Pertussis prevalence trend prediction method and device, electronic equipment and computer storage medium

Abstract

The invention provides a pertussis prevalence trend prediction method and device, electronic equipment and a computer storage medium. The pertussis epidemic trend prediction method comprises the following steps: S1, collecting daily newly-added case data of pertussis, and preprocessing the daily newly-added case data of pertussis to obtain case time sequence data; s2, decomposing the case time sequence data into a long-term trend component, a plurality of seasonal components and a residual component by using an MSTL method; performing noise reduction processing on the residual component to obtain a smooth residual component; s3, respectively inputting the long-term trend component and the seasonal component into an LSTM neural network model for training and prediction to obtain a predicted long-term trend component and a predicted seasonal component; s4, inputting the smooth residual component into an LSTM neural network model to obtain a predicted residual component; and S5, superposing the predicted long-term trend component, seasonal component and residual component, and outputting the finally predicted pertussis epidemic situation trend.

Description

Technical Field

[0001] The present invention relates to the technical field of disease prediction, and more specifically, to a method, device, electronic device and computer storage medium for predicting whooping cough epidemic trends. Background Art

[0002] Pertussis (whooping cough) is an acute respiratory infectious disease primarily caused by Bordetella pertussis. Although many countries around the world, including China, have effectively controlled the disease through vaccination, recent years have seen a resurgence of pertussis in some countries and regions, particularly in areas with high vaccination coverage. Consequently, the cyclical nature of pertussis epidemics has garnered significant attention and research. Peak epidemics typically occur every three to five years, and accurate prediction of pertussis epidemic trends is crucial for public health decision-making and preventive measures. Summary of the Invention

[0003] The purpose of the present invention is to provide a pertussis epidemic trend prediction method, device, electronic device and computer storage medium, wherein the pertussis epidemic trend prediction method can effectively and accurately predict the short-term fluctuations and long-term trends of the pertussis epidemic.

[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0005] In a first aspect, the present invention provides a method for predicting the epidemic trend of pertussis, comprising the following steps:

[0006] S1. Collecting daily new pertussis case data and preprocessing the daily new pertussis case data to obtain case time series data;

[0007] S2. Use the multiple seasonal trend decomposition (MSTL) method to decompose the case time series data into a long-term trend component, multiple seasonal components, and a residual component;

[0008] Performing noise reduction processing on the residual component to obtain a smoothed residual component;

[0009] S3. Inputting the long-term trend component and seasonal component into the LSTM neural network model for training and prediction, respectively, to obtain the predicted long-term trend component and seasonal component;

[0010] S4. Inputting the smoothed residual component into the LSTM neural network model to obtain a predicted residual component;

[0011] S5. Superimpose the predicted long-term trend component, seasonal component, and residual component to output the final predicted pertussis epidemic trend.

[0012] In step S1 of the present invention, the daily new pertussis case data includes but is not limited to the following information: reporting date, region where the case is located (such as province, city, county), number of new cases per day, diagnosis time, and data source.

[0013] In step S1 of the present invention, the preprocessing includes data cleaning and normalization.

[0014] Optionally, the data cleaning includes processing missing values and removing outliers.

[0015] Basic data cleaning was performed on the daily new pertussis case data to ensure data quality. The data was normalized to unify the numerical range and bring data of different dimensions and ranges to the same scale to facilitate subsequent modeling and analysis, while also improving the stability and convergence speed of model training.

[0016] The residual component after the MSTL decomposition is subjected to noise reduction processing to remove unnecessary high-frequency noise. Optionally, in step S2 of the present invention, the residual component is subjected to noise reduction processing using a fast Fourier transform (FFT).

[0017] The LSTM described in the present invention is a long short-term memory network, which is a special recurrent neural network (RNN) architecture.

[0018] The training parameters of the LSTM neural network model described in the present invention are: a time window T of 7 to 30, a batch size of 16 to 128, a learning rate of 0.001 to 0.01, and a number of training rounds of 50 to 200. As one embodiment, the time window T is 7, the batch size is 16, the learning rate is 0.01, and the number of training rounds is 100.

[0019] As one implementation method, the long-term trend component and seasonal component are respectively input into the LSTM neural network for training and prediction, and then the residual component processed in step S2 is separately trained and predicted. The long-term trend and seasonal components reflect the regularity and periodicity of the data, while the residual component contains more random noise and abnormal fluctuations. Separate training helps to adjust and optimize parameters for different characteristics, thereby improving the overall prediction accuracy. Although theoretically the three components can be input into the LSTM model at the same time for joint training, this approach may cause confusion when the model captures different dynamic characteristics, thereby affecting the learning effect and prediction accuracy. Therefore, in order to fully utilize the unique information of each component, a strategy of training separately and then superimposing is usually adopted.

[0020] The present invention uses the multiple seasonal trend decomposition (MSTL) method to decompose the case time series data into a long-term trend component, multiple seasonal components, and a residual component. This step will identify significant cyclical characteristics in pertussis epidemics, such as a short cycle of 1 year and a long cycle of 3-5 years.

[0021] The present invention inputs the smoothed residual component into an LSTM neural network model for training and prediction. The LSTM neural network model learns the short-term fluctuations in the residual component and improves the accuracy of the overall prediction.

[0022] In a second aspect, the present invention provides a device for predicting the epidemic trend of pertussis. The device comprises:

[0023] Input unit, used to obtain daily data on new cases of pertussis;

[0024] a processing unit, configured to process the daily new pertussis case data, wherein the processing includes preprocessing and multiple seasonal trend decomposition (MSTL) method processing;

[0025] An LSTM neural network model processing unit is used to input the processed long-term trend component, multiple seasonal components, and residual components into the LSTM neural network model for training and prediction;

[0026] The output unit is used to output the final predicted pertussis epidemic trend.

[0027] In a third aspect, the present invention further provides an electronic device. The electronic device includes at least one memory and at least one processor;

[0028] The at least one memory is coupled to the at least one processor, the at least one memory is used to store a computer program, the at least one processor is used to call the computer program, and the computer program includes instructions. When the instructions are executed by the at least one processor, the electronic device executes the pertussis epidemic trend prediction method as described in the first aspect above.

[0029] In a fourth aspect, the present invention further provides a computer storage medium comprising computer instructions, which, when executed on an electronic device, cause the electronic device to execute the pertussis epidemic trend prediction method described in the first aspect.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] Compared to traditional time series models, the pertussis epidemic trend forecasting method described in the present invention has the advantage of multiple period decomposition, which can more accurately predict the short-term fluctuations and long-term trends of the pertussis epidemic. At the same time, by utilizing the LSTM model, the present invention can effectively capture the nonlinear changes in pertussis epidemic data, improving the accuracy of the forecast. The pertussis epidemic trend forecasting method described in the present invention provides a reliable scientific basis, helping relevant departments to deploy prevention and control measures in advance and reduce the impact of pertussis on public health and the economy. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 Schematic diagram of the processing flow of the pertussis epidemic trend prediction method.

[0033] Figure 2 The present invention is a functional module block diagram of a device for predicting the epidemic trend of pertussis.

[0034] Figure 3 This is a structural block diagram of an electronic device. DETAILED DESCRIPTION

[0035] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0036] The following describes a method for predicting the epidemic trend of pertussis according to an embodiment of the present invention.

[0037] Specifically, the method for predicting the epidemic trend of pertussis includes the following steps:

[0038] S1. Collecting daily new pertussis case data and preprocessing the daily new pertussis case data to obtain case time series data;

[0039] The data on daily new cases of pertussis were obtained through official websites of local disease control and health commissions and in combination with publicly available epidemiological data at home and abroad. The data were used as the original data source and were desensitized before use.

[0040] The daily new pertussis case data used for training and prediction mainly includes the following information: reporting date, region where the case is located (such as province, city, or county), number of new cases per day, time of diagnosis, and data source.

[0041] The preprocessing includes data cleaning and normalization.

[0042] Data cleaning involves handling missing values and removing outliers. Specifically, during the data cleaning process, the integrity of the daily pertussis case data must first be checked. For missing values, various strategies can be employed: for example, for continuous variables (such as the number of new cases per day), linear interpolation or the mean of the preceding and following data can be used to fill in missing values. For discrete or categorical variables (such as regional information), the mode can be used or marked as "unknown" to ensure data integrity and continuity during subsequent model processing. Secondly, to remove outliers, statistical methods can be used to detect data deviations. For example, box plots can be used to determine the upper and lower quartile ranges of the data or to calculate the Z-score. If the number of new cases on a given day is significantly higher or lower than the range of the period (for example, exceeding the mean plus or minus three standard deviations), the data is considered an outlier. In this case, the data point can be removed or the median can be used to correct the outlier based on the surrounding data to maintain statistical consistency across the data.

[0043] The specific steps for normalization are as follows: First, select an appropriate data scaling method, such as minimum-maximum scaling or Z-score normalization. For example, using the minimum-maximum scaling method, calculate the minimum and maximum daily number of new cases in the dataset. Then, use the formula (x - min) / (max - min) to linearly map each data point to the range 0 to 1. For example, for a set of case numbers [50, 60, 55, 65, 70], if the minimum is 50 and the maximum is 70, the corresponding data will be converted to [0, 0.5, 0.25, 0.75, 1]. Another method is Z-score normalization, which first calculates the mean μ and standard deviation σ of the data. Then, use the formula (x - μ) / σ to transform the data so that the transformed data follows a distribution with a mean of 0 and a standard deviation of 1. For example, if the mean of the case data is 60 and the standard deviation is 7.07, then the data point 50 will become (50-60) / 7.07≈-1.41 after normalization. After completing these operations, all data input to the model will be within a uniform scale, which helps improve training efficiency and model prediction accuracy.

[0044] S2. Use the multiple seasonal trend decomposition (MSTL) method to decompose the case time series data into a long-term trend component, multiple seasonal components, and a residual component;

[0045] The residual component is subjected to noise reduction processing using Fast Fourier Transform (FFT) to obtain a smooth residual component.

[0046] Specifically, during FFT denoising, the original residual components are first subjected to a fast Fourier transform (FFT), converting them from the time domain to the frequency domain. This process yields the frequency components and their corresponding amplitudes. Next, based on spectrum analysis, an appropriate frequency threshold is determined to set high-frequency components with large amplitudes that represent noise to zero or perform appropriate attenuation. This threshold can be determined by observing the location of non-periodic noise peaks in the spectrum. An inverse fast Fourier transform (inverse FFT) is then performed on the processed spectrum to convert the signal back to the time domain. The resulting residual components are the smoothed signal after denoising. For example, for a segment of residual data [2, 5, 3, 6, 4, 7, 3, 5], an FFT transform reveals abnormal high-frequency amplitudes. By setting a threshold to zero these high-frequency components and then performing an inverse FFT, the resulting signal exhibits smoother fluctuations and significantly reduced noise, thereby better reflecting the true trend of the data.

[0047] S3. Inputting the long-term trend component and seasonal component into the LSTM neural network model for training and prediction, respectively, to obtain the predicted long-term trend component and seasonal component;

[0048] The training parameters of the LSTM neural network model are: time window T is 7, batch size is set to 16, learning rate is 0.01, and the number of training rounds is 100.

[0049] S4. Inputting the smoothed residual component into the LSTM neural network model to obtain a predicted residual component;

[0050] S5. Superimpose the predicted long-term trend component, seasonal component, and residual component to output the final predicted pertussis epidemic trend.

[0051] For example, Figure 1 As shown, the number of new pertussis cases per day for a province was collected for one consecutive year (e.g. Figure 1As shown in the "Daily Number of New Infections" on the left side of the figure), first use MSTL to perform multiple seasonal trend decomposition on the time series to obtain a long-term trend component (such as a curve reflecting the overall upward or downward trend throughout the year), several seasonal components (such as monthly or quarterly cyclical fluctuations) and residual components (high-frequency random fluctuation parts). Corresponding to step S3, the long-term trend component and each seasonal component are input into multiple parallel LSTM models for training. The hyperparameters of each LSTM model are set to time window T=7, batch size=16, learning rate=0.01, and number of training rounds=100 rounds. After the training is completed, predictions are made for the next several days (such as the next week or the next month) to obtain the predicted long-term trend component and each seasonal component. Corresponding to step S4, the residual component that has been denoised by FFT is input into the LSTM model of the same structure, and the same training parameters are used for learning to capture the random fluctuation characteristics and predict the future residual part. Finally, corresponding to step S5, the predicted long-term trend component, all seasonal components and residual components are numerically superimposed ( Figure 1 The "linear function L(t)" on the right side of the figure is used as an example), and we can finally get the overall prediction curve of the number of new cases per day in the future ( Figure 1 The overall prediction curve not only includes the overall trend, but also integrates periodic changes and random disturbances, thus effectively reflecting the comprehensive evolution trend of the pertussis epidemic in the short and long term.

[0052] This embodiment also provides a device for predicting the epidemic trend of pertussis.

[0053] The following describes an apparatus for executing the above-mentioned pertussis epidemic trend prediction method.

[0054] Figure 2 A functional module block diagram of the device provided in this embodiment is provided. The device for predicting the epidemic trend of pertussis includes:

[0055] Input unit 10 is used to obtain daily data on new cases of pertussis;

[0056] a processing unit 20 for processing the daily new pertussis case data, wherein the processing includes preprocessing and multiple seasonal trend decomposition (MSTL) method processing;

[0057] An LSTM neural network model processing unit 30 is used to input the processed long-term trend component, multiple seasonal components, and residual components into the LSTM neural network model for training and prediction;

[0058] The output unit 40 is used to output the final predicted pertussis epidemic trend.

[0059] The device provided in this embodiment and the pertussis epidemic trend prediction method provided in this application have the same concept. The specific implementation process is detailed in the full text of the specification and will not be repeated here.

[0060] This embodiment also provides an electronic device. The electronic device capable of executing the above-mentioned pertussis epidemic trend prediction method is described below.

[0061] Figure 3 A structural block diagram of an electronic device 100 for implementing a method for predicting a pertussis epidemic trend is shown. The electronic device 100 includes: at least one memory 101 and at least one processor 102; the at least one memory 101 is coupled to the at least one processor 102, the at least one memory 101 is used to store a computer program, and the at least one processor 102 is used to call the computer program. The computer program includes instructions. When the instructions are executed by the at least one processor, the electronic device executes the above-mentioned pertussis epidemic trend prediction method.

[0062] This embodiment also provides a computer storage medium. The following describes the computer storage medium containing the above-mentioned pertussis epidemic trend prediction method.

[0063] A computer storage medium includes computer instructions. When the computer instructions are executed on an electronic device, the electronic device executes the above-mentioned pertussis epidemic trend prediction method.

[0064] Obviously, the above embodiments of the present invention are merely examples for the purpose of clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Those skilled in the art will appreciate that other variations or modifications can be made based on the above description. It is not necessary and impossible to enumerate all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the claims of the present invention.

Claims

1. A method for predicting the epidemic trend of pertussis, characterized in that: The following steps are involved: S1. Collecting daily new pertussis case data and preprocessing the daily new pertussis case data to obtain case time series data; S2. Use the multiple seasonal trend decomposition (MSTL) method to decompose the case time series data into a long-term trend component, multiple seasonal components, and a residual component; Performing noise reduction processing on the residual component to obtain a smoothed residual component; S3. Inputting the long-term trend component and seasonal component into the LSTM neural network model for training and prediction, respectively, to obtain the predicted long-term trend component and seasonal component; S4. Inputting the smoothed residual component into the LSTM neural network model to obtain a predicted residual component; S5. Superimpose the predicted long-term trend component, seasonal component, and residual component to output the final predicted pertussis epidemic trend.

2. The method for predicting the epidemic trend of pertussis according to claim 1, characterized in that: In step S1, the preprocessing includes data cleaning and normalization.

3. The method for predicting the epidemic trend of pertussis according to claim 2, characterized in that: The data cleaning includes processing missing values and removing outliers.

4. The method for predicting the epidemic trend of pertussis according to claim 1, characterized in that: In step S2, the residual component is subjected to noise reduction processing using Fast Fourier Transform (FFT).

5. The method for predicting the epidemic trend of pertussis according to claim 1, characterized in that: The training parameters of the LSTM neural network model are as follows: time window T is 7-30, batch size is set to 16-128, learning rate is 0.001-0.01, and the number of training rounds is 50-200.

6. The method for predicting the epidemic trend of pertussis according to claim 1, characterized in that: In step S1, the daily new pertussis case data includes the following information: reporting date, region where the case is located, number of new cases per day, time of diagnosis, and data source.

7. A device for predicting the epidemic trend of pertussis, characterized in that: include: Input unit, used to obtain daily data on new cases of pertussis; a processing unit, configured to process the daily new pertussis case data, wherein the processing includes preprocessing and multiple seasonal trend decomposition (MSTL) method processing; An LSTM neural network model processing unit is used to input the processed long-term trend component, multiple seasonal components, and residual components into the LSTM neural network model for training and prediction; The output unit is used to output the final predicted pertussis epidemic trend.

8. An electronic device, characterized in that: comprising at least one memory and at least one processor; The at least one memory is coupled to the at least one processor, the at least one memory is used to store a computer program, the at least one processor is used to call the computer program, and the computer program includes instructions. When the instructions are executed by the at least one processor, the electronic device executes the pertussis epidemic trend prediction method as described in any one of claims 1 to 6.

9. A computer storage medium, characterized in that The method comprises computer instructions, which, when executed on an electronic device, enable the electronic device to execute the pertussis epidemic trend prediction method according to any one of claims 1 to 6.

Images