Virus propagation prediction method based on neural network

Abstract

The invention relates to the technical field of data processing, in particular to a virus propagation prediction method based on a neural network, and the method comprises the steps: collecting real-time data; inputting a model prediction threshold value; determining an abnormal propagation event; determining a propagation event type; judging the virus transmission risk; determining a risk level; and adjusting a propagation threshold value and performing early warning. According to the method, the real-time infection rate, the positive rate and the population flow change rate are combined, efficient analysis is performed by using the neural network model, and the virus propagation trend and risk can be accurately predicted. Through intelligent adjustment of the propagation threshold, changes of different epidemic situation development stages can be dynamically responded, abnormal propagation events can be recognized in advance, early warning can be given out in time, and public health departments can be helped to quickly recognize potential virus propagation risks and take effective control measures. The problems of low propagation dynamic change response speed and low prediction accuracy due to dependence on a traditional mathematical model and static parameter setting are effectively solved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a virus propagation prediction method based on a neural network. Background Art

[0002] With the acceleration of globalization, the speed and scope of virus transmission continue to expand, bringing unprecedented challenges to public health. Especially during virus outbreaks, quickly and accurately predicting the trends and risks of virus transmission is crucial for taking effective prevention and control measures in a timely manner. Traditional virus transmission prediction methods often rely on limited historical data and simple models, which are difficult to cope with complex and changeable real-life situations. Therefore, how to improve the accuracy and real-time nature of virus transmission prediction has become a major issue that needs to be solved urgently.

[0003] Patent document with publication number CN116741398A discloses a virus transmission prediction method and device, which includes: inputting the number of first viruses from the first infectious source in the current time period, the number of second viruses from the first infectious source in the previous time period, the number of third viruses from the second infectious source in the next time period, and the virus growth rate into a constructed sensor disease model to determine the virus impact factor; inputting the growth rate and the virus impact factor into a constructed boundary model to predict the actual transmission speed of the virus; wherein the sensor disease model is constructed based on a second-order integral difference equation of the time period, and the boundary model is constructed by the second-order integral difference equation after variable transformation based on preset conditions.

[0004] It can be seen that the virus transmission prediction method has the following problems: the method uses a second-order integral difference equation to construct a sensor disease model, which requires complex mathematical modeling and variable transformation, which makes the construction and tuning of the model relatively difficult, and the computational complexity is high, which is not conducive to real-time prediction; the method only relies on the constructed sensor disease model and boundary model. The construction of these models is affected by the environment, parameter selection and data quality. If the model is inaccurate, the prediction results will be affected, so the construction and verification of the model have high requirements. Summary of the invention

[0005] To this end, the present invention provides a virus propagation prediction method based on a neural network, which is used to overcome the problems of low response speed and prediction accuracy in reflecting dynamic changes in propagation in the prior art due to reliance on traditional mathematical models and static parameter settings through adaptive neural network model prediction and dynamic correction thresholds.

[0006] To achieve the above object, the present invention provides a virus propagation prediction method based on a neural network, comprising: Collect real-time infection rates of virus transmission, real-time positive rates of virus testing, and real-time change rates of population mobility from public health data sources; Inputting the real-time infection rate, the real-time positive rate, and the real-time change rate into a preset neural network prediction model to obtain a propagation threshold; Determine the occurrence of an abnormal transmission event according to the real-time infection rate, the real-time positive rate, and the transmission threshold, and form a first determination result; Determine a propagation type of the abnormal propagation event according to the first determination result and the real-time change rate; Determining the existence of a virus transmission risk according to the transmission type and the real-time positive rate, and forming a second determination result; Determine the severity level of the virus transmission risk according to the second determination result and the real-time change rate to form a risk level; Adjust the propagation threshold according to the number of times the risk level is formed within a preset adjustment period to form an adjusted propagation threshold; An early warning is issued for the virus propagation event corresponding to the risk level determined based on the adjusted propagation threshold.

[0007] Further, determining that an abnormal transmission event occurs according to the real-time infection rate, the real-time positive rate, and the transmission threshold, and forming a first determination result, includes: Calculate the standard deviation of the real-time infection rate within a preset first determination time period to form an infection rate fluctuation value; Calculate the standard deviation of the real-time positive rate within the preset first determination time to form a positive fluctuation value; The occurrence of an abnormal propagation event is determined according to the infection rate fluctuation value, the positive fluctuation value and the propagation threshold to form a first determination result.

[0008] Further, judging that an abnormal transmission event occurs according to the infection rate fluctuation value, the positive fluctuation value, and the transmission threshold, and forming a first judgment result includes: Calculating the relative deviation between the infection rate fluctuation value and the positive fluctuation value to form a fluctuation deviation value; Draw a change curve of the infection rate fluctuation value within the preset first determination time to form an infection fluctuation curve; Draw a change curve of the positive fluctuation value within the preset first determination time to form a positive fluctuation curve; Calculating the cosine similarity of the infection fluctuation curve and the positive fluctuation curve to form synchronization; The occurrence of an abnormal propagation event is determined according to the fluctuation deviation value, the synchronization degree and the propagation threshold to form a first determination result.

[0009] Further, determining the occurrence of an abnormal propagation event according to the fluctuation deviation value, the synchronization degree, and the propagation threshold includes: Calculating the relative deviation between the fluctuation deviation value and the synchronization degree to form a propagation consistency; When the propagation consistency is greater than the propagation threshold, it is determined that an abnormal propagation event occurs.

[0010] Further, determining the propagation type of the abnormal propagation event according to the first determination result and the real-time change rate includes: Calculate the standard deviation of the real-time change rate within the preset type determination time period to form a first change fluctuation value; Calculating the standard deviation of the number of first determination results formed within the preset type determination time period to form a first determination fluctuation value; The propagation type of the abnormal propagation event is determined according to the first change fluctuation value and the first determination fluctuation value.

[0011] Further, determining the propagation type of the abnormal propagation event according to the first change fluctuation value and the first determination fluctuation value includes: When the first change fluctuation value is greater than a preset change fluctuation threshold, and the first determination fluctuation value is greater than a preset determination fluctuation threshold, it is determined that the propagation type of the abnormal propagation event is large-scale propagation.

[0012] Further, judging that there is a risk of virus transmission according to the transmission type and the real-time positive rate, and forming a second judgment result, including: When it is determined that the propagation type of the abnormal propagation event is large-scale propagation, the change rate of the real-time positive rate per unit time within the preset second determination time length is calculated to form a positive change rate; Recording the duration of time when the positive change rate is greater than a preset positive change rate threshold to form a determination duration; Calculating the ratio of the determination duration to the preset second determination duration to form a duration ratio; When the continuous proportion is greater than the preset proportion threshold, it is determined that the virus transmission risk exists, forming a second determination result.

[0013] Further, determining the severity level of the virus transmission risk according to the second determination result and the real-time change rate to form a risk level includes: Calculating the standard deviation of the number of the second determination results formed within the preset level determination time to form a second determination fluctuation value; Calculating the standard deviation of the real-time change rate within the preset level determination time period to form a second change fluctuation value; The severity level is determined according to the second determination fluctuation value and the second change fluctuation value to form a risk level.

[0014] Further, determining the severity level according to the second determination fluctuation value and the second change fluctuation value to form a risk level includes: Calculating a relative deviation between the second determination fluctuation value and the second change fluctuation value to form a propagation index; When the propagation index is greater than a maximum value of a preset index range, the severity level is determined to form a risk level.

[0015] Further, adjusting the propagation threshold according to the number of times the risk level is formed within a preset adjustment period to form an adjusted propagation threshold includes: Calculate the standard deviation of the number of times the risk level is formed and the fluctuation value of the number of times formed; When the number fluctuation value is greater than a preset number fluctuation threshold, the propagation threshold is increased according to a relative deviation between the number fluctuation value and the preset number fluctuation threshold and a preset adjustment coefficient to form an adjusted propagation threshold.

[0016] Compared with the prior art, the beneficial effect of the present invention is that by combining the real-time infection rate, positive rate and population mobility change rate, the neural network model is used for efficient analysis, which can accurately predict the trend and risk of virus transmission. By intelligently adjusting the transmission threshold, it can dynamically respond to changes in different stages of epidemic development, identify abnormal transmission events in advance, and issue early warnings in a timely manner, which can help public health departments quickly identify potential virus transmission risks and take effective control measures, improve the efficiency and accuracy of epidemic prevention and control, and effectively solve the problem of low response speed and prediction accuracy of reflecting dynamic changes in transmission due to reliance on traditional mathematical models and static parameter settings.

[0017] Furthermore, by combining the fluctuations in the infection rate and the positivity rate, we can accurately assess abnormal fluctuations in virus transmission, discover potential transmission anomalies in advance, trigger early warnings in a timely manner, and help relevant departments quickly take effective prevention and control measures, thereby improving the sensitivity and response speed of epidemic prevention and control.

[0018] Furthermore, by quantifying the fluctuation deviation and synchronization, combined with the changing trend of historical data, the accuracy of identifying abnormal events of virus transmission can be improved. By introducing synchronization, the correlation between the infection rate and the positive rate fluctuation can be better judged, and potential abnormal transmission events can be effectively predicted and warned.

[0019] Furthermore, by introducing the concept of transmission consistency, the correlation between infection rate and positive rate fluctuations can be measured more accurately, thereby effectively identifying the occurrence of abnormal transmission events. By comparing the transmission consistency with the transmission threshold, the reliability and timeliness of abnormal transmission event determination can be improved, providing timely and accurate warning information for public health emergency response.

[0020] Furthermore, by calculating the rate of change and the fluctuation value of the judgment results, the dynamic changes in the transmission process can be accurately captured, which helps to distinguish different types of transmission events, and can more accurately classify abnormal transmission events, providing more targeted information for public health decision-making and improving the accuracy and efficiency of emergency response.

[0021] Furthermore, by setting appropriate fluctuation thresholds, it is possible to capture potential signs of rapid expansion of the virus in the early stages of transmission through accurate calculation of fluctuation values. This can not only accurately distinguish situations of large-scale transmission, but also avoid misjudgments and missed judgments, thereby ensuring the accuracy of the early warning system.

[0022] Furthermore, by introducing the calculation of the positive change rate and the duration of determination, the dynamic changes of virus transmission can be evaluated more accurately, thus avoiding misjudgment based on a single data point. By calculating the continuous proportion and judging the threshold, the risk of large-scale transmission can be identified in a timely manner, and measures can be taken in advance to effectively reduce the speed and scope of virus transmission, thereby ensuring public health safety.

[0023] Furthermore, by introducing standard deviation analysis of fluctuation values ​​and change rates, the determination of risk levels is made more dynamic and sensitive, and can more accurately reflect changes in transmission risks. Compared with simple static threshold determination, it has stronger adaptability in complex environments and can warn of abnormal situations in advance, thus providing effective support for the formulation of prevention and control measures.

[0024] Furthermore, through the calculation of the transmission index, the changes in transmission fluctuations are effectively integrated, which can accurately identify the severity of changes in the epidemic, reduce the limitations of single indicator judgment, and improve the accuracy of risk assessment and response capabilities. It not only improves the accuracy of risk judgment, but also can provide timely warnings, thereby providing strong data support for epidemic prevention and control.

[0025] Furthermore, according to the fluctuation of risk level, the transmission threshold is adjusted in real time to avoid overly rigid threshold settings, thereby enhancing the adaptability and accuracy of the early warning system. By dynamically adjusting the transmission threshold, the system's ability to respond to sudden changes in the epidemic situation is improved, which helps to identify and prevent potential virus transmission risks in advance. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a flowchart of the virus propagation prediction method based on neural network in this embodiment; Figure 2 This is a decision logic diagram for determining the occurrence of an abnormal propagation event in this embodiment; Figure 3 A decision logic diagram for determining the propagation type for this embodiment; Figure 4This is a decision logic diagram for determining the virus transmission risk in this embodiment. DETAILED DESCRIPTION

[0027] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0028] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0029] See also Figure 1 As shown, it is a flow chart of the virus propagation prediction method based on neural network in this embodiment; This implementation provides a virus propagation prediction method based on a neural network, comprising: Collect real-time infection rates of virus transmission, real-time positive rates of virus testing, and real-time change rates of population mobility from public health data sources; Inputting the real-time infection rate, the real-time positive rate, and the real-time change rate into a preset neural network prediction model to obtain a propagation threshold; Determine the occurrence of an abnormal transmission event according to the real-time infection rate, the real-time positive rate, and the transmission threshold, and form a first determination result; Determine a propagation type of the abnormal propagation event according to the first determination result and the real-time change rate; Determining the existence of a virus transmission risk according to the transmission type and the real-time positive rate, and forming a second determination result; Determine the severity level of the virus transmission risk according to the second determination result and the real-time change rate to form a risk level; Adjust the propagation threshold according to the number of times the risk level is formed within a preset adjustment period to form an adjusted propagation threshold; An early warning is issued for the virus propagation event corresponding to the risk level determined based on the adjusted propagation threshold.

[0030] The public health data source in this embodiment refers to a multi-dimensional, real-time updated health data set used to monitor and predict the spread of the virus. Specifically, these data come from public health institutions, medical units and other related monitoring platforms at all levels, covering multiple aspects of information, such as real-time infection rate, virus detection positive rate, population mobility and environmental factors (such as temperature and humidity). This embodiment is a public health monitoring system in a large city, which obtains and integrates epidemic-related data through multiple channels. First, the local disease control center collects and updates the virus infection situation in various places in real time through the public health infectious disease monitoring system, including the number of confirmed cases and positive detection rate. At the same time, major hospitals and clinics in the city provide data on patient health status through the electronic health record system. Secondly, the environmental monitoring system provides real-time air quality data, temperature changes, humidity and other environmental factors that may affect the spread of the virus in the city. At the same time, social media platforms and mobile health applications can also provide citizens' health behavior data and health status dynamics, which can reflect citizens' self-reported symptoms, health trends and regional health problems.

[0031] The preset adjustment period refers to the time period selected when adjusting the transmission threshold of the virus transmission prediction model. It depends on the changing pattern of historical data and the real-time needs of risk assessment. It is usually set between 3 days and 2 weeks to ensure the adequacy of the data and the stability of the model. In this embodiment, it is set to 7 days, which can capture the changing trend of the epidemic within a week, help to adjust the prediction model in time, and improve the accuracy of virus transmission risk prediction.

[0032] The preset neural network prediction model is designed to predict the risk of virus transmission. The model predicts the risk of the epidemic by learning a large amount of historical data, and is built based on time series data (such as infection rate, positive rate, etc.), and finally generates a deep learning model that can output prediction results based on real-time data. The following is a detailed description of the neural network model.

[0033] 1. Design ideas of the model: This preset neural network prediction model uses a long short-term memory network (LSTM), which is a special type of recurrent neural network (RNN) that is particularly suitable for processing data with time dependencies. In the prediction of virus transmission risk, time series data is crucial because the transmission trend often depends on the historical infection situation, epidemic changes, and the implementation of prevention and control measures. LSTM can capture this long-term dependency, making it an ideal choice.

[0034] 2 Initial model: Long short-term memory network (LSTM): 2.1LSTM structure: The LSTM model mainly consists of the following parts: Input layer: receives raw input data (time series data).

[0035] LSTM layer: processes time series data through forget gate, input gate and output gate.

[0036] Fully connected layer: maps the output of the LSTM layer to the prediction result.

[0037] Output layer: Output propagation threshold or other predicted values.

[0038] 2.2 Core parameters of LSTM: Each layer of an LSTM contains multiple parameters that determine how to process input data, how to maintain long-term memory, and how to output predictions.

[0039] Dimensions of input data: "batch_size": The number of data samples in each training batch.

[0040] "time_steps": The number of time steps for each time series.

[0041] "input_dim": the number of data features at each time step (such as infection rate, positive rate, etc.).

[0042] Weights and biases of LSTM layer: The LSTM layer contains four gates: input gate, forget gate, output gate, and candidate memory unit. Each gate has a set of weights and biases: Weight matrix (W): includes input gate weight (W_i), forget gate weight (W_f), output gate weight (W_o) and candidate memory unit weight (W_c). The size of each matrix is ​​"(input_dim, hidden_units)", where "hidden_units" is the number of hidden units in the LSTM layer.

[0043] Bias vector (b): Each gate has a bias term, typically of size (hidden_units,) .

[0044] Activation function: LSTM uses different activation functions to control the flow of information: Forget gate (f): Sigmoid function ("sigmoid"), whose output is between [0, 1], indicating the degree of forgetting of information in the memory unit.

[0045] Input gate (i): Sigmoid function that determines the degree of adoption of input information at the current time step.

[0046] Candidate memory cell (c'): tanh function ("tanh") used to generate candidate memories.

[0047] Output gate (o): Sigmoid function that determines the hidden state of the current time step.

[0048] 2.3 Hidden layer and output layer: Hidden state (h_t): LSTM iteratively updates the hidden state through time steps. The hidden state stores the long-term memory of the time series.

[0049] Memory cell (C_t): The memory cell of LSTM is also updated as the time step progresses, and it stores long-term information in the sequence.

[0050] Output layer: Finally, the output of the LSTM layer is passed to a fully connected layer or output layer to generate a prediction result.

[0051] 3. Training process: 3.1 Input data preparation: During training, the LSTM model uses historical data (such as past infection rates, positivity rates, etc.) as input. These data are usually time series data, and each data point contains information for one time step.

[0052] Training data: The input data set is formed by historical infection rate, positive rate, control measures and other data. The dimension of the input data is "(batch_size, time_steps, input_dim)".

[0053] Label data: The output of the model is the prediction result of the virus transmission risk, that is, the predicted value at a certain point in time. The label data is usually the infection probability, positive rate and other information at the target time step.

[0054] 3.2 Model training: Loss function: Use a suitable loss function (such as mean squared error MSE, cross entropy loss, etc.) to measure the difference between the predicted results and the actual labels.

[0055] Optimizer: Adam optimizer is usually used to optimize the parameters of the network and adjust the learning rate to accelerate training convergence.

[0056] Training process: Through the back propagation through time (BPTT), the LSTM model updates the weights and biases of each gate. The goal of training is to iteratively update the weights so that the model can predict the risk of virus transmission as accurately as possible.

[0057] 3.3 Preset training duration: The training time of the model is determined by two factors: The size of historical data: During training, LSTM needs to learn enough historical data to recognize time series patterns, so the size of the data set directly affects the training time.

[0058] Model convergence: During the training process, as the loss function decreases, the model gradually converges, and the training time is also affected by the model convergence speed.

[0059] 4. Trained model: After training, the model will get a set of final weights and biases, which represent the temporal patterns learned by the LSTM network. The final model outputs the propagation threshold based on the input time series data.

[0060] Trained model parameters: The weight matrix of the LSTM layer: the weights of the input gate, forget gate, output gate, and candidate memory cells.

[0061] Bias term of LSTM layer: bias term of each gate.

[0062] Fully connected layer weights and biases: weights and biases that map the LSTM output to the final prediction result.

[0063] 5. Use the trained model to output the propagation threshold: The trained LSTM model can be used to predict the risk of virus transmission (such as the transmission threshold) using the following steps: Input new data: In a real-time environment, input new historical data (such as new infection rate, positivity rate, etc.) into the trained LSTM model.

[0064] Model reasoning: The LSTM model predicts future risk values ​​based on historical data.

[0065] Output transmission threshold: The model outputs the predicted value to obtain the transmission threshold, which is used to assess the risk of future epidemic transmission.

[0066] The preset neural network prediction model of this embodiment is based on the LSTM network. By processing the time series information in the historical data, a deep learning model that can predict the risk of virus transmission (such as the transmission threshold) is finally obtained. The core parameters of the model include the weight matrix and bias term of the LSTM layer, as well as the optimizer and loss function used in training. Through training on historical data, the model can output accurate transmission thresholds to assist in epidemic prevention and control decisions.

[0067] By collecting the real-time infection rate, positive rate and population mobility change rate from the public health data source, the preset neural network prediction model is input for analysis to obtain the transmission threshold. Based on these real-time data and transmission thresholds, it is determined whether an abnormal transmission event has occurred, and the type of transmission is further determined. The virus transmission risk is determined by the real-time positive rate, and the risk severity level is evaluated in combination with the real-time change rate. According to the frequency of risk level changes within the preset adjustment period, the transmission threshold is adjusted, and an early warning is issued for the corresponding transmission event so that timely prevention and control measures can be taken.

[0068] By combining the real-time infection rate, positive rate and population mobility change rate, and using the neural network model for efficient analysis, the virus transmission trend and risk can be accurately predicted. By intelligently adjusting the transmission threshold, it can dynamically respond to changes in different stages of the epidemic development, identify abnormal transmission events in advance, and issue early warnings in a timely manner, which can help public health departments quickly identify potential virus transmission risks and take effective control measures, improve the efficiency and accuracy of epidemic prevention and control, and effectively solve the problem of low response speed and prediction accuracy in reflecting dynamic changes in transmission due to reliance on traditional mathematical models and static parameter settings.

[0069] Specifically, determining that an abnormal transmission event occurs according to the real-time infection rate, the real-time positive rate, and the transmission threshold, and forming a first determination result includes: Calculate the standard deviation of the real-time infection rate within a preset first determination time period to form an infection rate fluctuation value; Calculate the standard deviation of the real-time positive rate within the preset first determination time to form a positive fluctuation value; The occurrence of an abnormal propagation event is determined according to the infection rate fluctuation value, the positive fluctuation value and the propagation threshold to form a first determination result.

[0070] The preset first judgment time is a time window for calculating the real-time infection rate and positive rate fluctuations, and its length depends on the characteristics of virus transmission and the frequency of data collection. It is usually set between 15 hours and 3 days to ensure sufficient observation time for the fluctuation of virus transmission. In this embodiment, it is set to 24 hours to ensure that the trend of daily infection and detection changes can be fully reflected, which helps to identify abnormal transmission patterns in advance.

[0071] By calculating the standard deviation of the real-time infection rate and positive rate within the preset first judgment period, the infection rate fluctuation value and the positive fluctuation value are obtained, which reflect the fluctuation range of these two indicators. Then, according to these two fluctuation values ​​and the propagation threshold, the preset judgment logic is used to determine whether an abnormal propagation event occurs, and the first judgment result is formed.

[0072] By combining the fluctuations in the infection rate and positivity rate, we can accurately assess abnormal fluctuations in virus transmission, discover potential transmission anomalies in advance, trigger early warnings in a timely manner, and help relevant departments quickly take effective prevention and control measures, thereby improving the sensitivity and response speed of epidemic prevention and control.

[0073] Specifically, determining that an abnormal transmission event occurs according to the infection rate fluctuation value, the positive fluctuation value, and the transmission threshold, and forming a first determination result includes: Calculating the relative deviation between the infection rate fluctuation value and the positive fluctuation value to form a fluctuation deviation value; Draw a change curve of the infection rate fluctuation value within the preset first determination time to form an infection fluctuation curve; Draw a change curve of the positive fluctuation value within the preset first determination time to form a positive fluctuation curve; Calculating the cosine similarity of the infection fluctuation curve and the positive fluctuation curve to form synchronization; The occurrence of an abnormal propagation event is determined according to the fluctuation deviation value, the synchronization degree and the propagation threshold to form a first determination result.

[0074] First, the fluctuation values ​​of the infection rate and the positive rate are calculated, and the relative deviation between the two is analyzed by standard deviation to obtain the fluctuation deviation value. Then, the infection fluctuation curve and the positive fluctuation curve within the preset judgment time are drawn, and the cosine similarity of the two curves is further calculated to obtain the synchronization degree. Based on the comprehensive analysis of the fluctuation deviation value, synchronization degree and propagation threshold, it is judged whether an abnormal propagation event has occurred to form the first judgment result.

[0075] By quantifying fluctuation deviation and synchronization, combined with the changing trend of historical data, the accuracy of identifying abnormal events of virus transmission can be improved. By introducing synchronization, the correlation between infection rate and positive rate fluctuations can be better judged, and potential abnormal transmission events can be effectively predicted and warned.

[0076] Please continue reading Figure 2 As shown, it is a determination logic diagram for determining the occurrence of an abnormal propagation event in this embodiment; Determining the occurrence of an abnormal propagation event according to the fluctuation deviation value, the synchronization degree, and the propagation threshold includes: Calculating the relative deviation between the fluctuation deviation value and the synchronization degree to form a propagation consistency; When the propagation consistency is greater than the propagation threshold, it is determined that an abnormal propagation event occurs.

[0077] First, the relative deviation of the fluctuation deviation value and the synchronization degree is calculated to obtain the propagation consistency. Then the propagation consistency is compared with the preset propagation threshold. When the propagation consistency is greater than the propagation threshold, it is determined that an abnormal propagation event has occurred.

[0078] By introducing the concept of transmission consistency, the correlation between infection rate and positive rate fluctuations can be measured more accurately, thereby effectively identifying the occurrence of abnormal transmission events. By comparing transmission consistency with transmission threshold, the reliability and timeliness of abnormal transmission event determination can be improved, providing timely and accurate warning information for public health emergency response.

[0079] Specifically, determining the propagation type of the abnormal propagation event according to the first determination result and the real-time change rate includes: Calculate the standard deviation of the real-time change rate within the preset type determination time period to form a first change fluctuation value; Calculating the standard deviation of the number of first determination results formed within the preset type determination time period to form a first determination fluctuation value; The propagation type of the abnormal propagation event is determined according to the first change fluctuation value and the first determination fluctuation value.

[0080] The preset type determination time is used to determine the time range of the propagation type of the abnormal propagation event. It is set according to the speed of virus propagation and the frequency of data collection. It is usually set between 1 hour and 24 hours. In this embodiment, it is set to 6 hours. It can balance real-time and data stability, avoid affecting the judgment of the propagation type due to excessive fluctuations in a short period of time, and ensure sufficient data volume for accurate analysis.

[0081] According to the first judgment result and the real-time change rate, first calculate the standard deviation of the real-time change rate within the preset type determination time length to obtain the first change fluctuation value; then calculate the standard deviation of the number of first judgment results formed within the time length to obtain the first judgment fluctuation value; finally, determine the propagation type of the abnormal propagation event based on these two fluctuation values.

[0082] By calculating the rate of change and the fluctuation value of the judgment results, we can accurately capture the dynamic changes in the transmission process, help distinguish different types of transmission events, more accurately classify abnormal transmission events, provide more targeted information for public health decision-making, and improve the accuracy and efficiency of emergency response.

[0083] Please continue reading Figure 3 As shown, it is a decision logic diagram for determining the propagation type in this embodiment; Determining the propagation type of the abnormal propagation event according to the first change fluctuation value and the first determination fluctuation value includes: When the first change fluctuation value is greater than a preset change fluctuation threshold, and the first determination fluctuation value is greater than a preset determination fluctuation threshold, it is determined that the propagation type of the abnormal propagation event is large-scale propagation.

[0084] The preset change fluctuation threshold is used to determine the limit of the change rate fluctuation, which depends on the fluctuation characteristics of historical data and the sensitivity requirements for the spread of the epidemic. It is usually set between 0.02-0.1, and is set to 0.05 in this embodiment, which helps to accurately capture the fluctuation of the change rate, avoid ignoring subtle fluctuations that may cause large-scale spread due to excessively high thresholds, and can respond sensitively to sudden epidemics.

[0085] The preset judgment fluctuation threshold is used to determine the critical point of the fluctuation of the judgment result. It depends on the classification characteristics of the historical data and the risk level that needs to be responded to in advance. It is usually set between 0.05-0.3. In this embodiment, it is set to 0.1 to ensure that when an abnormal judgment is found, it can be marked and intervened in time to avoid the impact of delayed response.

[0086] First, calculate the fluctuation value of the real-time change rate within the preset type determination time length to obtain the first change fluctuation value; then, calculate the standard deviation of the number of first judgment results formed within the same time length to obtain the first judgment fluctuation value. Next, compare these two fluctuation values ​​with the preset change fluctuation threshold and judgment fluctuation threshold. If the first change fluctuation value is greater than the preset change fluctuation threshold, and the first judgment fluctuation value is also greater than the preset judgment fluctuation threshold, it is judged as large-scale transmission, and the epidemic situation with abnormal transmission and rapid expansion can be timely identified according to the change of the fluctuation value, and then classified and judged.

[0087] By setting appropriate fluctuation thresholds, we can capture potential signs of rapid expansion of the virus in the early stages of transmission through accurate calculation of fluctuation values. This can not only accurately distinguish situations of large-scale transmission, but also avoid misjudgments and missed judgments, thus ensuring the accuracy of the early warning system.

[0088] Please continue reading Figure 4 As shown, it is a decision logic diagram for determining the virus transmission risk in this embodiment; Determining the existence of a virus transmission risk according to the transmission type and the real-time positive rate, and forming a second determination result, including: When it is determined that the propagation type of the abnormal propagation event is large-scale propagation, the change rate of the real-time positive rate per unit time within the preset second determination time length is calculated to form a positive change rate; Recording the duration of time when the positive change rate is greater than a preset positive change rate threshold to form a determination duration; Calculating the ratio of the determination duration to the preset second determination duration to form a duration ratio; When the continuous proportion is greater than the preset proportion threshold, it is determined that the virus transmission risk exists, forming a second determination result.

[0089] The preset second judgment time is a time window used to measure the real-time positive rate change rate, which depends on the observation period of virus transmission and the speed of epidemic change. It is usually set between 24 hours and 72 hours. This embodiment is set to 48 hours, which can capture a more comprehensive virus transmission trend, balance the response speed and stability, and avoid unstable judgment results due to too short a time window.

[0090] The preset positive change rate threshold is used to determine whether the positive rate change is significant, which is related to the fluctuation range and spread speed of the epidemic, and is usually set between 0.5% and 2%. In this embodiment, it is set to 1%, which can effectively distinguish between normal fluctuations and potential transmission risks, and ensure timely response when the epidemic changes rapidly.

[0091] The preset percentage threshold is used to measure the ratio between the duration of the positive rate change and the total observation time. It depends on the urgency of virus transmission and the sensitivity of the response. It is usually set between 50% and 70%. In this embodiment, it is set to 60%, which can ensure that when the trend of the positive rate change continues for a long enough time, it can trigger the judgment of the virus transmission risk and avoid misjudgment due to short-term fluctuations.

[0092] First, according to the determination result that the propagation type of the abnormal propagation event is large-scale propagation, the real-time positive rate change rate per unit time is calculated within the preset second determination time, and the positive change rate is formed. Next, the duration during which the positive change rate is greater than the preset positive change rate threshold is recorded to form the determination duration. Then, the ratio of the determination duration to the preset second determination duration is calculated to obtain the continuous proportion. Finally, if the continuous proportion exceeds the preset proportion threshold, it is determined that there is a risk of virus transmission, and a second determination result is formed.

[0093] By introducing the calculation of the positive change rate and the duration of determination, the dynamic changes of virus transmission can be evaluated more accurately, thus avoiding misjudgment based on a single data point. By calculating the continuous proportion and judging the threshold, the risk of large-scale transmission can be identified in a timely manner, and measures can be taken in advance to effectively reduce the speed and scope of virus transmission, thereby ensuring public health safety.

[0094] Specifically, determining the severity level of the virus transmission risk according to the second determination result and the real-time change rate to form a risk level includes: Calculating the standard deviation of the number of the second determination results formed within the preset level determination time to form a second determination fluctuation value; Calculating the standard deviation of the real-time change rate within the preset level determination time period to form a second change fluctuation value; The severity level is determined according to the second determination fluctuation value and the second change fluctuation value to form a risk level.

[0095] The preset level determination time is the time window required to calculate the virus transmission risk level, which depends on the potential speed of virus transmission and the frequency of data collection. It is usually set to 24 to 72 hours. In this embodiment, it is set to 48 hours, which can ensure that it covers a sufficient time period to smooth data fluctuations, improve assessment accuracy, and reduce misjudgments caused by data fluctuations in a short period of time.

[0096] According to the second determination result and the real-time change rate, the standard deviation of the number of second determination results formed within the preset level determination time is first calculated to obtain the second determination fluctuation value. Then, the standard deviation of the real-time change rate within the same time range is calculated to form the second change fluctuation value. Then, based on these two fluctuation values, the severity level of the virus transmission risk is comprehensively evaluated and determined, and the risk level is finally output. This process helps to timely identify potential virus transmission risks by quantifying the size of the fluctuation and accurately evaluating the changes in risks.

[0097] By introducing standard deviation analysis of fluctuation values ​​and change rates, the determination of risk levels is made more dynamic and sensitive, and can more accurately reflect changes in transmission risks. Compared with simple static threshold determination, it has stronger adaptability in complex environments and can warn of abnormal situations in advance, thus providing effective support for the formulation of prevention and control measures.

[0098] Specifically, determining the severity level according to the second determination fluctuation value and the second change fluctuation value to form a risk level includes: Calculating a relative deviation between the second determination fluctuation value and the second change fluctuation value to form a propagation index; When the propagation index is greater than a maximum value of a preset index range, the severity level is determined to form a risk level.

[0099] The preset index range is a set interval used to define the range of variation of the transmission index, thereby determining the risk level of virus transmission. It depends on factors such as fluctuations in historical epidemic data, the speed of virus transmission, and public health policies. It is usually set between 0 and 1. The specific value can be adjusted based on the distribution of historical data. In this embodiment, it is set to between 0.5 and 0.8. It can be dynamically adjusted based on actual data fluctuations to ensure that risk assessment is neither too broad nor too sensitive, so as to achieve more accurate risk prediction and early warning effects.

[0100] First, the relative deviation between the second determination fluctuation value and the second change fluctuation value is calculated to obtain the propagation index. Then, the propagation index is compared with the preset index range. If the propagation index exceeds the maximum value of the preset range, the severity level of the virus propagation is determined based on this indicator, and the final risk level is formed.

[0101] Through the calculation of the transmission index, the changes in transmission fluctuations are effectively integrated, the severity of epidemic changes can be accurately identified, the limitations of single indicator judgment can be reduced, and the accuracy of risk assessment and response capabilities can be improved. It not only improves the accuracy of risk judgment, but also can provide timely warnings, thereby providing strong data support for epidemic prevention and control.

[0102] Specifically, adjusting the propagation threshold according to the number of times the risk level is formed within a preset adjustment period to form an adjusted propagation threshold includes: Calculate the standard deviation of the number of times the risk level is formed and the fluctuation value of the number of times formed; When the number fluctuation value is greater than the preset number fluctuation threshold, the propagation threshold is increased according to the relative deviation between the number fluctuation value and the preset number fluctuation threshold and the preset adjustment coefficient to form an adjusted propagation threshold, wherein the relative deviation between the number fluctuation value and the preset number fluctuation threshold is positively correlated with the adjusted propagation threshold.

[0103] The preset number of fluctuation thresholds is used to determine the tolerance range of fluctuations in the number of risk level formations. It depends on the fluctuations in historical data and the sensitivity to fluctuations. It is usually set between 2 or 3 times the standard deviation. In this embodiment, it is set to 2.5 times, which can more accurately detect situations where risk fluctuations are large, thereby adjusting the propagation threshold in time to avoid delays.

[0104] The preset adjustment coefficient is a multiple used to adjust the propagation threshold in the case of large fluctuations. It depends on the risk management strategy in actual applications and the sensitivity of the threshold adjustment. It is usually set between 1.2 and 2. In this embodiment, it is set to 1.5, which can balance the response to risk fluctuations and avoid unnecessary alarms caused by excessive adjustments.

[0105] According to the number of times the risk level is formed within the preset adjustment period, the standard deviation of the number of times is calculated to obtain the frequency fluctuation value. If the fluctuation value is greater than the preset frequency fluctuation threshold, the relative deviation between the frequency fluctuation value and the threshold is calculated, and then the propagation threshold is increased by combining the preset adjustment coefficient to form the adjusted propagation threshold. This adjustment process can help dynamically optimize the threshold according to the actual risk fluctuation, ensuring a more sensitive response mechanism.

[0106] According to the fluctuation of risk level, the transmission threshold is adjusted in real time to avoid overly rigid threshold settings, thereby enhancing the adaptability and accuracy of the early warning system. By dynamically adjusting the transmission threshold, the system's ability to respond to sudden changes in the epidemic situation is improved, which helps to identify and prevent potential virus transmission risks in advance.

[0107] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. A virus propagation prediction method based on neural network, characterized in that: include: Collect real-time infection rates of virus transmission, real-time positive rates of virus testing, and real-time change rates of population mobility from public health data sources; Inputting the real-time infection rate, the real-time positive rate, and the real-time change rate into a preset neural network prediction model to obtain a propagation threshold; Determine the occurrence of an abnormal transmission event according to the real-time infection rate, the real-time positive rate, and the transmission threshold, and form a first determination result; Determine a propagation type of the abnormal propagation event according to the first determination result and the real-time change rate; Determining the existence of a virus transmission risk according to the transmission type and the real-time positive rate, and forming a second determination result; Determine the severity level of the virus transmission risk according to the second determination result and the real-time change rate to form a risk level; Adjust the propagation threshold according to the number of times the risk level is formed within a preset adjustment period to form an adjusted propagation threshold; An early warning is issued for the virus propagation event corresponding to the risk level determined based on the adjusted propagation threshold.

2. The virus propagation prediction method based on neural network according to claim 1, characterized in that: Determining the occurrence of an abnormal transmission event according to the real-time infection rate, the real-time positive rate, and the transmission threshold, and forming a first determination result, including: Calculate the standard deviation of the real-time infection rate within a preset first determination time period to form an infection rate fluctuation value; Calculate the standard deviation of the real-time positive rate within the preset first determination time to form a positive fluctuation value; The occurrence of an abnormal propagation event is determined according to the infection rate fluctuation value, the positive fluctuation value and the propagation threshold to form a first determination result.

3. The virus propagation prediction method based on neural network according to claim 2 is characterized in that: Determining the occurrence of an abnormal transmission event according to the infection rate fluctuation value, the positive fluctuation value, and the transmission threshold, and forming a first determination result, including: Calculating the relative deviation between the infection rate fluctuation value and the positive fluctuation value to form a fluctuation deviation value; Draw a change curve of the infection rate fluctuation value within the preset first determination time to form an infection fluctuation curve; Draw a change curve of the positive fluctuation value within the preset first determination time to form a positive fluctuation curve; Calculating the cosine similarity of the infection fluctuation curve and the positive fluctuation curve to form synchronization; The occurrence of an abnormal propagation event is determined according to the fluctuation deviation value, the synchronization degree and the propagation threshold to form a first determination result.

4. The virus propagation prediction method based on neural network according to claim 3 is characterized in that: Determining the occurrence of an abnormal propagation event according to the fluctuation deviation value, the synchronization degree, and the propagation threshold includes: Calculating the relative deviation between the fluctuation deviation value and the synchronization degree to form a propagation consistency; When the propagation consistency is greater than the propagation threshold, it is determined that an abnormal propagation event occurs.

5. The virus propagation prediction method based on neural network according to claim 4 is characterized in that: Determining the propagation type of the abnormal propagation event according to the first determination result and the real-time change rate includes: Calculate the standard deviation of the real-time change rate within the preset type determination time period to form a first change fluctuation value; Calculating the standard deviation of the number of first determination results formed within the preset type determination time period to form a first determination fluctuation value; The propagation type of the abnormal propagation event is determined according to the first change fluctuation value and the first determination fluctuation value.

6. The method for predicting virus propagation based on a neural network according to claim 5, characterized in that: Determining the propagation type of the abnormal propagation event according to the first change fluctuation value and the first determination fluctuation value includes: When the first change fluctuation value is greater than a preset change fluctuation threshold, and the first determination fluctuation value is greater than a preset determination fluctuation threshold, it is determined that the propagation type of the abnormal propagation event is large-scale propagation.

7. The method for predicting virus propagation based on a neural network according to claim 6, characterized in that: Determining the existence of a virus transmission risk according to the transmission type and the real-time positive rate, and forming a second determination result, including: When it is determined that the propagation type of the abnormal propagation event is large-scale propagation, the change rate of the real-time positive rate per unit time within the preset second determination time length is calculated to form a positive change rate; Recording the duration of time when the positive change rate is greater than a preset positive change rate threshold to form a determination duration; Calculating the ratio of the determination duration to the preset second determination duration to form a duration ratio; When the continuous proportion is greater than the preset proportion threshold, it is determined that the virus transmission risk exists, forming a second determination result.

8. The method for predicting virus propagation based on a neural network according to claim 7, characterized in that: Determining the severity level of the virus transmission risk according to the second determination result and the real-time change rate to form a risk level includes: Calculating the standard deviation of the number of the second determination results formed within the preset level determination time to form a second determination fluctuation value; Calculating the standard deviation of the real-time change rate within the preset level determination time period to form a second change fluctuation value; The severity level is determined according to the second determination fluctuation value and the second change fluctuation value to form a risk level.

9. The method for predicting virus propagation based on a neural network according to claim 8, characterized in that: Determining the severity level according to the second determination fluctuation value and the second change fluctuation value to form a risk level includes: calculating a relative deviation between the second determination fluctuation value and the second change fluctuation value to form a propagation index; When the propagation index is greater than a maximum value of a preset index range, the severity level is determined to form a risk level.

10. The virus propagation prediction method based on neural network according to claim 9, characterized in that: Adjusting the propagation threshold according to the number of times the risk level is formed within a preset adjustment period to form an adjusted propagation threshold includes: Calculate the standard deviation of the number of times the risk level is formed and the fluctuation value of the number of times formed; When the number fluctuation value is greater than a preset number fluctuation threshold, the propagation threshold is increased according to a relative deviation between the number fluctuation value and the preset number fluctuation threshold and a preset adjustment coefficient to form an adjusted propagation threshold.

Images