge points, such as mathematical function concept explanation and function exercise, the content similarity value is set to 1.0, indicating the highest similarity. When two scenes involve related knowledge points, such as algebra basics and geometry introduction, which belong to the category of mathematical foundation, the content similarity value is set to 0.7, indicating moderate similarity. When two scenes belong to completely different disciplines, such as physics experiments and Chinese reading, the content similarity value is set to 0.2, indicating low similarity. In the construction process of the content similarity matrix, all scenes are paired, and the similarity value is determined according to the knowledge point correlation degree table to form a symmetric matrix structure, and each element in the matrix represents the content similarity of the corresponding scene pair.

[0053] Time similarity calculation is based on the overlap degree of user's time window of using educational scenes. The time window refers to the time range of user's concentrated use of a certain educational scene in a specific time period. The overlapping part of the time window is calculated by comparing the active time periods of two scenes. In the specific calculation process, the starting and ending use time of each educational scene is determined, and then the intersection part of the two time periods is found as the overlapping time. The union of the time window refers to the total coverage range of the use time period of two scenes, which is calculated by merging the two time periods and removing the overlapping part. The time similarity value is equal to the length of the overlapping time divided by the length of the union time, which reflects the similarity of the two educational scenes in the time use mode. In the generation process of the time similarity matrix, all scene combinations in the educational scene sampling dataset are traversed, the time window overlap ratio of each scene pair is calculated, and the calculation result is filled into the corresponding position of the matrix to form a data structure reflecting the similarity of the time use mode.

[0054] The user similarity calculation comprehensively considers the user group characteristics of multiple dimensions such as grade, subject, and learning ability. The grade characteristics are classified by the education stage of the user, such as user groups in primary school, junior high school, and high school stages. The subject characteristics are divided according to the main course field that the user learns, such as user categories of science inclination, liberal arts inclination, or comprehensive learning. The learning ability characteristics are evaluated by analyzing the historical learning performance and test scores of the user, including learning speed, understanding ability, memory ability, and the like. In the calculation process of user similarity, a weight coefficient is assigned to each user group characteristic, then the difference degree of the two education scene main user groups in each characteristic dimension is calculated, and finally the weighted sum of the differences in each dimension is obtained to obtain the comprehensive similarity value. The user similarity matrix records the user group similarity degree of all scene pairs, and the higher the matrix element value indicates that the target user groups of the two scenes are more similar.

[0055] The weighted fusion process linearly combines the three similarity matrices according to the preset weights, the content similarity weight is set to 0.5, the time similarity weight is set to 0.3, and the user similarity weight is set to 0.2. The weight allocation is determined based on the influence degree of different factors on scene clustering in the IPTV education platform. In the fusion calculation, the matrix elements at corresponding positions are multiplied by the corresponding weights, and then the three products are added to obtain the comprehensive similarity matrix. The iterative clustering algorithm classifies scenes based on the comprehensive similarity matrix. In the algorithm initialization stage, 8 cluster center positions are randomly selected, corresponding to 8 main education scene types such as classroom live broadcast, interactive answering, resource download, online examination, group discussion, experiment demonstration, course playback, and homework submission. In the iteration process, the comprehensive similarity distance of each scene data point to each cluster center is calculated, the data point is assigned to the cluster center with the highest similarity, and then the cluster center position is recalculated according to the new clustering result. The process is repeated until the cluster center position is stable or the maximum iteration number is reached.

[0056] In a specific embodiment, step S3 comprises:

[0057] 8 kinds of education scenes in the standardized education scene type set are set as hypergraph nodes, including classroom live broadcast, interactive answering, resource download, online examination, group discussion, experiment demonstration, course playback, and homework submission nodes, to obtain an education scene node set;

[0058] Based on the education scene node set, a hyperedge connection is constructed, wherein each hyperedge connection has multiple scene nodes with teaching logic association. Classroom live broadcast, interactive answering, and course playback are combined into a first hyperedge, online examination, experiment demonstration, and homework submission are combined into a second hyperedge, to obtain an education space-time hypergraph structure;

[0059] The scene nodes in the education space-time hypergraph structure are processed by an education scene mask matrix, the mask value is set to 0 when two scenes have a teaching logical association, and the mask value is set to negative infinity when there is no association, and a scene association mask matrix is obtained.

[0060] The scene association mask matrix is input into a hypergraph neural network for multi-layer graph convolution coding, the scene time sequence features are extracted through time dimension convolution, and the scene geographical distribution features are extracted through space dimension convolution, and a scene feature vector representation is obtained.

[0061] Specifically, the construction process of the education space-time hypergraph structure in the application is specially designed for the multi-scene association characteristics of the IPTV education platform, and solves the technical problem that the traditional graph structure cannot represent the high-order relationship between multiple scenes. The establishment of the education scene node set is based on the identification of 8 core scene types in the standardized education scene type set, each scene type is abstracted as a node in the hypergraph, the classroom live node represents real-time teaching activities, the interactive answering node represents the interaction between teachers and students, the resource download node represents learning material acquisition, the online examination node represents learning evaluation activities, the group discussion node represents collaborative learning, the experimental demonstration node represents practical teaching activities, the course playback node represents review and consolidation, and the homework submission node represents learning achievement display. In the node setting process, each node is assigned a unique identifier and attribute information, the attribute information includes scene function type, resource demand, user interaction mode and other characteristic data, and the initial connection relationship between nodes is temporarily empty, waiting for the establishment of subsequent hyperedge connection.

[0062] The construction of the hyperedge connection is based on the teaching logical association, the hyperedge is the core concept of the hypergraph structure, unlike the edge in the traditional graph which can only connect two nodes, the hyperedge can connect multiple nodes at the same time, representing the complex multi-element relationship between these nodes. The first hyperedge connects the classroom live, interactive answering and course playback nodes, this connection reflects the teaching process logic, the classroom live provides the main teaching content, the interactive answering verifies the learning effect, and the course playback supports review and consolidation, and the three scenes form a closed loop structure in the teaching process. The second hyperedge connects the online examination, experimental demonstration and homework submission nodes, embodying the logical relationship of learning evaluation and verification, the experimental demonstration shows the practical application of theoretical knowledge, the online examination detects the knowledge mastery degree, and the homework submission records the learning achievement, and the three scenes jointly constitute a learning evaluation system. The weight of the hyperedge is determined by calculating the teaching association strength between the connected nodes, the association strength is comprehensively calculated based on the use frequency, time sequence, content dependency relationship and other factors between the scenes, and the higher the weight value, the closer the teaching logical association between the nodes in the hyperedge.

[0063] The education scene mask matrix processing is a key data preprocessing step in the hypergraph neural network, and the mask matrix controls the information propagation between different scene nodes through the attention mechanism. In the construction process of the mask matrix, an 8x8 matrix structure is established, and the rows and columns of the matrix correspond to the 8 education scene nodes, respectively. Then, according to the teaching logic association relationship between the nodes, the mask value is filled in. When two scene nodes are located in the same hyperedge, it means that they have a direct teaching logic association, and the corresponding mask value is set to 0, which allows the attention mechanism to normally exchange information between the two nodes. When two scene nodes are not in any common hyperedge, it means that they lack direct teaching logic association, and the corresponding mask value is set to negative infinity, which will be converted to a probability close to zero in the softmax calculation, thereby preventing irrelevant scene information interference. The diagonal elements of the mask matrix are all set to 0, indicating the association relationship between each scene node and itself. The scene association mask matrix is a symmetric matrix, reflecting the bidirectional logical association between scenes.

[0064] The multi-layer graph convolutional encoding process of the hypergraph neural network takes the scene association mask matrix as input for feature extraction. Graph convolution is a deep learning technique specifically designed for graph-structured data, which can update the feature representation of the current node by aggregating the information of neighboring nodes. The time dimension convolution is specifically used to extract the time sequence features of the scene. By analyzing the usage patterns, switching frequency, duration, and other time-related attributes of the scene in different time periods, a feature vector reflecting the time behavior of the scene is generated. The spatial dimension convolution is specifically used to extract the geographical distribution features of the scene. By analyzing the preference differences of users in different regions, the influence of network conditions on scene performance, and the requirements of device types on scene adaptation, a feature vector reflecting the spatial distribution of the scene is generated. In the multi-layer convolution processing, the first layer of convolution extracts the basic node features and neighbor information, the second layer of convolution further aggregates the neighbor information in a wider range based on the first layer, and the third layer of convolution forms the global scene association features, outputting a 128-dimensional scene feature vector representation. This vector contains information in four dimensions: content features, time features, spatial features, and association features.

[0065] Taking the construction of a hypergraph of a high school chemistry IPTV education platform as an example, the standardized education scene type set includes four core scenes of chemistry theory explanation, experiment demonstration, exercise, and simulated examination. When the education scene node set is established, the attribute information of the chemistry theory explanation node includes the function of imparting theoretical knowledge, high bandwidth demand, and one-way interaction mode, and the experiment demonstration node includes the function of practical operation demonstration, ultra-high-definition video demand, and observation interaction mode. In the construction of the hyperedge connection, the first hyperedge connects the chemistry theory explanation node and the exercise node, reflecting the logical relationship between theoretical learning and practical application, and the second hyperedge connects the experiment demonstration node and the simulated examination node, embodying the relevance of experiment observation and ability detection. When the education scene mask matrix is processed, the chemistry theory explanation and the exercise are located in the same hyperedge, and the mask value is set to 0, while the chemistry theory explanation and the simulated examination are not in the same hyperedge, and the mask value is set to negative infinity. In the convolution coding process of the hypergraph neural network, the time dimension convolution finds that the chemistry theory explanation is mainly used in the morning classroom time, and the experiment demonstration is concentrated in the afternoon experiment class time. The spatial dimension convolution analysis finds that there are differences in network demand between the experiment demonstration scene in the laboratory environment and the ordinary classroom environment. The generated scene feature vector accurately reflects the multi-dimensional characteristics and complex correlation of the chemistry education scene, solving the technical problem that the traditional method cannot process high-order dependency relationship between multiple scenes.

[0066] In a specific embodiment, step S4 comprises:

[0067] Based on the scene feature vector representation, an education scene priority parameter is set, wherein the online examination scene priority is set to 10, the classroom live scene priority is set to 8, the interactive answer question scene priority is set to 6, and the resource download scene priority is set to 4, to obtain an education priority parameter matrix;

[0068] The education priority parameter matrix is multiplied by the scene response delay data to calculate the delay generation value of each education scene, and a scene delay generation cost matrix is obtained;

[0069] According to the scene delay generation cost matrix, a resource optimization objective function is constructed, the optimization objective is to minimize the sum of all scene delay generation values, and a constraint condition is set that the total bandwidth is greater than or equal to the sum of the product of the bandwidth demand of each scene and the number of students, to obtain a constraint optimization model;

[0070] The constraint optimization model is input into an education scene oriented optimization algorithm for iterative solution, the optimal solution space is searched through population evolution and elite reservation strategy, and an IPTV scene resource allocation scheme is obtained.

[0071] Specifically, the setting of the education scenario priority parameter is quantified based on the importance indicators extracted in the scenario feature vector representation, and the priority value reflects the key degree of different education scenarios in the teaching process and the urgent demand for network resources. The online examination scenario priority is set to 10, and this value is determined based on the real-time requirement of the examination and the non-repeatability feature. Any delay or interruption in the examination process will directly affect the student's score evaluation, so the highest level of resource guarantee is required. The classroom live broadcast scenario priority is set to 8, which takes into account the real-time demand of live broadcast and the continuity requirement of teacher-student interaction. Delay in live broadcast will affect the teaching effect but has a certain fault tolerance compared to examination. The interactive answering scenario priority is set to 6, and although the answering activity requires real-time response, the impact of single delay is relatively small, and the students have time to wait for network recovery. The resource download scenario priority is set to 4, and the download activity has the lowest real-time requirement, and the delay mainly affects the user experience without affecting the core function of teaching. In the construction process of the education priority parameter matrix, the priority values of the 8 education scenarios are arranged in a one-dimensional vector in the order of the scenarios, and each element in the vector corresponds to the priority weight of a scenario.

[0072] The scenario response delay data is derived from the real-time monitoring module of the IPTV education platform, which continuously records the time interval from the user's request to the successful loading of each education scenario. The delay data includes network transmission delay, server processing delay, content loading delay and other components, and the average response delay value of each scenario is obtained through historical data statistical analysis. In the multiplication operation process of the education priority parameter matrix and the scenario response delay data, each element in the priority parameter vector is multiplied by the delay data of the corresponding scenario, and the calculation formula is delay value = scenario priority * scenario response delay. The product reflects the negative impact of the delay of this scenario on the overall teaching quality. The higher the delay value, the more serious the delay problem of the scenario, and more resources need to be invested for optimization. The formation of the scenario delay cost matrix organizes the delay values of all scenarios into a matrix form, and each element in the matrix represents the delay cost value of the corresponding scenario. This matrix serves as the core data input of the resource optimization objective function.

[0073] The construction of the constraint optimization model is based on the establishment of a mathematical optimization framework of the scene delay cost matrix. The objective function is set to minimize the total value of all scene delay costs, which is obtained by adding all elements in the scene delay cost matrix. The task of the optimization algorithm is to find a resource allocation strategy that can minimize the total value. The setting of the constraint condition ensures the feasibility and rationality of the resource allocation scheme. The bandwidth constraint condition requires that the total bandwidth be greater than or equal to the sum of the product of the bandwidth demand of each scene and the number of students. This constraint calculates the total bandwidth demand of each scene by counting the bandwidth demand of each user and the number of students currently using the scene, and then adds the bandwidth demands of all scenes to obtain the total demand of the system, ensuring that the allocated bandwidth resources can meet the actual use demand. The server load constraint condition limits the CPU usage rate to be less than a preset threshold, preventing server overload and causing system performance degradation. The constraint optimization model combines the objective function and the constraint condition into a mathematical model, providing a clear solution framework for the subsequent optimization algorithm.

[0074] The education scene-oriented optimization algorithm EDU-NGO is an intelligent optimization technology based on the improvement of the Northern Goshawk optimization algorithm, which is specially designed for the resource allocation characteristics of the IPTV education platform. The population evolution process of the algorithm simulates the hunting behavior of the goshawk. Each goshawk individual represents a possible resource allocation scheme, and the individual's gene code contains the bandwidth allocation proportion of 8 education scenes and the server allocation strategy parameters. In the initialization stage, an initial population containing multiple goshawk individuals is randomly generated. The gene values of each individual are randomly set within a reasonable range to ensure the diversity of the initial solution. In the fitness evaluation process, each individual is substituted into the constraint optimization model for calculation. The individual's fitness level is determined by the comprehensive score of the objective function value and the constraint violation degree. The higher the fitness, the better the resource allocation scheme. The position update operation guides the moving direction of other individuals based on the position information of the current best individual. The update formula is new position equals to current position plus the product of random factor and the difference between optimal position. The random factor controls the size of the update step, ensuring that the algorithm has both exploration ability and convergence. The elite reservation strategy selects the top 20% individuals with the highest fitness after each iteration to directly enter the next generation population, avoiding the loss of excellent solutions in the evolution process, and providing a learning target for the remaining individuals.

[0075] In a specific embodiment, the execution step of inputting the constraint optimization model into the education scene-oriented optimization algorithm for iterative solution can specifically include the following steps:

[0076] Initialize the goshawk population as the fitness function of the constraint optimization model, where each goshawk individual contains the bandwidth allocation proportion of 8 education scenes and the server allocation strategy parameters, obtaining the initial optimization population;

[0077] The fitness of each hawk individual in the initial optimization population is evaluated, and the individual fitness value is determined by calculating the weighted combination of the total sum of the scenario delay cost matrix and the constraint violation degree, and the population fitness distribution is obtained;

[0078] The hawk position updating operation is performed based on the population fitness distribution, the new position coordinates are calculated by adding the current position to the product of the random factor and the difference between the optimal position, and the key education scenarios such as online examination are automatically increased in resource allocation weight, and the updated population is obtained;

[0079] The elite reservation mechanism is applied to the updated population, the top 20% individuals with the highest fitness value are selected as the elite solution and reserved to the next generation population, and the global optimal solution is output after 200 iterations, and the IPTV scene resource allocation scheme is obtained.

[0080] Specifically, the initialization process of the hawk population takes the constraint optimization model as the fitness evaluation standard, each hawk individual represents a resource allocation scheme, the gene code of the individual includes the bandwidth allocation proportion of 8 education scenes and the server allocation strategy parameters, the bandwidth allocation proportion data structure is an 8-dimensional vector, each element in the vector represents the occupation proportion of the corresponding education scene in the total bandwidth, and the sum of all elements is equal to 1 to ensure the integrity of bandwidth allocation. The server allocation strategy parameters include the load distribution of each scene on different server nodes, and the deployment position of the scene is controlled by the load proportion of the edge server and the cloud server. In the generation process of the initial optimization population, the random number generator assigns values to the gene parameters of each individual within a reasonable range, the random range of the bandwidth allocation proportion is determined according to the historical use frequency of each education scene, the upper limit of the initial allocation proportion of the high-frequency use scene is relatively high, and the allocation proportion upper limit of the low-frequency use scene is relatively low, which ensures the rationality and feasibility of the initial solution.

[0081] The fitness evaluation process comprehensively evaluates each hawk individual in the initial optimization population, and the calculation of the fitness function is designed based on the weighted combination of the sum of the scenario delay cost matrix and the constraint violation degree. The sum of the scenario delay cost matrix is obtained by substituting the individual's resource allocation scheme into the delay calculation model. In the specific calculation process, the actual bandwidth obtained by each education scenario is calculated according to the individual's bandwidth allocation proportion, and then the expected response delay of the scenario is estimated based on the inverse proportional relationship between bandwidth and delay. The expected delay is multiplied by the priority of the scenario to obtain the delay value of the scenario, and finally the total delay cost is obtained by adding up the delay values of all scenarios. The calculation of the constraint violation degree involves the checking of bandwidth constraints and server load constraints. When the individual's resource allocation scheme causes the bandwidth demand of some scenarios to be unable to meet, the violation degree is quantified according to the size of the gap, and a violation penalty is also generated when the server load exceeds the limit. The fitness value is calculated by the difference between the reciprocal of the total delay cost and the constraint violation penalty. The higher the fitness value, the better the individual's resource allocation scheme. The population fitness distribution records the fitness ranking and statistical information of all individuals.

[0082] The hawk position updating operation is guided based on the optimal individual position in the population fitness distribution, and the calculation process of the position updating simulates the behavior pattern of the hawk approaching the prey. The current position data is derived from the individual's gene encoding value in the last iteration, the optimal position data is derived from the gene encoding value of the individual with the highest fitness in the population, and the random factor is generated by the pseudo-random number generator between zero and one, controlling the step size of the position updating. The calculation of the new position coordinates adds the product of the current position and the random factor and the difference between the optimal position to the current position. This calculation process ensures that the individual can move towards the optimal solution while maintaining a certain randomness to avoid premature convergence to a local optimal solution. The key education scenario detection mechanism plays a dynamic adjustment role in the position updating process. When the online examination scenario is detected to be in progress, the algorithm automatically increases the resource allocation weight of this scenario. The specific implementation is to add a fixed increment to the bandwidth allocation proportion of the online examination scenario in the position updating calculation, while proportionally reducing the allocation proportion of other non-key scenarios, ensuring that the key scenario obtains sufficient resource guarantee. The updated population contains the new gene encoding of all individuals after position updating, preparing data for the next round of fitness evaluation.

[0083] The elite reservation mechanism selects and saves excellent individuals in the updated population, and the selection criteria of the elite individuals are determined based on the ranking result of the fitness value. The algorithm ranks all individuals in the updated population in descending order of fitness value, and then selects the top 20% individuals as elite solutions. The number of elite individuals is equal to the total number of populations multiplied by the reservation rate and rounded up. The implementation process of elite reservation directly copies the selected excellent individuals to the next generation population without further mutation or crossover operation, ensuring that excellent solutions are not lost in the evolution process. The remaining population positions are filled by genetic operations such as selection, crossover, and mutation on other individuals in the current population to form the next generation population. The maximum number of iterations is set to 200 by the iteration control mechanism, and each iteration includes a complete operation process such as fitness evaluation, position update, and elite reservation. The iteration termination conditions include reaching the maximum number of iterations or the optimal solution of the population in consecutive generations has no significant improvement. The output of the global optimal solution is selected from the iteration results, and the individual with the highest fitness value is the IPTV scene resource allocation scheme, which includes the optimal bandwidth allocation ratio and server deployment strategy for each education scene.

[0084] In a specific embodiment, step S5 comprises:

[0085] Based on the IPTV scene resource allocation scheme, the network bandwidth is pre-allocated, and the bandwidth requirements of each education scene are reserved in the corresponding network channel according to the priority to obtain a bandwidth pre-allocation configuration;

[0086] According to the IPTV scene resource allocation scheme, the server resources are pre-allocated, and the computing resources are allocated to the edge nodes and cloud servers according to the scene type to obtain a server pre-allocation configuration;

[0087] The bandwidth pre-allocation configuration and the server pre-allocation configuration are combined to construct a resource scheduling mapping table, which records the network channel identifier and the server node identifier corresponding to each education scene to obtain a scene resource mapping relationship;

[0088] When receiving a user scene switching request, the pre-allocated resources of the target scene are found according to the scene resource mapping relationship, and the corresponding network channel and server node are activated through a resource switching instruction to obtain a scene switching execution result.

[0089] Specifically, the pre-allocation of network bandwidth is implemented based on the bandwidth allocation ratio in the IPTV scenario resource allocation scheme, and the hierarchical reservation mechanism divides the network bandwidth into different priority channels according to the priority parameters of the education scenarios. The data processing process of bandwidth pre-allocation extracts the bandwidth proportion data of each education scenario in the resource allocation scheme, then calculates the specific bandwidth allocation of each scenario according to the proportion, and then allocates the bandwidth to the corresponding network channel according to the priority of the scenario. High-priority scenarios such as online exams are allocated to the first priority channel, medium-priority scenarios such as classroom live broadcasts are allocated to the second priority channel, and low-priority scenarios such as resource downloads are allocated to the third priority channel. The identification of the network channel adopts the combination of channel ID and priority label, each channel is allocated a unique digital identifier and priority code, and the bandwidth isolation and priority scheduling between channels are realized through the QoS mechanism, ensuring that the bandwidth demand of high-priority scenarios is satisfied first. The data structure of bandwidth pre-allocation configuration contains key information such as scenario identification, channel identification, allocated bandwidth, priority level, etc. The configuration data is stored in table form and updated in real time.

[0090] The pre-allocation of server resources calculates the distributed deployment of computing resources according to the server allocation strategy in the IPTV scenario resource allocation scheme, and the selection of edge nodes and cloud servers is based on the computing demand characteristics of scenario types and the geographical distribution of users. The data processing process of computing resource allocation involves quantitative allocation of multi-dimensional resources such as CPU core number, memory capacity, and storage space, and each education scenario determines the required computing resource size according to its computing complexity and concurrent user number. Edge nodes mainly undertake delay-sensitive scenarios such as classroom live broadcasts and interactive question answering, which require local computing to reduce network transmission delay, and cloud servers mainly undertake computing-intensive scenarios such as large file download and data analysis, making full use of the powerful computing capacity of the cloud. The decision algorithm of server pre-allocation calculates the optimal server deployment scheme according to factors such as the geographical distribution of scenarios, user access patterns, and network topology structure, and allocates the computing tasks of the same scenario to multiple server nodes through a load balancing algorithm to avoid single-point overload. The data structure of server pre-allocation configuration records detailed information such as scenario identification, server node identification, allocated computing resource quantity, and load distribution ratio, and the configuration data is stored and synchronized through a distributed database.

[0091] The construction process of the resource scheduling mapping table fuses the bandwidth pre-allocation configuration and the server pre-allocation configuration, and the mapping table adopts a multi-level index structure to quickly locate the resource configuration information of a specific scene. The data processing logic of the mapping table establishes an index structure with a scene identifier as a primary key, then stores the network channel identifier and the server node identifier corresponding to each scene as associated data, and records detailed configuration parameters of resources such as bandwidth capacity, CPU core number, and memory size. The mapping relationship is established by using a hash algorithm to achieve fast lookup, the scene identifier is calculated by using a hash function to obtain a storage address, and the corresponding resource configuration information is directly accessed, and the time complexity of the lookup is constant. The data synchronization mechanism of the mapping table ensures that the mapping information in multiple network nodes and server clusters is consistent, and when the resource configuration is changed, the update information is broadcast to all related nodes through a message queue mechanism. The data structure of the scene resource mapping relationship adopts a key-value pair form, the key is a scene identifier, and the value is a composite data structure containing network channel information and server node information, which supports fast reading and updating operations.

[0092] The data processing process of the scene switching execution responds to a scene switching request of a user, and quickly activates a network channel and a server node through a resource switching instruction. The scene switching request of the user includes basic information such as a user identifier, a source scene identifier, a target scene identifier, and a switching timestamp, a request processing module verifies the user authority and the legality of the scene switching, then finds corresponding pre-allocated resource information in the scene resource mapping relationship according to the target scene identifier. The resource finding process directly locates the resource configuration record of the target scene through the hash index of the mapping table, extracts key information such as a network channel identifier and a server node identifier, and checks the current state and availability of the resource. The generation of the resource switching instruction includes an instruction for activating a network channel and an instruction for starting a server node, the network channel activation instruction includes configuration information such as a channel identifier, a QoS parameter, and a flow control strategy, and the server node starting instruction includes detailed parameters such as a node identifier, a computing resource allocation, and a storage resource allocation. The instruction execution process realizes automatic operation through a network management protocol and a server management interface, the activation of the network channel realizes flow routing and QoS guarantee through a flow table rule issued by an SDN controller, and the starting of the server node quickly deploys a computing environment required by a scene through containerization technology. The scene switching execution result includes feedback information such as a switching state, a switching time consumption, and a resource allocation state, and the result data is used for monitoring switching performance and optimizing subsequent resource allocation strategies.

[0093] The above describes the method for multi-scene interaction control in the IPTV education platform in the embodiments of the present application, and the following describes the system for multi-scene interaction control in the IPTV education platform in the embodiments of the present application. Please refer to Figure 2 An embodiment of the system for multi-scene interaction control in the IPTV education platform in the embodiments of the present application includes:

[0094] a sampling module, configured to sample IPTV user behavior data according to an educational behavior Latin hypercube sampling, to generate an educational scene sampling dataset by layer sampling according to an educational scene weight factor and a learning progress correlation coefficient;

[0095] a calculation module, configured to calculate a scene clustering center according to content similarity, time similarity and user similarity, to obtain a standardized educational scene type set;

[0096] a construction module, configured to construct an educational space-time hypergraph by taking the standardized educational scene type set as a node, to obtain a scene feature vector representation, wherein a hyperedge is connected to a scene combination having a teaching logic correlation;

[0097] an establishment module, configured to establish a resource optimization model with a target function being a minimum product of a scene response delay and an educational priority, to determine an IPTV scene resource allocation scheme;

[0098] a response module, configured to pre-allocate network bandwidth and server resources according to the IPTV scene resource allocation scheme, and to perform resource scheduling when responding to a user scene switching request.

[0099] The above Figure 2 The multi-scene interaction control system in the IPTV educational platform in the embodiment of the application is described in detail from the perspective of a modular functional entity, and the multi-scene interaction control device in the IPTV educational platform in the embodiment of the application is described in detail from the perspective of hardware processing.

[0100] Referring to Figure 3 , the embodiment of the application further provides a multi-scene interaction control device in an IPTV educational platform. The multi-scene interaction control device in the IPTV educational platform can be a server, and the internal structure thereof can be as shown in Figure 3 . The multi-scene interaction control device in the IPTV educational platform includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. The processor of the computer is used to provide computing and control capabilities. The memory of the multi-scene interaction control device in the IPTV educational platform includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the multi-scene interaction control device in the IPTV educational platform is used to store the corresponding data in the embodiment. The network interface of the multi-scene interaction control device in the IPTV educational platform is used to communicate with external terminals through network connection. The computer program is executed by the processor to implement the above method.

[0101] Those skilled in the art can understand, Figure 3 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the multi-scene interaction control device in the IPTV education platform to which the scheme of the present application is applied.

[0102] The present application also provides a computer readable storage medium, which can be a non-volatile computer readable storage medium, or a volatile computer readable storage medium, and the computer readable storage medium stores instructions, and when the instructions are run on a computer, the computer executes the steps of the multi-scene interaction control method in the IPTV education platform.

[0103] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, system and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0104] The integrated unit, if realized in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical scheme of the present application or the part of the prior art that essentially contributes or the whole or part of the technical scheme can be embodied in the form of a software product, which is stored in a storage medium and includes a plurality of instructions for causing an IPTV education platform multi-scene interaction control device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The foregoing storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program code storage media.

[0105] The above embodiments are only used to illustrate the technical scheme of the present application, rather than limit it; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical scheme recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical scheme deviate from the spirit and scope of the technical scheme of each embodiment of the present application.

Claims

1. A method for multi-scene interaction control in an IPTV education platform, characterized in that, The method comprises: Step S1: using education behavior Latin hypercube sampling, sampling IPTV user behavior data according to education scene weight factor and learning progress correlation coefficient to generate an education scene sampling data set, comprising: dividing IPTV user behavior data into an n-dimensional hypercube space according to user ID, timestamp, scene type, network state, and device type to obtain a user behavior space matrix; uniformly dividing each dimension of the user behavior space matrix into k intervals to ensure that the number of sampling points in each interval is equal, thereby obtaining a uniformly distributed sampling grid; assigning different weights to different education scenes in the uniformly distributed sampling grid based on an education scene weight factor α, wherein the weight of a classroom live scene is 1.0 times α, the weight of an interactive answering scene is 0.8 times α, and the weight of a resource downloading scene is 0.6 times α, thereby obtaining a scene weight distribution matrix; combining a learning progress correlation coefficient β with the scene weight distribution matrix to calculate the weight of each sampling point by multiplying the scene priority by α and adding the product of the progress correlation and β, thereby obtaining the education scene sampling data set; Step S2: calculating a scene clustering center according to content similarity, time similarity, and user similarity of the education scene sampling data set to obtain a standardized education scene type set; Step S3: constructing an education space-time hypergraph with the standardized education scene type set as nodes, wherein a hyperedge connects a combination of scenes having a teaching logic association, and a scene feature vector representation is obtained; Step S4: establishing a resource optimization model with a target function of minimizing the product of scene response delay and education priority based on the scene feature vector representation to determine an IPTV scene resource allocation scheme; Step S5: pre-allocating network bandwidth and server resources according to the IPTV scene resource allocation scheme, and performing resource scheduling when responding to a user scene switching request.

2. The method of claim 1, wherein the IPTV education platform is characterized in that, The step S2 comprises: calculating content similarity of scene data in the education scene sampling data set, wherein the similarity value of a same knowledge point scene is 1.0, the similarity value of related knowledge point scenes is 0.7, and the similarity value of different subject scenes is 0.2, thereby obtaining a content similarity matrix; calculating time similarity based on time window information in the education scene sampling data set, determining a time similarity value through the ratio of the overlapping part of the time window to the union set of the time window, thereby obtaining a time similarity matrix; calculating user similarity according to user group characteristics in the education scene sampling data set, thereby obtaining a user similarity matrix; weighting and fusing the content similarity matrix, the time similarity matrix, and the user similarity matrix according to weights of 0.5, 0.3, and 0.2, calculating 8 clustering center positions through an iterative clustering algorithm, and obtaining a standardized education scene type set.

3. The method of claim 1, wherein the IPTV education platform is characterized by, The step S3 comprises: setting 8 education scenes in the standardized education scene type set as hypergraph nodes, including a classroom live node, an interactive answering node, a resource downloading node, an online examination node, a group discussion node, an experimental demonstration node, a course playback node, and a homework submission node, thereby obtaining an education scene node set; constructing hyper-edges based on the set of education scene nodes, wherein each hyper-edge has a plurality of scene nodes with teaching logical association, and the first hyper-edge is composed of live class, interactive answering and course playback, and the second hyper-edge is composed of online examination, experiment demonstration and homework submission, so as to obtain an education space-time hypergraph structure; performing education scene mask matrix processing on the scene nodes in the education space-time hypergraph structure, wherein the mask value is set to 0 when two scenes have teaching logical association, and the mask value is set to negative infinity when there is no association, so as to obtain a scene association mask matrix; inputting the scene association mask matrix into a hypergraph neural network for multi-layer graph convolution coding, extracting scene time sequence features through time dimension convolution and extracting scene geographical distribution features through space dimension convolution, so as to obtain a scene feature vector representation.

4. The method of claim 1, wherein the IPTV education platform is characterized by, The step S4 comprises: setting an education scene priority parameter based on the scene feature vector representation, wherein the online examination scene priority is set to 10, the live class scene priority is set to 8, the interactive answering scene priority is set to 6, and the resource download scene priority is set to 4, so as to obtain an education priority parameter matrix; performing product operation on the education priority parameter matrix and scene response delay data to calculate the delay generation value of each education scene, so as to obtain a scene delay generation cost matrix; constructing a resource optimization objective function based on the scene delay generation cost matrix, wherein the optimization objective is to minimize the sum of all scene delay generation values, and a constraint condition is set that the total bandwidth is greater than or equal to the product sum of the bandwidth demand of each scene and the number of students, so as to obtain a constraint optimization model; inputting the constraint optimization model into an education scene oriented optimization algorithm for iterative solution, searching for the optimal solution space through population evolution and elite reservation strategy, and obtaining an IPTV scene resource allocation scheme.

5. The method of claim 4, wherein the IPTV education platform is characterized in that, The constraint optimization model is input into an education scene oriented optimization algorithm for iterative solution, the optimal solution space is searched through population evolution and elite reservation strategy, and an IPTV scene resource allocation scheme is obtained. The constraint optimization model is input into an education scene oriented optimization algorithm for iterative solution, the optimal solution space is searched through population evolution and elite reservation strategy, and an IPTV scene resource allocation scheme is obtained. The constraint optimization model is input into an education scene oriented optimization algorithm for iterative solution, the optimal solution space is searched through population evolution and elite reservation strategy, and an IPTV scene resource allocation scheme is obtained. The constraint optimization model is input into an education scene oriented optimization algorithm for iterative solution, the optimal solution space is searched through population evolution and elite reservation strategy, and an IPTV scene resource allocation scheme is obtained. The constraint optimization model is input into an education scene oriented optimization algorithm for iterative solution, the optimal solution space is searched through population evolution and elite reservation strategy, and an IPTV scene resource allocation scheme is obtained.

6. The method of claim 1, wherein the IPTV education platform is characterized by, The step S5 comprises: Pre-allocate network bandwidth according to the IPTV scene resource allocation scheme, reserve bandwidth demand of each education scene to corresponding network channel according to priority, obtain bandwidth pre-allocation configuration; Pre-allocate server resource according to the IPTV scene resource allocation scheme, allocate computing resource to edge node and cloud server according to scene type, obtain server pre-allocation configuration; Combine the bandwidth pre-allocation configuration and the server pre-allocation configuration to build a resource scheduling mapping table, which records network channel identifier and server node identifier corresponding to each education scene, obtain scene resource mapping relationship; When receiving a user scene switching request, find pre-allocated resource of the target scene according to the scene resource mapping relationship, activate corresponding network channel and server node through resource switching instruction, obtain scene switching execution result.

7. A multi-scene interactive control system in an IPTV education platform, characterized in that, The IPTV education platform multi-scene interaction control method is implemented by using the IPTV education platform multi-scene interaction control system, which comprises: A sampling module is configured to sample IPTV user behavior data according to education scene weight factor and learning progress correlation coefficient by using education behavior Latin hypercube sampling, and generate education scene sampling data set; A computing module is configured to calculate scene clustering center according to content similarity, time similarity and user similarity based on the education scene sampling data set, and obtain standardized education scene type set; A building module is configured to build education space-time hypergraph by taking the standardized education scene type set as node, and obtain scene feature vector representation by connecting scene combination with teaching logic association through hyperedge; An establishing module is configured to establish resource optimization model with target function being minimization of product of scene response delay and education priority according to the scene feature vector representation, and determine IPTV scene resource allocation scheme; A response module is configured to pre-allocate network bandwidth and server resource according to the IPTV scene resource allocation scheme, and execute resource scheduling when responding to user scene switching request.

8. A multi-scene interactive control device in an IPTV education platform, characterized in that, The IPTV education platform multi-scene interaction control method is implemented by using the IPTV education platform multi-scene interaction control system, which comprises:

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program, when executed by the processor, causes the processor to perform the IPTV education platform multi-scene interaction control method. The computer program, when executed by the processor, causes the processor to perform the IPTV education platform multi-scene interaction control method.

Images