A method and system for constructing an avian influenza occurrence risk prediction model based on an el nino index

Abstract

The application discloses a kind of based on the construction method and system of avian influenza occurrence risk prediction model of El Nino index, belong to risk prediction technical field, including respectively constructing H5 subtype avian influenza event data set and multiple El Nino index data set and climate-avian influenza data set;Regression model is constructed to avian influenza event number on climate index based on climate-avian influenza data set;Multiple deep learning models are constructed based on multiple El Nino index data set, and optimal prediction model is selected;Future avian influenza event number is predicted based on regression model and optimal prediction model.The application uses MEI as the core representative variable of El Nino climate factor, combines the historical epidemiological characteristics and time factor of avian influenza virus, realizes the dynamic prediction of avian influenza occurrence risk under future climate condition by quantitative analysis of the influence of climate change on highly pathogenic avian influenza transmission, and can accurately predict the high-risk space-time distribution of H5 subtype strain.

Description

Technical Field

[0001] This invention relates to the field of risk prediction technology, and more specifically to a method and system for constructing a risk prediction model for avian influenza based on the El Niño index. Background Technology

[0002] Avian influenza is a dangerous animal infectious disease that poses a threat to poultry farming, the ecological environment, and human health. The H5 subtype of avian influenza virus is one of the most concerning subtypes globally. This subtype encompasses multiple highly pathogenic and low-pathogenic strains, and its spread not only threatens the healthy development of livestock farming but also poses a significant challenge to global public health security. In recent years, the epidemic pattern of the H5 subtype of avian influenza virus has shown high complexity, especially in certain regions and seasons, where its transmission trajectory and outbreak frequency are closely related to climatic conditions.

[0003] Global climate change, particularly the climate anomalies triggered by El Niño, is considered a significant driver of avian influenza virus transmission and host behavior. As a global climate anomaly, El Niño's multifaceted impacts on the climate system and ecological environment have been extensively studied. Increasing evidence suggests that El Niño may play a crucial role in the outbreak and spread of avian influenza by influencing climate conditions such as temperature, humidity, and precipitation. Therefore, it is necessary to select an indicator that comprehensively reflects the intensity and duration of El Niño to quantify its impact.

[0004] El Niño significantly alters global meteorological conditions such as temperature, precipitation, and humidity. These climate changes can affect not only the survival and transmission capacity of the H5 subtype avian influenza virus in the environment but also indirectly drive viral transmission dynamics by altering migratory bird routes, poultry farming environments, and ecosystem balance. In particular, the highly pathogenic H5 subtype strain, due to its high pathogenicity, broad host spectrum, and sensitivity to environmental conditions, is more susceptible to the impact of El Niño-induced climate anomalies on its epidemic patterns. However, current research on H5 subtype avian influenza risk prediction based on the MEI index is still lacking, and there is an urgent need to develop more accurate prediction tools.

[0005] Therefore, how to provide a method and system for constructing a prediction model for the occurrence risk of avian influenza that can accurately predict the H5 subtype strain is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] In view of this, the present invention provides a method and system for constructing an avian influenza occurrence risk prediction model based on the El Niño index. Taking the H5 subtype avian influenza event as the research object, the MEI is used as the core representative variable of El Niño climate factors. Combined with the historical epidemiological characteristics and time factors of avian influenza virus, the impact of climate change on the spread of highly pathogenic avian influenza is quantitatively analyzed to achieve dynamic prediction of the risk of avian influenza occurrence under future climate conditions, and can accurately predict the high-risk spatiotemporal distribution of H5 subtype strains.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] A method for constructing an avian influenza outbreak risk prediction model based on the El Niño index includes the following steps:

[0009] H5 subtype avian influenza event dataset and multivariate El Niño index dataset were constructed respectively, and data processing was performed on the H5 subtype avian influenza event dataset and the multivariate El Niño index dataset to obtain climate-avian influenza dataset;

[0010] A regression model of avian influenza event numbers on climate indices was constructed based on the aforementioned climate-avian influenza dataset.

[0011] Multiple deep learning models are constructed based on the multivariate El Niño index dataset. The multiple deep learning models are evaluated to obtain the optimal prediction model for future climate change indices.

[0012] The number of future avian influenza events is predicted based on the regression model and the optimal prediction model of the future climate change index.

[0013] Preferably, a regression model of avian influenza event numbers on climate indices is constructed based on the climate-avian influenza dataset, including:

[0014] Based on the aforementioned climate-avian influenza dataset, several influencing factors affecting the number of avian influenza events were identified;

[0015] Multiple generalized additive models were constructed by randomly combining the aforementioned influencing factors as independent variables and using the number of avian influenza events as the dependent variable.

[0016] Calculate the AIC values ​​of the multiple generalized additive models according to the AIC criterion, and select the generalized additive model with the smallest AIC value as the optimal model;

[0017] The parameters of the optimal model were optimized using smoothing and approximation methods to obtain a regression model of avian influenza event numbers on climate indices.

[0018] Preferably, the multiple deep learning models constructed based on the multivariate El Niño index dataset include: multilayer perceptron, convolutional neural network, recurrent neural network, long short-term memory network, and convolutional neural network-long short-term memory network deep learning model.

[0019] Preferably, the evaluation metrics for evaluating the plurality of deep learning models include: mean absolute error, mean absolute percentage error, and root mean square error.

[0020] Preferably, predicting the number of future avian influenza events based on the regression model and the optimal prediction model of the future climate change index includes:

[0021] Predict climate change in future time periods based on the optimal prediction model of the future climate change index;

[0022] By incorporating future climate change as an independent variable into the regression model, the predicted number of avian influenza events in the future time period is obtained.

[0023] On the other hand, the present invention provides a system for constructing a prediction model for the occurrence risk of avian influenza based on the El Niño index, comprising:

[0024] The dataset construction module is used to construct the H5 subtype avian influenza event dataset and the multivariate El Niño index dataset respectively, and to process the H5 subtype avian influenza event dataset and the multivariate El Niño index dataset to obtain the climate-avian influenza dataset.

[0025] The regression model building module is used to build a regression model of the number of avian influenza events on the climate index based on the climate-avian influenza dataset.

[0026] The prediction model building module is used to build multiple deep learning models based on the multivariate El Niño index dataset, evaluate the multiple deep learning models, and obtain the optimal prediction model for future climate change indices.

[0027] The prediction module is used to predict the number of future avian influenza events based on the regression model and the optimal prediction model of the future climate change index.

[0028] As can be seen from the above technical solution, compared with the prior art, this invention discloses a method and system for constructing an avian influenza outbreak risk prediction model based on the El Niño index. By integrating the H5 subtype avian influenza event dataset and the multivariate El Niño index dataset, this invention can comprehensively consider the impact of climate change on avian influenza events, thereby constructing a more accurate prediction model. In particular, by constructing and evaluating multiple deep learning models and selecting the optimal model for predicting future climate change indices, the accuracy of predicting the number of future avian influenza events is further improved. When constructing the regression model of the number of avian influenza events on the climate index, this invention adopts a generalized additive model and enhances the robustness and adaptability of the model by randomly combining influencing factors and selecting the optimal model. This method can cope with complex and variable climate conditions and improve the model's predictive performance under different environments. By introducing deep learning technologies, such as multilayer perceptrons, convolutional neural networks, and recurrent neural networks, this invention can efficiently process and analyze large amounts of climate data and avian influenza event data, thereby quickly constructing a prediction model. This greatly improves the efficiency of prediction, enabling the model to respond to climate change more quickly and provide timely and effective support for avian influenza prevention and control. Attached Figure Description

[0029] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0030] Figure 1 A flowchart of the model construction method provided for this invention;

[0031] Figure 2 For statistical modeling workflow;

[0032] Figure 3 The MEI value for the next 3 months is predicted using an RNN model;

[0033] Figure 4 To use a statistical regression model to predict the number of highly pathogenic avian influenza events in the next three months;

[0034] Figure 5 A framework diagram of the model building system is provided for this invention. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] This invention discloses a method for constructing a predictive model for the occurrence risk of avian influenza based on the El Niño index, such as... Figure 1 As shown, it includes the following steps:

[0037] H5 subtype avian influenza event dataset and multivariate El Niño index dataset were constructed separately, and data processing was performed on the H5 subtype avian influenza event dataset and multivariate El Niño index dataset to obtain climate-avian influenza dataset.

[0038] Specifically, the constructed global H5 subtype avian influenza dataset includes the location (continent, country, city) and discovery date (accurate to year, month, and day) of each avian influenza event. A multivariate El Niño dataset, represented by the El Niño index (MEI), is then constructed, with each value corresponding to the average of the preceding two months. R language version 4.3.2 is used to clean and filter invalid data, and the dates of the two datasets are aligned and unified.

[0039] The specific steps for constructing the H5 subtype avian influenza event dataset are as follows: Filter the original data frame, keeping only the "Animal.type", "Region", and "Observation.date..dd.mm.yyyy." columns. In the "Animal.type" column, keep only the Domestic and Wild classes. Then, change the date in the "Observation.date..dd.mm.yyyy." column from a day-to-month format and delete blank data.

[0040] The specific steps to construct the MEI dataset are as follows: Use the pivot_longer function in the tidyr package (version 1.3.1) to convert the original data frame from wide format to long format, and at the same time modify the date names, changing the column names such as DJ (the average of December and January), JF, FM...ON, ND, DJ, etc. to the full name of the month represented by the first letter.

[0041] The month and year were merged to match the date column format of the H5 subtype avian influenza event dataset. Finally, the dates of the two datasets were aligned and merged into a new dataset for backup.

[0042] A regression model of avian influenza event numbers on climate indices was constructed based on the climate-avian influenza dataset.

[0043] Multiple deep learning models were constructed based on a multivariate El Niño index dataset. These models were then evaluated to obtain the optimal prediction model for future climate change indices.

[0044] The number of future avian influenza events is predicted based on regression models and the optimal prediction model of the future climate change index.

[0045] Furthermore, based on the distribution of avian influenza events, and by comparing the effects of various statistical models (such as Poisson regression and pseudo-Poisson regression models, negative binomial generalized linear models, and generalized additive models), the Generalized Additive Models (GAM) were ultimately selected as the main model framework (expressed as Y = β0 + f1(X1) + f2(X2) + ... + f p (X p Based on the climate-avian influenza dataset, a regression model of avian influenza event numbers on climate indices is constructed, including:

[0046] Multiple factors influencing the number of avian influenza events were identified based on the climate-avian influenza dataset;

[0047] Multiple generalized additive models were constructed by randomly combining multiple influencing factors as independent variables and using the number of avian influenza events as the dependent variable.

[0048] Calculate the AIC values ​​of multiple generalized additive models according to the AIC criterion, and select the generalized additive model with the smallest AIC value as the optimal model;

[0049] The optimal model's parameters are optimized using smoothing and approximation methods (see the detailed process). Figure 2 The method of comparison was used to improve the prediction accuracy and generalization of the model, and a regression model of avian influenza event numbers on climate indices was obtained.

[0050] Furthermore, several deep learning models built based on the multivariate El Niño index dataset include: multilayer perceptron, convolutional neural network, recurrent neural network, long short-term memory network, and convolutional neural network-long short-term memory network deep learning models. These models are implemented using the TensorFlow framework in Python 3.10.14 to predict the MEI index.

[0051] First, the cleaned time-series data is converted into a supervised learning format to be used as training data for input features (X) and output labels (Y). By constructing historical and future time-series data, the model's input features and output labels are effectively separated, supporting univariate, multivariate, and single-step or multi-step prediction tasks. It automatically handles time-step misalignment and missing values, ensuring data integrity and is suitable for common deep learning models, providing standardized data input for time-series analysis models. Then, a sliding window is defined; input features and output labels are separated to facilitate model training; the training and validation sets are divided in an 8:2 ratio to ensure the model's generalization ability; and the data is normalized to enhance the stability of model training.

[0052] Deep learning models MLP, CNN, RNN, LSTM, and CNN-LSTM were constructed respectively. Each model was trained 200 times, and the prediction time step nout was set to 3.

[0053] MLP uses fully connected (Dense) layers as its main components. The first layer has 100 neurons and uses the ReLU activation function to introduce non-linearity. The input dimension is the product of the time step and the number of features. The second layer has 60 neurons and no activation function is set, defaulting to linear activation. The output layer has 3 neurons.

[0054] The CNN model first needs to adjust the shape of the input data to fit the input requirements; the one-dimensional convolutional layer is used to extract local patterns in the time series, using 64 convolutional kernels with a kernel size of 2, and using the ReLU activation function to increase the model's non-linear expressive power; the max pooling layer is used for dimensionality reduction and extraction of key features, with a pooling window size of 2; the flattening layer converts the pooled two-dimensional feature map into a one-dimensional vector for input to the output layer; the first layer of the output layer has 50 neurons, and the second layer has 3 neurons.

[0055] The RNN's input layer has 50 neurons, and the output layer has 3 neurons.

[0056] The first and second layers of the LSTM each have 50 LSTM units, and the output layer has 3 LSTM units.

[0057] The CNN-LSTM model first reshapes the data into a four-dimensional array, dividing the time series data into two sub-sequences and processing each sub-sequence independently. It then extracts local pattern features of the sub-sequences through the convolutional layers of the CNN, setting 64 convolutional kernels of size 1. The features are then reduced in dimensionality through pooling layers and flattened into one-dimensional vectors. The flattened sub-sequence features are input into an LSTM layer containing 50 units, and the output layer has 3 neurons.

[0058] By comparing the model evaluation metrics (loss function, set to Mean Square Error, MSE in this invention) on the validation set, the optimal model was determined to be an RNN. Next, an RNN model for time series prediction was constructed. First, the training, validation, and test sets were divided in an 8:1:1 ratio. A 50-unit RNN layer was added to the model, followed by an output layer with 3 neurons, which generated the final prediction result. During model compilation, MSE was selected as the loss function, and the Adam optimizer was used to adjust the model parameters. Metrics such as Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Root Mean Square Error (RMSE) were used to evaluate the model's performance. During training, the model received input data strictly in chronological order without shuffling the data to maintain the integrity of the time series. The training parameters were set to 200 iterations, processing 32 data points per iteration. By monitoring the model's performance in real-time on the validation set, it was ensured that the model could fully capture the feature changes in the time series and achieve reasonable predictions of the MEI values ​​for the next three months.

[0059] like Figure 3 As shown, the deep learning model RNN predicts the MEI value for the next three months with good results; the predicted values ​​for the next three months are: 0.9 for December 2023, 0.9 for January 2024, and 0.8 for February 2024.

[0060] Furthermore, the evaluation metrics for evaluating multiple deep learning models include: mean absolute error, mean absolute percentage error, and root mean square error.

[0061] Furthermore, based on regression models and optimal prediction models of future climate change indices, the number of future avian influenza events is predicted, including:

[0062] Predicting climate change in future time periods based on the optimal prediction model of the future climate change index;

[0063] By incorporating future climate change as an independent variable into the regression model, the predicted number of avian influenza events in the future period can be obtained.

[0064] like Figure 4 As shown, the 3-month MEI prediction values ​​were put into the constructed and optimized statistical regression model to predict the number of highly pathogenic avian influenza events in the next 3 months. The corresponding 3-month prediction values ​​were 14.520938, 12.017535, and 9.857997.

[0065] On the other hand, this invention provides a system for constructing a prediction model for the risk of avian influenza based on the El Niño index, such as...Figure 5 As shown, it includes:

[0066] The dataset construction module is used to construct the H5 subtype avian influenza event dataset and the multivariate El Niño index dataset respectively, and to process the H5 subtype avian influenza event dataset and the multivariate El Niño index dataset to obtain the climate-avian influenza dataset.

[0067] The regression model building module is used to build a regression model of the number of avian influenza events on the climate index based on the climate-avian influenza dataset.

[0068] The prediction model building module is used to build multiple deep learning models based on the multivariate El Niño index dataset, evaluate the multiple deep learning models, and obtain the optimal prediction model for future climate change indices.

[0069] The prediction module is used to predict the number of future avian influenza events based on regression models and the optimal prediction model of the future climate change index.

[0070] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0071] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for constructing a predictive model for the occurrence risk of avian influenza based on the El Niño index, characterized in that, Includes the following steps: H5 subtype avian influenza event dataset and multivariate El Niño index dataset were constructed respectively, and data processing was performed on the H5 subtype avian influenza event dataset and the multivariate El Niño index dataset to obtain climate-avian influenza dataset; A regression model of avian influenza event numbers on climate indices was constructed based on the aforementioned climate-avian influenza dataset. Multiple deep learning models are constructed based on the multivariate El Niño index dataset. The multiple deep learning models are evaluated to obtain the optimal prediction model for future climate change indices. The number of future avian influenza events is predicted based on the regression model and the optimal prediction model of the future climate change index, including: Predict climate change in future time periods based on the optimal prediction model of the future climate change index; By incorporating future climate change as an independent variable into the regression model, the predicted number of avian influenza events in the future time period is obtained.

2. The method for constructing an avian influenza outbreak risk prediction model based on the El Niño index according to claim 1, characterized in that, Based on the aforementioned climate-avian influenza dataset, a regression model of avian influenza event numbers on climate indices is constructed, including: Based on the aforementioned climate-avian influenza dataset, several influencing factors affecting the number of avian influenza events were identified; Multiple generalized additive models were constructed by randomly combining the aforementioned influencing factors as independent variables and using the number of avian influenza events as the dependent variable. Calculate the AIC values ​​of the multiple generalized additive models according to the AIC criterion, and select the generalized additive model with the smallest AIC value as the optimal model; The parameters of the optimal model were optimized using smoothing and approximation methods to obtain a regression model of avian influenza event numbers on climate indices.

3. The method for constructing a prediction model for the risk of avian influenza based on the El Niño index according to claim 1, characterized in that, Several deep learning models built based on the aforementioned multivariate El Niño index dataset include: multilayer perceptron, convolutional neural network, recurrent neural network, long short-term memory network, and convolutional neural network-long short-term memory network deep learning model.

4. The method for constructing a prediction model for the risk of avian influenza based on the El Niño index according to claim 3, characterized in that, The evaluation metrics for assessing the multiple deep learning models include: mean absolute error, mean absolute percentage error, and root mean square error.

5. A system for constructing a predictive model for the occurrence risk of avian influenza based on the El Niño index, characterized in that, include: The dataset construction module is used to construct the H5 subtype avian influenza event dataset and the multivariate El Niño index dataset respectively, and to process the H5 subtype avian influenza event dataset and the multivariate El Niño index dataset to obtain the climate-avian influenza dataset. The regression model building module is used to build a regression model of the number of avian influenza events on the climate index based on the climate-avian influenza dataset. The prediction model building module is used to build multiple deep learning models based on the multivariate El Niño index dataset, evaluate the multiple deep learning models, and obtain the optimal prediction model for future climate change indices. The prediction module is used to predict the number of future avian influenza events based on the regression model and the optimal prediction model of the future climate change index, including predicting climate change in the future time period based on the optimal prediction model of the future climate change index. By incorporating future climate change as an independent variable into the regression model, the predicted number of avian influenza events in the future time period is obtained.

Images