A cloud computing-based classroom teaching data management method and system

Abstract

This invention belongs to the field of artificial intelligence, specifically providing a cloud-based classroom teaching data management method and system. The method includes: acquiring classroom teaching data from multiple heterogeneous data sources within the classroom; preprocessing the acquired data and transmitting it to a cloud system composed of edge cloud nodes and central cloud nodes for further processing; the processing includes: issuing early warnings for teaching anomalies based on preset rules, and extracting features from the teaching data to obtain feature data; further including in-depth analysis of the feature data; and providing teaching assistance applications to at least one of the teacher, student, and academic management terminals based on the results of the in-depth analysis and / or the anonymous teaching dataset. This invention ensures immediate feedback and rapid processing of dynamic classroom data, adapts to the real-time intervention needs of classroom teaching, and generates personalized suggestions that fit teachers' teaching habits and the learning situation of the class.

Description

Technical Field

[0001] This invention belongs to the field of artificial intelligence technology, and in particular relates to a cloud computing-based method and system for classroom teaching data management. Background Technology

[0002] To overcome the limitations of traditional manual recording and decentralized management of teaching data, cloud computing-based teaching data management technology has gradually become a research hotspot in the education field. The idea is to rely on cloud computing power to realize centralized storage, analysis and sharing of teaching data, replacing the traditional local server architecture, which to some extent solves the problems of insufficient teaching data storage capacity and inconvenient access across scenarios.

[0003] Currently, existing cloud-based classroom teaching data management technologies have formed preliminary application schemes, mainly building a system around three major aspects: data collection, cloud processing, and teaching applications. However, existing technologies have significant shortcomings in cloud collaborative architecture, teaching-specific algorithms, and data processing mechanisms, resulting in insufficient real-time performance, accuracy, and efficiency in teaching data management, and failing to fully leverage the empowering value of data in classroom teaching. Summary of the Invention

[0004] The purpose of this invention is to provide a cloud computing-based classroom teaching data management method and system to at least solve the technical problems mentioned in the background section.

[0005] To achieve the above objectives, the present invention provides the following technical solution.

[0006] According to an embodiment of the present invention, a cloud computing-based classroom teaching data management method is provided, comprising the following steps:

[0007] Step S1: Obtain classroom teaching data from multiple heterogeneous data sources within the classroom;

[0008] Step S2: Preprocess the acquired data and transmit it to a cloud system consisting of edge cloud nodes and central cloud nodes for further processing; wherein:

[0009] The edge cloud nodes are deployed locally on campus and are used to process teaching data in real time. The real-time processing includes: issuing early warnings of teaching anomalies based on preset rules, and extracting features from the teaching data to obtain feature data.

[0010] The central cloud node is used to perform in-depth analysis on the feature data uploaded by the edge cloud node. The in-depth analysis includes at least: generating student learning status analysis results based on the student dynamic learning profile algorithm, and generating teaching adjustment suggestions based on the teacher dynamic recommendation model.

[0011] The results of the deep analysis are made secure and anonymized to generate a reusable anonymous teaching dataset;

[0012] Step S3: Based on the results of the deep analysis and / or the anonymized teaching dataset, provide teaching assistance applications to at least one of the teacher's end, student's end, and academic affairs management end.

[0013] Furthermore, in step S1, the multiple heterogeneous data sources include classroom smart terminals, audio and video acquisition devices, environmental sensing devices, and the academic affairs management system.

[0014] Furthermore, in step S2, the preset rules include scene threshold rules and behavior feature association rules;

[0015] The scenario threshold rules set quantitative thresholds corresponding to the classroom environment and individual student status.

[0016] The behavioral feature association rules are used to analyze at least two teaching data features to trigger an early warning.

[0017] After the warning is triggered, the edge cloud node synchronously pushes a warning prompt to the teacher's end, which includes the warning type, the objects involved, and the duration of the abnormality, and retains the warning log and the corresponding original data.

[0018] Furthermore, in step S2, the step of extracting features from the teaching data to obtain feature data includes:

[0019] A lightweight neural network model is used to extract hierarchical features from the verified teaching data, including text features of classroom interaction statements, student behavior features, audio and speech features, and basic features of knowledge point relevance. Through the teaching semantic fusion unit, the basic features are semantically associated with the preset subject knowledge graph, and redundant features that are irrelevant to the teaching objectives are filtered out.

[0020] Weights are assigned to each feature based on its importance in teaching, resulting in feature data containing feature type annotations, weight labels, and teaching semantic labels.

[0021] Furthermore, the lightweight neural network model includes a multimodal input module, a depthwise separable convolution module, an instructional semantic attention module, and a feature fusion output module connected in sequence.

[0022] in:

[0023] The multimodal input module is used to receive and process text, behavioral image, and audio teaching data in parallel to extract initial features of each modality.

[0024] The depthwise separable convolution module uses a depthwise separable convolution structure to perform lightweight convolution processing on the initial features;

[0025] The teaching semantic attention module is used to assign attention weights to the features after lightweight convolution processing based on a prior knowledge graph that is dynamically associated with the current teaching scenario.

[0026] The feature fusion output module is used to fuse and filter the weighted multimodal features and output the feature data for transmission.

[0027] Furthermore, in the instructional semantic attention module, the prior knowledge graph is associated with at least subject knowledge points, teaching skill points, and classroom behavior patterns. The allocation of attention weights is used to strengthen features related to the current teaching objectives and generate interpretable weight allocation data.

[0028] Furthermore, in step S2, the step of generating student learning status analysis results based on the student dynamic learning profile algorithm includes:

[0029] A dynamic learning profile algorithm for students is constructed using a fusion architecture of cognitive diagnostic model and long short-term memory network. The algorithm takes feature data uploaded from edge cloud as input and combines it with a pre-set subject knowledge graph to construct a three-dimensional dynamic image of students’ knowledge mastery, ability dimension and learning behavior model.

[0030] The knowledge point association features, classroom interaction semantic features, and attention temporal features in the feature data are mapped to a high-dimensional input vector, and the contribution weight of each feature is calculated, represented as follows:

[0031]

[0032] In the formula, Represents the normalized weight of the i-th feature. Let K represent the feature vector of the i-th class, and K represent the set of concept nodes in the subject knowledge graph. Represents the semantic similarity function. This represents the teaching scenario adaptation factor, where m represents the total number of feature categories.

[0033] The LSTM network captures the temporal evolution of features, updates the 3D profile incrementally with a preset time window, and outputs a quantitative value of knowledge point mastery based on a cognitive diagnostic model. A comprehensive learning state index, used to characterize the analysis results of student learning status, is calculated and expressed as:

[0034]

[0035] In the formula, Indicates a composite index. This represents the weighting coefficients for each dimension of the profile. This represents the scores for each dimension. The score indicates the degree of mastery of the knowledge points. This indicates the ability-fit score. This indicates the score for behavioral pattern assessment.

[0036] Furthermore, in step S2, the step of generating teaching adjustment suggestions based on the teacher dynamic recommendation model includes:

[0037] The dynamic recommendation model for teachers is generated by taking the results of student learning status analysis, teachers' teaching history data, subject knowledge graphs and edge cloud feature data as inputs to generate teaching adjustment suggestions.

[0038] From historical teaching data, teacher interaction frequency, knowledge point breakdown methods, teaching pace, and resource preferences were extracted. Using the K-means clustering algorithm, these were categorized into three style clusters: logical breakdown type, interactive guidance type, and resource-dependent type. The similarity coefficient between the target teacher and their respective style cluster was calculated, and expressed as:

[0039]

[0040] In the formula, The style similarity coefficient. The weight of the style feature of class p, Let p be the feature vector of the target teacher. Let p be the feature mean vector of the p-th class belonging to the style cluster. The cosine similarity function;

[0041] Based on the analysis of students' learning status, common knowledge weaknesses in the class, individual ability differences, and classroom attention distribution characteristics are identified. Combining this with the knowledge point correlation of the subject knowledge graph, the priority score of each potential adjustment plan is calculated, as follows:

[0042]

[0043] In the formula, P represents the priority score. To assign weights to the learning situation, M represents the degree of matching between the plan and the class's learning situation. C represents the implementation cost coefficient, and C represents the implementation complexity of the solution.

[0044] Schemes that meet the priority score requirements are selected, and teaching adjustment suggestions are generated based on the teacher's style.

[0045] Furthermore, the teacher dynamic recommendation model includes:

[0046] The multi-source input fusion layer is used to process the input data and obtain the fusion feature matrix. The input data includes student learning status analysis result vector, teacher teaching history style feature vector, subject knowledge graph association matrix and classroom real-time interactive feedback feature vector.

[0047] The teacher style feature processing layer includes a built-in K-means clustering submodule and a style adaptation submodule. The clustering submodule is used to divide teacher styles into three clusters: logical decomposition type, interactive guidance type, and resource-dependent type. It calculates the feature fit between the target teacher and its cluster using cosine similarity. The style adaptation submodule dynamically adjusts the feature weights based on the fit.

[0048] The learning situation adaptation decision layer includes a learning situation matching sub-mechanism and a collaborative filtering sub-mechanism. In the learning situation matching sub-mechanism, common weaknesses in the class and individual ability differences are used as decision nodes to screen a set of potential teaching adjustment schemes that are suitable for the learning situation. In the collaborative filtering sub-mechanism, the adaptation priority of the schemes is calculated by combining the historical recommendation and adoption data of teachers with the same style. The output results of the two mechanisms are weighted and fused to obtain a set of candidate schemes.

[0049] According to another embodiment of the present invention, a cloud-based classroom teaching data management system is provided, comprising the following modules:

[0050] The data acquisition unit is used to acquire classroom teaching data from multiple heterogeneous data sources within the classroom.

[0051] A data processing unit is used to preprocess the acquired data and transmit it to a cloud system composed of edge cloud nodes and a central cloud node for further processing. The edge cloud nodes are deployed locally on campus and are used for real-time processing of teaching data. The real-time processing includes: issuing early warnings of teaching anomalies based on preset rules, and extracting features from the teaching data to obtain feature data. The central cloud node is used to perform in-depth analysis on the feature data uploaded by the edge cloud nodes. The in-depth analysis includes at least: generating student learning status analysis results based on a student dynamic learning profile algorithm, and generating teaching adjustment suggestions based on a teacher dynamic recommendation model. The results of the in-depth analysis are then security- and anonymized to generate a reusable anonymous teaching dataset.

[0052] The data application unit is used to provide teaching assistance applications to at least one of the teacher's end, student's end, and academic affairs management end based on the results of the deep analysis and / or the anonymous teaching dataset.

[0053] Compared with existing technologies, the beneficial effects of the cloud computing-based classroom teaching data management method and system of this invention are:

[0054] This invention utilizes localized deployment of edge cloud nodes to pre-position lightweight processing tasks such as real-time early warning and feature extraction, avoiding bandwidth redundancy in transmitting all data to the central cloud. Combined with a lightweight neural network model specific to teaching scenarios, it enables real-time response to teaching anomaly warnings and simultaneously reduces the burden on the central cloud for deep analysis. This solves the problems of high latency and wasted computing power of edge nodes in traditional single cloud architectures, ensuring immediate feedback and rapid processing of dynamic classroom data, and adapting to the real-time intervention needs of classroom teaching.

[0055] This invention employs a dynamic student learning profiling algorithm that integrates a cognitive diagnostic model with an LSTM network. By combining a subject knowledge graph with the algorithm, a three-dimensional dynamic profile is constructed. Through time-series updates and quantitative calculations, the algorithm identifies students' knowledge weaknesses, skill gaps, and behavioral patterns, overcoming the limitations of traditional static profiles that are disconnected from classroom dynamics. The dynamic teacher recommendation model integrates teacher style clustering, learning situation adaptation, and collaborative filtering to generate personalized suggestions that align with teachers' teaching habits and the learning situation of the class, avoiding the problem of insufficient targeting in general recommendation algorithms. Attached Figure Description

[0056] The accompanying drawings, which form part of this application, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0057] In the attached diagram:

[0058] Figure 1 This is a flowchart illustrating the implementation of the cloud computing-based classroom teaching data management method of the present invention.

[0059] Figure 2 This is a sub-flowchart of the cloud computing-based classroom teaching data management method of the present invention;

[0060] Figure 3 This is a structural block diagram of the lightweight neural network model provided by the present invention;

[0061] Figure 4 This is another sub-flowchart of the cloud computing-based classroom teaching data management method of the present invention;

[0062] Figure 5 The structural block diagram of the teacher dynamic recommendation model provided by this invention;

[0063] Figure 6 This is a structural block diagram of the cloud computing-based classroom teaching data management system of the present invention. Detailed Implementation

[0064] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0065] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0066] Figure 1 This is a flowchart illustrating the implementation of the cloud computing-based classroom teaching data management method of the present invention.

[0067] Please refer to Figure 1 In one embodiment of the present invention, a cloud computing-based classroom teaching data management method is provided, comprising the following steps:

[0068] Step S1: Obtain classroom teaching data from multiple heterogeneous data sources within the classroom;

[0069] In step S1, the multiple heterogeneous data sources of the present invention include classroom smart terminals, audio and video acquisition devices, environmental sensing devices, and academic affairs management systems. Classroom smart terminals can be student tablets, teacher teaching terminals, etc., used to collect student answer results, classroom interaction feedback (such as raising hands, annotations), and teaching resource access records. Audio and video acquisition devices can be classroom cameras, microphones, etc., used to collect students' facial emotional characteristics, behavioral actions (such as daydreaming, interactive participation), classroom voice dialogues, and teacher lecture audio / video. Environmental sensing devices can be noise sensors, light sensors, etc., used to collect environmental data affecting teaching effectiveness, such as classroom noise decibels and light intensity. Academic affairs systems, such as school academic affairs platforms and homework management systems, are used to collect student attendance records, historical academic performance, homework submission status, and basic data on knowledge point mastery.

[0070] Please continue to refer to Figure 1 The cloud-based classroom teaching data management method of this invention further includes:

[0071] Step S2: Preprocess the acquired data and transmit it to the cloud system consisting of edge cloud nodes and central cloud nodes for further processing;

[0072] The present invention includes a data preprocessing step, which includes noise reduction and standardization conversion to remove invalid interference information and unify the data format. In the noise reduction process, noise specific to the teaching scenario, such as environmental noise, invalid data generated by equipment mis-triggered, and background interference pixels in image acquisition, is removed.

[0073] In one implementation of the present invention, the edge cloud node is deployed locally on campus for real-time processing of teaching data. The real-time processing includes: providing early warning of teaching anomalies based on preset rules, and extracting features from the teaching data to obtain feature data.

[0074] Specifically, the preset rules of this invention include scene threshold rules and behavioral feature association rules;

[0075] The scenario threshold rules set quantitative thresholds corresponding to the classroom environment and individual student status. In this invention, edge cloud nodes collect classroom environment data and individual student status data in real time and compare them with the preset quantitative thresholds in real time. Once the data reaches or exceeds the threshold, the corresponding basic warning is immediately triggered, such as abnormal environmental noise or abnormal student inattention. By setting quantitative thresholds, this invention avoids the lag of teachers' subjective judgment and ensures the consistency of warnings.

[0076] The behavioral feature association rules are used to analyze at least two teaching data features to trigger an early warning.

[0077] After the warning is triggered, the edge cloud node synchronously pushes a warning prompt to the teacher's end, which includes the warning type, the objects involved, and the duration of the abnormality, and retains the warning log and the corresponding original data.

[0078] Please refer to Figure 2 In step S2 of the present invention, the step of extracting features from the teaching data to obtain feature data includes:

[0079] Step S211: Using a lightweight neural network model, perform hierarchical feature extraction on the verified teaching data, extracting text features of classroom interaction statements, student behavior and action features, audio and speech features, and basic features of knowledge point relevance. Through the teaching semantic fusion unit, semantically associate the basic features with the preset subject knowledge graph and filter out redundant features that are irrelevant to the teaching objectives.

[0080] Please refer to the details. Figure 3 The lightweight neural network model provided in this embodiment of the invention includes a multimodal input module, a depthwise separable convolution module, a teaching semantic attention module, and a feature fusion output module connected in sequence. Specifically: the multimodal input module is used to receive and process text-based, behavioral image-based, and audio-based teaching data in parallel to extract initial features for each modality; the depthwise separable convolution module uses a depthwise separable convolution structure to perform lightweight convolution processing on the initial features; the teaching semantic attention module is used to assign attention weights to the features after lightweight convolution processing based on a prior knowledge graph dynamically associated with the current teaching scenario; and the feature fusion output module is used to fuse and filter the weighted multimodal features and output the feature data for transmission.

[0081] Furthermore, in the instructional semantic attention module, the prior knowledge graph is associated with at least subject knowledge points, teaching skill points, and classroom behavior patterns. The allocation of attention weights is used to strengthen features related to the current teaching objectives and generate interpretable weight allocation data.

[0082] Please continue to refer to Figure 2 The present invention further includes the step of extracting features from teaching data to obtain feature data, which further includes:

[0083] Step S212: Assign weights to each feature based on the importance of teaching, and finally obtain feature data containing feature type annotations, weight labels and teaching semantic labels.

[0084] In the hierarchical feature extraction step S211 of this embodiment, multimodal basic features are first obtained, and then associated with the subject knowledge graph through the teaching semantic fusion unit to achieve integrated processing of feature extraction and semantic filtering. This ensures the comprehensiveness of features and effectively eliminates irrelevant and redundant information, reducing the computational complexity of subsequent analysis. Step S212 assigns weights and adds labels based on the importance of teaching, so that the feature data has clear priority and interpretability, ensuring that teaching-related features receive key attention. At the same time, feature type annotation and semantic labels also provide input basis for the in-depth analysis of the central cloud node.

[0085] Please continue to refer to Figure 1 The central cloud node of the present invention is used to perform deep analysis on the feature data uploaded by the edge cloud node. The deep analysis includes at least: generating student learning status analysis results based on the student dynamic learning profile algorithm, and generating teaching adjustment suggestions based on the teacher dynamic recommendation model; and performing security and anonymization processing on the results of the deep analysis to generate a reusable anonymous teaching dataset.

[0086] Please continue to refer to Figure 1 The cloud-based classroom teaching data management method of this invention further includes:

[0087] Step S3: Based on the results of the deep analysis and / or the anonymized teaching dataset, provide teaching assistance applications to at least one of the teacher's end, student's end, and academic affairs management end.

[0088] As can be seen, this invention, through local deployment of edge cloud nodes, forwards lightweight processing tasks such as real-time early warning and feature extraction, avoiding bandwidth redundancy in transmitting all data to the central cloud. Combined with a lightweight neural network model specific to teaching scenarios, it enables real-time response to teaching anomaly warnings, while simultaneously reducing the burden on the central cloud for deep analysis. This solves the problems of high latency and wasted computing power of traditional single cloud architectures, ensuring immediate feedback and rapid processing of dynamic classroom data, and adapting to the real-time intervention needs of classroom teaching.

[0089] Please refer to Figure 4 In one implementation of the present invention, step S2, the step of generating student learning status analysis results based on the student dynamic learning profile algorithm, includes:

[0090] Step S221: A dynamic learning profile algorithm for students is constructed using a fusion architecture of cognitive diagnostic model and long short-term memory network. The algorithm takes feature data uploaded from edge cloud as input and combines it with a preset subject knowledge graph to construct a three-dimensional dynamic image of students’ knowledge mastery, ability dimension and learning behavior model.

[0091] Step S222: Map the knowledge point association features, classroom interaction semantic features, and attention temporal features in the feature data into a high-dimensional input vector, and calculate the contribution weight of each feature, expressed as:

[0092]

[0093] In the formula, Represents the normalized weight of the i-th feature. Let K represent the feature vector of the i-th class, and K represent the set of concept nodes in the subject knowledge graph. Represents the semantic similarity function. This represents the teaching scenario adaptation factor, where m represents the total number of feature categories.

[0094] Step S223: Capture the temporal evolution of features using an LSTM network, update the 3D profile incrementally with a preset time window, and output a quantitative value of knowledge point mastery based on the cognitive diagnostic model. Calculate the comprehensive learning state index, which characterizes the analysis results of the student's learning state, as follows:

[0095]

[0096] In the formula, Indicates a composite index. This represents the weighting coefficients for each dimension of the profile. This represents the scores for each dimension. The score indicates the degree of mastery of the knowledge points. This indicates the ability-fit score. This indicates the score for behavioral pattern assessment.

[0097] The fusion architecture in step S221 of this invention combines the advantages of cognitive diagnostic models in assessing knowledge mastery with the strengths of LSTM networks in capturing temporal features. The three-dimensional dynamic profile comprehensively covers the dimensions of student learning, avoiding the one-sidedness of traditional single-dimensional profiles. Step S222 adjusts the contribution weight of each feature to match the actual teaching needs through semantic similarity calculation and teaching scenario adaptation factors, ensuring that the core features play a leading role in profile construction. Step S223 uses LSTM networks to achieve dynamic incremental updates of the profile, solving the problem of the disconnect between traditional static profiles and real-time classroom dynamics. The comprehensive learning status index intuitively represents the student's learning status in a quantitative way, providing a clear and quantifiable basis for generating subsequent teaching adjustment suggestions.

[0098] Furthermore, in step S2, the step of generating teaching adjustment suggestions based on the teacher dynamic recommendation model includes:

[0099] The dynamic recommendation model for teachers is generated by taking the results of student learning status analysis, teachers' teaching history data, subject knowledge graphs and edge cloud feature data as inputs to generate teaching adjustment suggestions.

[0100] From historical teaching data, teacher interaction frequency, knowledge point breakdown methods, teaching pace, and resource preferences were extracted. Using the K-means clustering algorithm, these were categorized into three style clusters: logical breakdown type, interactive guidance type, and resource-dependent type. The similarity coefficient between the target teacher and their respective style cluster was calculated, and expressed as:

[0101]

[0102] In the formula, The style similarity coefficient. The weight of the style feature of class p, Let p be the feature vector of the target teacher. Let p be the feature mean vector of the p-th class belonging to the style cluster. The cosine similarity function;

[0103] Based on the analysis of students' learning status, common knowledge weaknesses in the class, individual ability differences, and classroom attention distribution characteristics are identified. Combining this with the knowledge point correlation of the subject knowledge graph, the priority score of each potential adjustment plan is calculated, as follows:

[0104]

[0105] In the formula, P represents the priority score. To assign weights to the learning situation, M represents the degree of matching between the plan and the class's learning situation. C represents the implementation cost coefficient, and C represents the implementation complexity of the solution.

[0106] Schemes that meet the priority score requirements are selected, and teaching adjustment suggestions are generated based on the teacher's style.

[0107] For further details, please refer to Figure 5 The teacher dynamic recommendation model provided by this invention includes:

[0108] The multi-source input fusion layer is used to process the input data and obtain the fusion feature matrix. The input data includes student learning status analysis result vector, teacher teaching history style feature vector, subject knowledge graph association matrix and classroom real-time interactive feedback feature vector.

[0109] The teacher style feature processing layer includes a built-in K-means clustering submodule and a style adaptation submodule. The clustering submodule is used to divide teacher styles into three clusters: logical decomposition type, interactive guidance type, and resource-dependent type. It calculates the feature fit between the target teacher and its cluster using cosine similarity. The style adaptation submodule dynamically adjusts the feature weights based on the fit.

[0110] The learning situation adaptation decision layer includes a learning situation matching sub-mechanism and a collaborative filtering sub-mechanism. In the learning situation matching sub-mechanism, common weaknesses in the class and individual ability differences are used as decision nodes to screen a set of potential teaching adjustment schemes that are suitable for the learning situation. In the collaborative filtering sub-mechanism, the adaptation priority of the schemes is calculated by combining the historical recommendation and adoption data of teachers with the same style. The output results of the two mechanisms are weighted and fused to obtain a set of candidate schemes.

[0111] This invention employs a dynamic student learning profiling algorithm that integrates a cognitive diagnostic model with an LSTM network. By combining a subject knowledge graph with the algorithm, a three-dimensional dynamic profile is constructed. Through time-series updates and quantitative calculations, the algorithm identifies students' knowledge weaknesses, skill gaps, and behavioral patterns, overcoming the limitations of traditional static profiles that are disconnected from classroom dynamics. The dynamic teacher recommendation model integrates teacher style clustering, learning situation adaptation, and collaborative filtering to generate personalized suggestions that align with teachers' teaching habits and the learning situation of the class, avoiding the problem of insufficient targeting in general recommendation algorithms.

[0112] Please refer to Figure 6 In another embodiment of the present invention, a cloud-based classroom teaching data management system is provided, comprising the following modules:

[0113] Data acquisition unit 11 is used to acquire classroom teaching data from multiple heterogeneous data sources within the classroom;

[0114] The data processing unit 12 is used to preprocess the acquired data and transmit it to a cloud system composed of edge cloud nodes and a central cloud node for further processing. The edge cloud nodes are deployed locally on campus and are used to process teaching data in real time. The real-time processing includes: issuing early warnings of teaching anomalies based on preset rules, and extracting features from the teaching data to obtain feature data. The central cloud node is used to perform in-depth analysis on the feature data uploaded by the edge cloud nodes. The in-depth analysis includes at least: generating student learning status analysis results based on a student dynamic learning profile algorithm, and generating teaching adjustment suggestions based on a teacher dynamic recommendation model. The results of the in-depth analysis are then security- and anonymized to generate a reusable anonymous teaching dataset.

[0115] Data application unit 13 is used to provide teaching assistance applications to at least one of the teacher terminal, student terminal and academic affairs management terminal based on the results of the deep analysis and / or the anonymous teaching dataset.

[0116] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0117] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0118] In the above embodiments of this application, the information collected is information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data all comply with relevant laws, regulations and standards, take necessary protective measures, do not violate public order and good morals, and provide corresponding operation entry points for users to choose to authorize or refuse.

[0119] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0120] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0121] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0122] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to related technologies, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0123] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A cloud computing-based classroom teaching data management method, characterized in that, Includes the following steps: Step S1: Obtain classroom teaching data from multiple heterogeneous data sources within the classroom; Step S2: Preprocess the acquired data and transmit it to a cloud system consisting of edge cloud nodes and central cloud nodes for further processing; wherein: The edge cloud nodes are deployed locally on campus and are used to process teaching data in real time. The real-time processing includes: issuing early warnings of teaching anomalies based on preset rules, and extracting features from the teaching data to obtain feature data. The central cloud node is used to perform in-depth analysis on the feature data uploaded by the edge cloud node. The in-depth analysis includes at least: generating student learning status analysis results based on the student dynamic learning profile algorithm, and generating teaching adjustment suggestions based on the teacher dynamic recommendation model. The results of the deep analysis are made secure and anonymized to generate a reusable anonymous teaching dataset; Step S3: Based on the results of the deep analysis and / or the anonymized teaching dataset, provide teaching assistance applications to at least one of the teacher's end, student's end, and academic affairs management end.

2. The classroom teaching data management method based on cloud computing according to claim 1, characterized in that, In step S1, the multiple heterogeneous data sources include classroom smart terminals, audio and video acquisition devices, environmental sensing devices, and the academic affairs management system.

3. The classroom teaching data management method based on cloud computing according to claim 2, characterized in that, In step S2, the preset rules include scene threshold rules and behavior feature association rules; The scenario threshold rules set quantitative thresholds corresponding to the classroom environment and individual student status. The behavioral feature association rules are used to analyze at least two teaching data features to trigger an early warning. After the warning is triggered, the edge cloud node synchronously pushes a warning prompt to the teacher's end, which includes the warning type, the objects involved, and the duration of the abnormality, and retains the warning log and the corresponding original data.

4. The classroom teaching data management method based on cloud computing according to claim 3, characterized in that, In step S2, the step of extracting features from the teaching data to obtain feature data includes: A lightweight neural network model is used to extract hierarchical features from the verified teaching data, including text features of classroom interaction statements, student behavior features, audio and speech features, and basic features of knowledge point relevance. Through the teaching semantic fusion unit, the basic features are semantically associated with the preset subject knowledge graph, and redundant features that are irrelevant to the teaching objectives are filtered out. Weights are assigned to each feature based on its importance in teaching, resulting in feature data containing feature type annotations, weight labels, and teaching semantic labels.

5. The classroom teaching data management method based on cloud computing according to claim 4, characterized in that, The lightweight neural network model includes a multimodal input module, a depthwise separable convolution module, an instructional semantic attention module, and a feature fusion output module connected in sequence. in: The multimodal input module is used to receive and process text, behavioral image, and audio teaching data in parallel to extract initial features of each modality. The depthwise separable convolution module uses a depthwise separable convolution structure to perform lightweight convolution processing on the initial features; The teaching semantic attention module is used to assign attention weights to the features after lightweight convolution processing based on a prior knowledge graph that is dynamically associated with the current teaching scenario. The feature fusion output module is used to fuse and filter the weighted multimodal features and output the feature data for transmission.

6. The cloud computing-based classroom teaching data management method according to claim 5, characterized in that, In the instructional semantic attention module, the prior knowledge graph is associated with at least subject knowledge points, teaching skill points, and classroom behavior patterns. The allocation of attention weights is used to strengthen features related to the current teaching objectives and generate interpretable weight allocation data.

7. The cloud computing-based classroom teaching data management method according to claim 6, characterized in that, In step S2, the step of generating student learning status analysis results based on the student dynamic learning profile algorithm includes: constructing a student dynamic learning profile algorithm using a fusion architecture of cognitive diagnostic model and long short-term memory network, taking feature data uploaded from edge cloud as input, and combining it with a preset subject knowledge graph to construct a three-dimensional dynamic image of student knowledge mastery, ability dimension and learning behavior model. The knowledge point association features, classroom interaction semantic features, and attention temporal features in the feature data are mapped to a high-dimensional input vector, and the contribution weight of each feature is calculated, represented as follows: In the formula, Represents the normalized weight of the i-th feature. Let K represent the feature vector of the i-th class, and K represent the set of concept nodes in the subject knowledge graph. Represents the semantic similarity function. This represents the teaching scenario adaptation factor, where m represents the total number of feature categories. The LSTM network captures the temporal evolution of features, updates the 3D profile incrementally with a preset time window, and outputs a quantitative value of knowledge point mastery based on a cognitive diagnostic model. A comprehensive learning state index, used to characterize the analysis results of student learning status, is calculated and expressed as: In the formula, Indicates a composite index. This represents the weighting coefficients for each dimension of the profile. This represents the scores for each dimension. The score indicates the degree of mastery of the knowledge points. This indicates the ability-fit score. This indicates the score for behavioral pattern assessment.

8. The cloud computing-based classroom teaching data management method according to claim 7, characterized in that, Step S2, the step of generating teaching adjustment suggestions based on the teacher dynamic recommendation model, includes: The dynamic recommendation model for teachers is generated by taking the results of student learning status analysis, teachers' teaching history data, subject knowledge graphs and edge cloud feature data as inputs to generate teaching adjustment suggestions. From historical teaching data, teacher interaction frequency, knowledge point breakdown methods, teaching pace, and resource preferences were extracted. Using the K-means clustering algorithm, these were categorized into three style clusters: logical breakdown type, interactive guidance type, and resource-dependent type. The similarity coefficient between the target teacher and their respective style cluster was calculated, and expressed as: In the formula, The style similarity coefficient. The weight of the style feature of class p, Let p be the feature vector of the target teacher. Let p be the feature mean vector of the p-th class belonging to the style cluster. The cosine similarity function; Based on the analysis of students' learning status, common knowledge weaknesses in the class, individual ability differences, and classroom attention distribution characteristics are identified. Combining this with the knowledge point correlation of the subject knowledge graph, the priority score of each potential adjustment plan is calculated, as follows: In the formula, P represents the priority score. To assign weights to the learning situation, M represents the degree of matching between the plan and the class's learning situation. C represents the implementation cost coefficient, and C represents the implementation complexity of the solution. Schemes that meet the priority score requirements are selected, and teaching adjustment suggestions are generated based on the teacher's style.

9. The cloud computing-based classroom teaching data management method according to claim 8, characterized in that, The teacher dynamic recommendation model includes: The multi-source input fusion layer is used to process the input data and obtain the fusion feature matrix. The input data includes student learning status analysis result vector, teacher teaching history style feature vector, subject knowledge graph association matrix and classroom real-time interactive feedback feature vector. The teacher style feature processing layer includes a built-in K-means clustering submodule and a style adaptation submodule. The clustering submodule is used to divide teacher styles into three clusters: logical decomposition type, interactive guidance type, and resource-dependent type. It calculates the feature fit between the target teacher and its cluster using cosine similarity. The style adaptation submodule dynamically adjusts the feature weights based on the fit. The learning situation adaptation decision layer includes a learning situation matching sub-mechanism and a collaborative filtering sub-mechanism. In the learning situation matching sub-mechanism, the common weaknesses of the class and the differences in individual abilities are used as decision nodes to screen a set of potential teaching adjustment schemes that are suitable for the learning situation. In the collaborative filtering sub-mechanism, the historical recommendation and adoption data of teachers with the same style are combined to calculate the adaptation priority of the schemes. The output results of the two mechanisms are weighted and fused to obtain a set of candidate schemes.

10. A management system for implementing the cloud computing-based classroom teaching data management method as described in any one of claims 1 to 9, characterized in that, Includes the following modules: The data acquisition unit is used to acquire classroom teaching data from multiple heterogeneous data sources within the classroom. The data processing unit is used to preprocess the acquired data and transmit it to the cloud system consisting of edge cloud nodes and central cloud nodes for further processing. The edge cloud nodes are deployed locally on campus for real-time processing of teaching data. This real-time processing includes: issuing early warnings of teaching anomalies based on preset rules, and extracting features from the teaching data to obtain feature data. The central cloud node is used to perform in-depth analysis of the feature data uploaded by the edge cloud nodes. This in-depth analysis includes at least: generating student learning status analysis results based on a student dynamic learning profile algorithm, and generating teaching adjustment suggestions based on a teacher dynamic recommendation model. The results of the in-depth analysis are then security- and anonymized to generate a reusable anonymous teaching dataset. The data application unit is used to provide teaching assistance applications to at least one of the teacher's end, student's end, and academic affairs management end based on the results of the deep analysis and / or the anonymous teaching dataset.

Images