An internet-based respiratory infectious disease prevention and control intelligent training system

Abstract

This invention relates to the field of intelligent training technology, specifically to an internet-based intelligent training system for the prevention and control of respiratory infectious diseases. In this invention, a long short-term memory network is used to deeply model learners' learning progress, identify weaknesses in the learning process, and accurately push corresponding content, effectively providing personalized learning path recommendations. A graph neural network-based learning resource scheduling and content push method dynamically adjusts the frequency and order of learning content pushes to match learners' progress, ensuring that learning content is pushed at the appropriate time and with suitable difficulty. By constructing an interaction graph among learners and combining it with influence assessment, key prevention and control knowledge is disseminated preferentially through learners with high influence, optimizing the knowledge dissemination path. Combined with the application of the SIR (Self-Improving Influence) propagation model, the system predicts learners' progress in learning prevention and control knowledge, identifies learning bottlenecks, and significantly improves learning effectiveness.

Description

Technical Field

[0001] This invention relates to the field of smart training technology, and in particular to a smart training system for the prevention and control of respiratory infectious diseases based on the Internet. Background Technology

[0002] The field of smart training technology aims to provide an efficient, personalized, and flexible learning experience through modern information technology, breaking through the limitations of time and space, enabling learners to access learning resources anytime, anywhere, improving learning efficiency and quality, and helping learners to grasp their learning progress and effects in real time through intelligent assessment and feedback mechanisms, achieving personalized learning paths and goals.

[0003] The purpose of the internet-based intelligent training system for respiratory infectious disease prevention and control is to help government departments, health institutions, medical personnel, and the public master effective respiratory infectious disease prevention and control measures and emergency response capabilities through digital and networked methods. By providing convenient online learning, prevention and control knowledge can be disseminated rapidly, enhancing awareness and response capabilities to infectious diseases and reducing public health risks caused by information asymmetry and untimely learning.

[0004] Existing technologies have limitations in terms of processing logic and technical means. They rely on standardized learning paths and fixed content delivery, failing to personalize learning based on individual differences. This results in some learners not mastering key knowledge in a timely manner and lacks the ability to accurately analyze learner behavior and progress. It is also impossible to identify weaknesses in learners' knowledge acquisition or effectively analyze interactions among learners. Consequently, the dissemination of prevention and control knowledge cannot effectively leverage the influence among learners. Furthermore, there is a lack of precise prediction and dynamic adjustment mechanisms based on learner progress in terms of adjusting the frequency, order, and difficulty of learning content delivery. This results in some learners not receiving timely help and support, affecting the continuity and motivation of learning. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing an internet-based intelligent training system for the prevention and control of respiratory infectious diseases.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A smart training system for the prevention and control of respiratory infectious diseases based on the Internet includes:

[0007] Learner Behavior Analysis Module: Collects learners' learning duration, frequency, and answer accuracy data; analyzes learners' mastery of knowledge on respiratory infectious disease prevention and control; calculates activity and participation; groups learners; and generates learner behavioral characteristic data.

[0008] Personalized learning path recommendation module: Based on the learner's behavioral characteristic data, the module uses a long short-term memory network to analyze the learner's learning progress on prevention and control knowledge, identifies weak links, pushes review content related to the weak links, and pushes advanced knowledge points for learners with fast learning progress, generating personalized learning path recommendation data and adjusting the learning path.

[0009] Knowledge dissemination path optimization module: Analyze the interaction behavior between learners, construct an interaction relationship diagram, evaluate the influence, optimize the knowledge dissemination path, prioritize the dissemination of key prevention and control knowledge through influential learners, and generate optimized dissemination paths;

[0010] Learning progress prediction and scheduling module: Based on the learner's behavioral characteristic data and personalized learning path recommendation data, predict the learner's progress and bottlenecks in learning prevention and control knowledge, identify learning difficulties, and generate learning progress prediction data.

[0011] Learning resource scheduling and content push module: Combining the learning progress prediction data and propagation optimization path, a graph neural network is used to dynamically adjust the frequency, order and difficulty of learning content push, and the SIR propagation model is used to prioritize the push of review content and advanced learning tasks, generating a learning content push plan.

[0012] As a further aspect of the present invention, the learner behavior analysis module includes:

[0013] Data acquisition submodule: Collects learners' learning duration, frequency, and answer accuracy data, stores the specific duration and frequency of each learning session, records the correctness of answers, and generates learner behavior data;

[0014] Behavior Analysis Submodule: Based on the learner behavior data, by quantifying the learning duration, frequency and answer accuracy data, analyze the learner's learning input and mastery of each knowledge, calculate the mastery and participation of each prevention and control knowledge, and generate learner mastery analysis data;

[0015] Learner grouping submodule: Based on the learners' mastery data analysis, and according to the learners' learning performance, clustering methods are used to identify groups of learners with similar behaviors, and different learner groups are divided to generate learner behavioral characteristic data.

[0016] As a further aspect of the present invention, the personalized learning path recommendation module includes:

[0017] Learning progress analysis submodule: Based on the learner behavioral characteristic data, the Long Short-Term Memory Network is used to analyze the learning completion status of each learner in the respiratory infectious disease prevention and control knowledge module. Combined with historical data, the current learning progress of the learners is evaluated, the current learning status of the learners is calculated, and the learner's learning progress data is obtained.

[0018] Weakness identification submodule: Based on the learner's learning progress data, compare the learner's learning performance in different modules with the overall group performance, identify the learner's weak points, calculate the gap of the learner in specific prevention and control knowledge points, and obtain the learner's weak point data;

[0019] The learning content recommendation submodule analyzes learners' knowledge gaps based on the data on their weak areas, recommends review content related to those gaps, and pushes advanced knowledge points to learners who are making rapid progress, generating personalized learning path recommendation data.

[0020] As a further aspect of the present invention, the Long Short-Term Memory network adopts the formula:

[0021]

[0022] in: This indicates the learner's learning status at the current time step. This indicates the learner's learning status at the previous time step. This represents the input data at the current time step. Data representing the individual characteristics of learners, This indicates the difficulty level of the current learning module. Represents the weight matrix. Indicates the bias term. This represents the activation function.

[0023] As a further aspect of the present invention, the knowledge dissemination path optimization module includes:

[0024] Interactive Behavior Analysis Submodule: Based on the interactive data of learners in the process of learning about respiratory infectious disease prevention and control, collect discussion records, Q&A interactions and information sharing frequency among learners, construct an interaction relationship diagram, mark the interactive participation of each learner, record the connection strength and interaction frequency between learners, and obtain learner interaction data;

[0025] Influence Assessment Submodule: Based on the learner interaction data, calculate the interaction frequency and duration of each learner in the interaction relationship graph, identify high-influence learners by analyzing the learners' participation intensity and influence scope, mark them as key players in disseminating prevention and control knowledge, and obtain learner influence data;

[0026] The dissemination path optimization submodule generates an optimized dissemination path by adjusting the dissemination order and prioritizing learners with high influence to disseminate key prevention and control knowledge, based on the learner influence data.

[0027] As a further aspect of the present invention, the learning progress prediction and scheduling module includes:

[0028] Progress prediction submodule: Based on the learner behavior characteristic data and personalized learning path recommendation data, analyze the learner's learning progress in each prevention and control knowledge module, predict the learner's future learning progress by comparing historical learning records with current behavior, and obtain learning progress prediction data.

[0029] Bottleneck identification submodule: Based on the learning progress prediction data, it tracks the learner's learning progress in prevention and control knowledge, identifies learners with lagging progress, and locates learning bottlenecks and obtains learner bottleneck data based on the learner's learning performance.

[0030] Progress scheduling submodule: Based on the learner bottleneck data, adjust the allocation of learning resources, prioritize providing appropriate review resources for learners with slow progress, and generate learning progress prediction data.

[0031] As a further aspect of the present invention, the learning resource scheduling and content push module includes:

[0032] Progress assessment submodule: Based on the learning progress prediction data and the transmission optimization path, the module tracks the learner's learning progress in respiratory infectious disease prevention and control knowledge, calculates the learner's learning completion rate in each module, analyzes the learner's learning status, determines whether there is a progress lag, and obtains learning progress assessment data.

[0033] Content Priority Adjustment Submodule: Based on the learning progress assessment data, a graph neural network is used to identify modules in which learners are weak in learning prevention and control knowledge. By analyzing the learners' weaknesses and combining them with the learning progress, the priority of pushing learning content is adjusted, prioritizing the push of review modules and modules with lower difficulty, and generating content push priority data.

[0034] The push plan generation submodule: Based on the content push priority data, it adopts the SIR propagation model, combines the learners' learning needs, dynamically adjusts the push order, frequency and difficulty of learning content, and calculates the propagation speed of learning content by simulating the propagation process of learning content. Based on the propagation speed, it generates a suitable push plan and generates a learning content push plan.

[0035] As a further aspect of the present invention, the graph neural network adopts the formula:

[0036]

[0037] in: Indicates the first In the next iteration, the learning content The hidden state, Indicates the first In the next iteration, the learning content The hidden state, Representation and learning content The set of connected adjacent nodes. Indicates learning content The weight matrix, Indicates learning content The adjacent module weight matrix, For activation function, These are weighting coefficients. For learning content The learning difficulty level Learning content Learning performance value, Indicates the content being studied. Adjacent units In the Hidden state in the next iteration.

[0038] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0039] 1. In this invention, a long short-term memory network is used to deeply model the learner's learning progress, identify weak links in the learning process, and accurately push corresponding content, effectively providing personalized learning path recommendations;

[0040] 2. In this invention, a learning resource scheduling and content push method based on graph neural networks is used to dynamically adjust the frequency and order of learning content push, match the learner's progress, and ensure that learning content is pushed at the appropriate time and with appropriate difficulty, thereby reducing the situation where learning content does not match the learner's ability.

[0041] 3. In this invention, by constructing an interaction graph among learners and combining it with influence assessment, key prevention and control knowledge is disseminated preferentially through learners with high influence, thus optimizing the knowledge dissemination path. By combining the application of the SIR dissemination model, the progress of learners in learning prevention and control knowledge is predicted, learning bottlenecks are identified, and corresponding measures are taken to adjust the learning plan. This effectively improves the personalization of the learning path, the accuracy of learning content delivery, and the efficiency of knowledge dissemination, significantly improving the learning effect. Attached Figure Description

[0042] Figure 1 This is a system flowchart of the present invention;

[0043] Figure 2 This is a schematic diagram of the system framework of the present invention. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0045] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0046] Please see Figure 1 This invention provides a technical solution: an internet-based intelligent training system for the prevention and control of respiratory infectious diseases, comprising:

[0047] Learner Behavior Analysis Module: Collects learners' learning duration, frequency, and answer accuracy data; analyzes learners' mastery of knowledge on respiratory infectious disease prevention and control; calculates activity and participation; groups learners; and generates learner behavioral characteristic data.

[0048] Personalized learning path recommendation module: Based on learner behavioral characteristic data, it uses long short-term memory network to analyze learners' learning progress on prevention and control knowledge, identify weak links, push review content related to the weak links, and push advanced knowledge points for learners with fast learning progress, generate personalized learning path recommendation data and adjust the learning path;

[0049] Knowledge dissemination path optimization module: Analyze the interaction behavior between learners, construct an interaction relationship diagram, evaluate the influence, optimize the knowledge dissemination path, prioritize the dissemination of key prevention and control knowledge through influential learners, and generate optimized dissemination paths;

[0050] Learning progress prediction and scheduling module: Based on learner behavioral characteristic data and personalized learning path recommendation data, predict learners’ progress and bottlenecks in learning prevention and control knowledge, identify learning difficulties, and generate learning progress prediction data.

[0051] The learning resource scheduling and content delivery module combines learning progress prediction data and propagation optimization paths, uses graph neural networks to dynamically adjust the frequency, order, and difficulty of learning content delivery, and employs the SIR propagation model to prioritize the delivery of review content and advanced learning tasks, generating a learning content delivery plan.

[0052] Please see Figure 2 The learner behavior analysis module includes:

[0053] Data acquisition submodule: Collects learners' learning duration, frequency, and answer accuracy data, stores the specific duration and frequency of each learning session, records the correctness of answers, and generates learner behavior data;

[0054] Behavior Analysis Submodule: Based on learner behavior data, this module quantifies learning duration, frequency, and answer accuracy to analyze learners' learning input and mastery of each piece of knowledge, calculates the degree of mastery and participation in each prevention and control knowledge, and generates learner mastery analysis data.

[0055] Learner grouping submodule: Based on learners’ mastery data analysis, clustering methods are used to identify learner groups with similar behaviors according to learners’ learning performance, and different learner groups are divided to generate learner behavioral characteristic data;

[0056] The data acquisition submodule collects learning duration, frequency, and accuracy data based on learners' learning activities. It uses Python's `time` module to obtain the start and end times of each learning session, calculates the learning duration, stores each learning session's record using a `DataFrame` structure from the pandas library, records the frequency of each session, merges multiple learning sessions using the `groupby` function in pandas, calculates the mean and standard deviation of the frequencies using the NumPy library, calculates the accuracy of each learning session using the `mean` function in NumPy, stores the correctness of each learner's answers, and finally generates learner behavior data.

[0057] Behavioral Analysis Submodule: Based on the collected learner behavioral data, the KMeans clustering algorithm in scikit-learn is used. By setting the number of clusters to 5, the learning duration, learning frequency, and answer accuracy data are quantitatively analyzed. The fit function of KMeans is used to fit the data, calculate the distance from each data point to the cluster center, and generate learners' learning input and mastery of each prevention and control knowledge module based on the data. The SilhouetteScore method of scikit-learn is further used to evaluate the clustering effect. The learner data of each cluster is classified using pandas to obtain the mastery and participation of each prevention and control knowledge module, and generate learner mastery analysis data.

[0058] Learner Grouping Submodule: Based on learners' mastery levels, this module analyzes data using the DBSCAN algorithm from scikit-learn, with parameters eps=0.5 and min_samples=5. It performs cluster analysis on the learner data, using DBSCAN's fit function to automatically identify different learner groups. Core points are extracted using DBSCAN's core_sample_indices attribute, and learners are divided into different learning groups based on these core points. The pandas groupby function is then used to group the data according to the learners' cluster labels. Further behavioral characteristic analysis is performed on each group to derive their learning characteristics, generating learner behavioral characteristic data.

[0059] Please see Figure 2 The personalized learning path recommendation module includes:

[0060] Learning Progress Analysis Submodule: Based on learner behavioral characteristic data, the module uses a long short-term memory network to analyze each learner's learning completion status in the respiratory infectious disease prevention and control knowledge module. Combined with historical data, it assesses the learner's current learning progress, calculates the learner's current learning status, and obtains learner learning progress data.

[0061] Weakness Identification Submodule: Based on learner learning progress data, compare learners' learning scores in different modules with the overall group scores to identify learners' weak points, calculate learners' gaps in specific prevention and control knowledge points, and obtain learner's weak point data;

[0062] The learning content recommendation submodule analyzes learners' knowledge gaps based on data on their weak areas, recommends review content related to those weak areas, and pushes advanced knowledge points to learners who are making rapid progress, generating personalized learning path recommendation data.

[0063] Learning Progress Analysis Submodule: Based on learner behavioral characteristic data, a Long Short-Term Memory (LSTM) network is used to analyze the learning progress of each learner in the respiratory infectious disease prevention and control knowledge module. The LSTM layer is trained on historical data with 128 units and a learning rate of 0.001. Batch Normalization is used for normalization. When inputting data, each learner's learning duration, frequency, and answer accuracy are used as time-series data to model their learning progress. A time step of 50 is set, and the Adam optimizer is used for training. The mean squared error loss function is set to evaluate the learner's current learning progress and generate learner learning progress data.

[0064] The weak link identification submodule: Based on learner learning progress data, the KMeans clustering algorithm is used to compare learners' learning scores in different modules with the overall group scores. The fit_predict function is used to cluster the data. By calculating the gap between learners' learning scores in each module and the group average score, learners are classified based on the cluster labels generated in the KMeans model. The weak links of learners in specific prevention and control knowledge points are identified. The performance gap of learners in weak links is further calculated. The mean function of NumPy is used to calculate the average gap of each group in weak links, generating learner weak link data.

[0065] The learning content recommendation submodule: Based on learner weakness data, it uses a collaborative filtering algorithm to analyze learners' knowledge gaps. By setting the number of factors to 50 and the learning rate to 0.005, and using SVD for matrix factorization, it trains the learning set using the `trainset.build_full_trainset()` method on the historical interaction data between learners and learning content to obtain a learning preference matrix for each learner. Combining the learner's weaknesses, it recommends review content related to those weaknesses. For learners who are making rapid progress, it recommends advanced knowledge points. The `predict` function predicts each learner's preferences to generate personalized learning path recommendation data.

[0066] Long Short-Term Memory (LSTM) networks use the following formula:

[0067]

[0068] in: This indicates the learner's learning status at the current time step. This indicates the learner's learning status at the previous time step. This represents the input data at the current time step. Data representing the individual characteristics of learners, This indicates the difficulty level of the current learning module. Represents the weight matrix. Indicates the bias term. Indicates the activation function;

[0069] Execution process: First, This represents the learner's learning state at the previous time step, providing historical information for calculating the current state and reflecting the learner's learning trajectory. It also represents the input data for the current time step. This includes learner behavioral characteristics, such as learning duration, task completion rate, and frequency of interaction, to help the model assess the learner's current learning status; individual characteristic data. Includes learner background information, such as age, learning habits, and health status, which helps adjust learning progress analysis to better suit individual learning characteristics and difficulty levels. This is used to measure the difficulty of the current knowledge module, ensuring that the learning pace can be adjusted according to the learner's learning ability for modules of different difficulty levels. Input data will be processed through a weight matrix. Weighted summation is performed to calculate a comprehensive input signal, which is then added to the bias term. This provides the model with the final adjustment to the current learning state, and finally, through the activation function. A nonlinear transformation is applied to the weighted sum to determine the learner's learning state at the current moment. It reflects the learner's learning progress and helps to assess the learner's learning status in the respiratory infectious disease prevention and control knowledge module in real time.

[0070] Please see Figure 2 The knowledge dissemination path optimization module includes:

[0071] Interactive Behavior Analysis Submodule: Based on the interactive data of learners in the process of learning about respiratory infectious disease prevention and control, collect discussion records, Q&A interactions and information sharing frequency among learners, construct an interaction relationship diagram, mark the interactive participation of each learner, record the connection strength and interaction frequency between learners, and obtain learner interaction data;

[0072] Influence Assessment Submodule: Based on learner interaction data, calculate the interaction frequency and duration of each learner in the interaction relationship graph. By analyzing the learner's participation intensity and influence scope, identify high-influence learners, mark them as key players in disseminating prevention and control knowledge, and obtain learner influence data.

[0073] The dissemination path optimization submodule generates an optimized dissemination path by adjusting the dissemination order and prioritizing learners with high influence to disseminate key prevention and control knowledge.

[0074] Interactive Behavior Analysis Submodule: Based on learners' interaction data during the learning process of respiratory infectious disease prevention and control knowledge, this module collects discussion records, Q&A interactions, and information sharing frequencies among learners. Using a graph database, it models the interactions between learners by setting node types as learners and relationship types as interactions. The Cypher query language is used for data input, calculating the interaction frequency and participation of each learner. The NEO4J apoc function library is used to calculate the connection strength between each learner and other learners. The apoc.coll.score() function is executed to analyze the interaction frequency between learners, marking each learner's interaction participation, resulting in an interaction relationship graph. The module records the connection strength and interaction frequency between learners, generating learner interaction data.

[0075] The influence assessment submodule calculates the frequency and duration of each learner's interaction based on learner interaction data in the interaction relationship graph. It uses the PageRank function to sort each learner node in the graph, sets alpha=0.85 as a damping factor, and iteratively calculates the influence of each node. It controls the precision by setting max_iter=100 and tol=1e-6, and calculates the learner's participation intensity and influence range. It uses NumPy's mean function to calculate the learner's influence range, identifies high-influence learners, marks them as key players in disseminating prevention and control knowledge, and generates learner influence data.

[0076] The propagation path optimization submodule adjusts the propagation order based on learner influence data, prioritizing learners with high influence to spread key prevention and control knowledge. The parameter of the greedy_color function is set to graph. The greedy_color function calculates the propagation order for each learner and determines the propagation order based on the learner's influence data, generating an optimized propagation path.

[0077] Please see Figure 2 The learning progress prediction and scheduling module includes:

[0078] Progress prediction submodule: Based on learner behavior characteristic data and personalized learning path recommendation data, analyze learners’ learning progress in each prevention and control knowledge module, and predict learners’ future learning progress by comparing historical learning records with current behavior, and obtain learning progress prediction data.

[0079] Bottleneck identification submodule: Based on learning progress prediction data, it tracks learners' learning progress in prevention and control knowledge, identifies learners who are lagging behind, and locates learning bottlenecks based on learners' learning performance, and obtains learner bottleneck data.

[0080] Progress scheduling submodule: Based on learner bottleneck data, adjust the allocation of learning resources, prioritize providing appropriate review resources for learners with slow progress, and generate learning progress prediction data;

[0081] Progress prediction submodule: Based on learner behavior feature data and personalized learning path recommendation data, it adopts a linear regression algorithm, uses the fit function to train on historical learning records and current behavior data, sets the input features as learning duration, frequency and answer accuracy, and the target variable as learning progress, performs model fitting, and uses the predict function to predict the learner's future learning progress to generate learning progress prediction data.

[0082] Bottleneck identification submodule: Based on learning progress prediction data, the K-nearest neighbor algorithm is used to track the learner's learning progress. The fit function is used to train historical progress data, generate a model and make predictions. The predict function is used to classify the current learner's progress and calculate the learner's progress lag degree, identify learners with progress lag, locate learning bottlenecks, and generate learner bottleneck data.

[0083] The progress scheduling submodule, based on learner bottleneck data, employs a greedy algorithm (specifically a greedy selection strategy that prioritizes allocating review resources to learners lagging behind). By setting the progress differences in learner bottleneck data as the basis for resource allocation, it provides appropriate review resources to learners with slower learning progress one by one, sorts them according to the degree of learning progress lag, prioritizes allocating more resources to lagging learners, and generates learning progress prediction data.

[0084] Please see Figure 2 The learning resource scheduling and content delivery module includes:

[0085] Progress assessment submodule: Based on learning progress prediction data and transmission optimization path, it tracks learners' learning progress in respiratory infectious disease prevention and control knowledge, calculates learners' learning completion rate in each module, analyzes learners' learning status, determines whether there is a progress lag, and obtains learning progress assessment data.

[0086] Content Priority Adjustment Submodule: Based on learning progress assessment data, a graph neural network is used to identify modules in which learners are weak in learning epidemic prevention knowledge. By analyzing learners' weaknesses and combining them with learning progress, the priority of pushing learning content is adjusted, prioritizing review modules and modules with lower difficulty, and generating content push priority data.

[0087] Push plan generation submodule: Based on content push priority data, using the SIR propagation model, and combined with learners' learning needs, dynamically adjust the push order, frequency and difficulty of learning content, and calculate the propagation speed of learning content by simulating the propagation process of learning content, and generate a suitable push plan based on the propagation speed.

[0088] Progress Assessment Submodule: Based on learning progress prediction data and propagation optimization paths, this module uses a support vector machine to track learners' learning progress in respiratory infectious disease prevention and control knowledge. It uses the `fit` function to train the model on historical learning data, generating a learning progress assessment model. Inputting the learning progress prediction data, it calculates the learner's completion rate in each module, analyzes the learner's learning status, determines if there are any progress delays, and uses the `predict` function to assess the learner's current learning progress, generating learning progress assessment data.

[0089] Content Priority Adjustment Submodule: Based on learning progress assessment data, a graph convolutional network is used with hidden_channels=64 and learning rate=0.005 to identify modules where learners perform weakly in learning epidemic prevention knowledge. By constructing a graph data structure, the learner's learning scores in each module are input, and the GCNConv layer is used to perform convolution operations on the graph data. By setting epochs=100 and batch_size=32, the GraphSAGE method is used for training. Combined with the learning progress, the learner's weak points are analyzed, and the priority of pushing learning content is adjusted, prioritizing review modules and modules with lower difficulty, generating content push priority data.

[0090] The push plan generation submodule: Based on content push priority data, it adopts the SIR propagation model and combines learners' learning needs to dynamically adjust the push order, frequency, and difficulty of learning content. By setting the initial state S0=0.9, I0=0.1, R0=0.0, infection rate beta=0.3, recovery rate gamma=0.1, it uses the odeint function to simulate the propagation process of learning content, and controls the time step by setting t_span=[0, 100] to calculate the propagation speed of learning content, adjust the push order, frequency, and difficulty, and generate a learning content push plan.

[0091] Graph neural networks, using the formula:

[0092]

[0093] in: Indicates the first In the next iteration, the learning content The hidden state, Indicates the first In the next iteration, the learning content The hidden state, Representation and learning content The set of connected adjacent nodes. Indicates learning content The weight matrix, Indicates learning content The adjacent module weight matrix, For activation function, These are weighting coefficients. For learning content The learning difficulty level Learning content Learning performance value, Indicates the content being studied. Adjacent units In the Hidden state at the next iteration;

[0094] Execution process: First, Indicates learning content The learning status at the time of the last iteration reflects the learner's historical learning progress in this module. Is related to the learning content The related set of adjacent nodes represents the set of nodes connected to the adjacent nodes. Other learning content and the interrelationships between modules were used to adjust learning priorities, and the influence of each adjacent node was normalized using coefficients. To ensure a balance in the relationships between modules during computation, a weight matrix is ​​applied to the state of the current learning content and the states of adjacent modules, and then a weighted sum is performed to reflect the relative importance of each module in the push. The weighted sum is obtained through an activation function. Then, a nonlinear transformation is performed to obtain the updated value of the current learning state. , These are weighting coefficients used to adjust the difficulty level of the modules. and learning performance value Impact on push notification priority, learning difficulty level This reflects the complexity of the module, while the learning performance value This indicates the learner's performance in this module; content with poor performance will be pushed to them first for review.

[0095] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. An Internet-based respiratory infectious disease prevention and control intelligent training system, characterized in that, The system includes: Learner Behavior Analysis Module: Collects learners' learning duration, frequency, and answer accuracy data; analyzes learners' mastery of knowledge on respiratory infectious disease prevention and control; calculates activity and participation; groups learners; and generates learner behavioral characteristic data. Personalized learning path recommendation module: Based on the learner's behavioral characteristic data, the module uses a long short-term memory network to analyze the learner's learning progress on prevention and control knowledge, identifies weak links, pushes review content related to the weak links, and pushes advanced knowledge points for learners with fast learning progress, generating personalized learning path recommendation data and adjusting the learning path. Knowledge dissemination path optimization module: Analyze the interaction behavior between learners, construct an interaction relationship diagram, evaluate the influence, optimize the knowledge dissemination path, prioritize the dissemination of key prevention and control knowledge through influential learners, and generate optimized dissemination paths; Learning progress prediction and scheduling module: Based on the learner's behavioral characteristic data and personalized learning path recommendation data, predict the learner's progress and bottlenecks in learning prevention and control knowledge, identify learning difficulties, and generate learning progress prediction data. Learning resource scheduling and content push module: Combining the learning progress prediction data and propagation optimization path, a graph neural network is used to dynamically adjust the frequency, order and difficulty of learning content push, and the SIR propagation model is used to prioritize the push of review content and advanced learning tasks, and generate a learning content push plan. The learner behavior analysis module includes: Data acquisition submodule: Collects learners' learning duration, frequency, and answer accuracy data, stores the specific duration and frequency of each learning session, records the correctness of answers, and generates learner behavior data; Behavior Analysis Submodule: Based on the learner behavior data, by quantifying the learning duration, frequency and answer accuracy data, analyze the learner's learning input and mastery of each knowledge, calculate the mastery and participation of each prevention and control knowledge, and generate learner mastery analysis data; Learner grouping submodule: Based on the learners' mastery data analysis, and according to the learners' learning performance, clustering methods are used to identify groups of learners with similar behaviors, and different learner groups are divided to generate learner behavioral characteristic data. 2.The Internet-based respiratory infectious disease prevention and control intelligent training system according to claim 1, characterized in that, The personalized learning path recommendation module includes: Learning progress analysis submodule: Based on the learner behavioral characteristic data, the Long Short-Term Memory Network is used to analyze the learning completion status of each learner in the respiratory infectious disease prevention and control knowledge module. Combined with historical data, the current learning progress of the learners is evaluated, the current learning status of the learners is calculated, and the learner's learning progress data is obtained. Weakness identification submodule: Based on the learner's learning progress data, compare the learner's learning performance in different modules with the overall group performance, identify the learner's weak points, calculate the gap of the learner in specific prevention and control knowledge points, and obtain the learner's weak point data; The learning content recommendation submodule analyzes learners' knowledge gaps based on the data on their weak areas, recommends review content related to those gaps, and pushes advanced knowledge points to learners who are making rapid progress, generating personalized learning path recommendation data. 3.The Internet-based respiratory infectious disease prevention and control intelligent training system according to claim 1, characterized in that, The Long Short-Term Memory (LSTM) network uses the following formula:

4. wherein: represents the learning state of the learner at the current time step, represents the learning state of the learner at the previous time step, represents the input data at the current time step, represents the individual characteristic data of the learner, represents the difficulty coefficient of the current learning module, represents the weight matrix, represents the bias term, represents the activation function.

5. The Internet-based intelligent training system for the prevention and control of respiratory infectious diseases according to claim 1, characterized in that, The knowledge dissemination path optimization module includes: Interactive Behavior Analysis Submodule: Based on the interactive data of learners in the process of learning about respiratory infectious disease prevention and control, collect discussion records, Q&A interactions and information sharing frequency among learners, construct an interaction relationship diagram, mark the interactive participation of each learner, record the connection strength and interaction frequency between learners, and obtain learner interaction data; Influence Assessment Submodule: Based on the learner interaction data, calculate the interaction frequency and duration of each learner in the interaction relationship graph, identify high-influence learners by analyzing the learners' participation intensity and influence scope, mark them as key players in disseminating prevention and control knowledge, and obtain learner influence data; The dissemination path optimization submodule generates an optimized dissemination path by adjusting the dissemination order and prioritizing learners with high influence to disseminate key prevention and control knowledge, based on the learner influence data.

6. The Internet-based intelligent training system for the prevention and control of respiratory infectious diseases according to claim 1, characterized in that, The learning progress prediction and scheduling module includes: Progress prediction submodule: Based on the learner behavior characteristic data and personalized learning path recommendation data, analyze the learner's learning progress in each prevention and control knowledge module, predict the learner's future learning progress by comparing historical learning records with current behavior, and obtain learning progress prediction data. Bottleneck identification submodule: Based on the learning progress prediction data, it tracks the learner's learning progress in prevention and control knowledge, identifies learners with lagging progress, and locates learning bottlenecks and obtains learner bottleneck data based on the learner's learning performance. Progress scheduling submodule: Based on the learner bottleneck data, adjust the allocation of learning resources, prioritize providing appropriate review resources for learners with slow progress, and generate learning progress prediction data.

7. The Internet-based intelligent training system for the prevention and control of respiratory infectious diseases according to claim 1, characterized in that, The learning resource scheduling and content delivery module includes: Progress assessment submodule: Based on the learning progress prediction data and the transmission optimization path, the module tracks the learner's learning progress in respiratory infectious disease prevention and control knowledge, calculates the learner's learning completion rate in each module, analyzes the learner's learning status, determines whether there is a progress lag, and obtains learning progress assessment data. Content Priority Adjustment Submodule: Based on the learning progress assessment data, a graph neural network is used to identify modules in which learners are weak in learning prevention and control knowledge. By analyzing the learners' weaknesses and combining them with the learning progress, the priority of pushing learning content is adjusted, prioritizing the push of review modules and modules with lower difficulty, and generating content push priority data. The push plan generation submodule: Based on the content push priority data, it adopts the SIR propagation model, combines the learners' learning needs, dynamically adjusts the push order, frequency and difficulty of learning content, and calculates the propagation speed of learning content by simulating the propagation process of learning content. Based on the propagation speed, it generates a suitable push plan and generates a learning content push plan.

8. The Internet-based intelligent training system for the prevention and control of respiratory infectious diseases according to claim 1, characterized in that, The graph neural network uses the following formula: in: Indicates the first In the next iteration, the learning content The hidden state, Indicates the first In the next iteration, the learning content The hidden state, Representation and learning content The set of connected adjacent nodes. Indicates learning content The weight matrix, Indicates learning content The adjacent module weight matrix, For activation function, These are weighting coefficients. For learning content The learning difficulty level Learning content Learning performance value, Indicates the content being studied. Adjacent units In the Hidden state in the next iteration.

Images