Animal epidemic disease monitoring data statistical analysis system and method

Abstract

The invention relates to an animal epidemic disease monitoring data statistical analysis system and method. The method comprises the following steps: a multi-source data fusion module performs standardized fusion processing based on input case data, environment data and population data to obtain a spatio-temporal joint data matrix; a risk probability prediction module performs area scanning shape prediction on geographic feature data and population distribution data in the matrix, outputs a dynamic scanning area set, and performs environment and small sample correction to obtain risk probability distribution; and the hierarchical prevention and control visualization module performs spatial aggregation test based on the distribution, outputs a significant epidemic disease aggregation area, and generates a hierarchical prevention and control thermodynamic diagram in combination with an environment correction factor. According to the system, through multi-source data fusion, geographically adaptive dynamic window prediction and spatial aggregation test, the animal epidemic disease risk probability can be accurately analyzed, an epidemic disease aggregation area is output, a hierarchical prevention and control thermodynamic diagram is generated, and the animal epidemic disease monitoring level and the prevention and control visualization capability are further improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of computers, and particularly relates to an animal epidemic disease monitoring data statistical analysis system and method. BACKGROUND

[0002] With the development of animal epidemic disease monitoring and spatial statistical technology, an epidemic disease aggregation detection method based on spatial scanning statistics has appeared. This method can calculate the deviation of the case number and the expected value by moving the scanning window, identify the area with abnormally high incidence, and provide a new technical means for epidemic disease monitoring. However, on the one hand, this method mainly relies on a single data source such as case reports or environmental monitoring data for animal epidemic disease monitoring, and has insufficient data fusion capability, cannot effectively integrate multi-source heterogeneous data, and leads to a serious information island phenomenon. On the other hand, this method usually uses a preset circular or elliptical window to traverse the area, ignores the restriction of geographical features on the spread of the epidemic disease, and often cannot comprehensively reflect the spatio-temporal dynamic characteristics of the epidemic disease. SUMMARY

[0003] Therefore, it is necessary to provide an animal epidemic disease monitoring data statistical analysis system and method to improve the accuracy and timeliness of epidemic disease monitoring and enhance the scientific nature of risk prediction in view of the above technical problems.

[0004] In a first aspect, the application provides an animal epidemic disease monitoring data statistical analysis system, comprising:

[0005] a multi-source data fusion module configured to perform standardized fusion processing based on input case data, environmental data and population data to obtain a spatio-temporal joint data matrix, wherein the case data includes geographical coordinates, a time stamp and a species code of animal cases, the environmental data includes temperature, precipitation and vegetation index, and the population data includes farm density and animal stock;

[0006] a risk probability prediction module configured to:

[0007] input geographical feature data and population distribution data in the spatio-temporal joint data matrix into a dynamic window prediction model to perform regional scanning shape prediction, and output a dynamic scanning region set;

[0008] perform environmental correction and small sample stabilization processing according to the dynamic scanning region set to obtain a risk probability distribution, wherein the risk probability distribution includes a basic risk parameter and an actual joint likelihood ratio, and the basic risk parameter includes an initial risk value, an environmental correction factor and a small sample correction term;

[0009] a hierarchical prevention and control visualization module configured to:

[0010] perform spatial aggregation testing based on the risk probability distribution by using a graphics processing unit, and output a significant epidemic disease aggregation area;

[0011] According to the significance disease aggregation area and the environmental correction factor, weight superposition processing is performed to generate a hierarchical prevention and control heat map, and the environmental correction factor includes a temperature deviation coefficient, a precipitation cumulative deviation value and a vegetation index abnormality degree.

[0012] In one of the embodiments, the multi-source data fusion module comprises:

[0013] The data processing subunit is configured to:

[0014] According to the missing geographic coordinates in the input case data, inverse distance weighted interpolation processing is performed to obtain case data after interpolation completion;

[0015] The temperature, precipitation and vegetation index in the environmental data are normalized to obtain normalized environmental data;

[0016] Based on the density of the breeding farm and the animal stock in the population data, logarithmic transformation processing is performed to obtain logarithmically transformed population data;

[0017] The data fusion subunit is configured to perform weight fusion processing on the case data after interpolation completion, the normalized environmental data and the logarithmically transformed population data to obtain a spatio-temporal joint data matrix.

[0018] In one of the embodiments, the risk probability prediction module comprises a dynamic window prediction subunit configured to:

[0019] Based on the river buffer and road network topology in the geographic feature data, feature extraction is performed to obtain a passability matrix, and the river buffer is a region within 500 meters from a river;

[0020] The density gradient in the population data and the passability matrix are subjected to feature fusion processing to obtain an input feature matrix, and the density gradient is generated by calculating the difference in animal stock of adjacent grids;

[0021] The input feature matrix is input into a dynamic window prediction model based on a convolutional neural network to perform sliding window probability prediction, and a probability distribution map is output;

[0022] According to the probability distribution map, threshold truncation processing is performed to obtain a dynamic scanning area set.

[0023] In one of the embodiments, the risk probability prediction module further comprises a risk probability synthesis subunit configured to:

[0024] According to the actual case number and the expected case number in the dynamic scanning area set, basic likelihood ratio calculation processing is performed to obtain an initial risk value, and the expected case number is generated by regression prediction based on historical epidemic data and animal stock;

[0025] According to the deviation of temperature, precipitation and vegetation index in the environmental data from the historical mean, Gaussian penalty processing is performed to obtain an environmental correction factor;

[0026] According to the regions in the dynamic scanning region set in which the expected number of cases is less than 1, gamma function stabilization processing is performed to obtain a small sample correction term;

[0027] In combination with the initial risk value, the environmental correction factor and the small sample correction term, a basic risk parameter is obtained, and joint likelihood ratio synthesis processing is performed to generate an actual joint likelihood ratio;

[0028] A risk probability distribution is constructed by the basic risk parameter and the actual joint likelihood ratio.

[0029] In one of the embodiments, the calculation formula of the environmental correction factor is:

[0030]

[0031] Wherein, L is the environmental correction factor, e k represents the current environmental factor, including temperature, precipitation, vegetation index, μ k is the historical mean of the environmental factor, and σ k is the historical standard deviation of the environmental factor, and k is the total number of environmental factors.

[0032] In one of the embodiments, the hierarchical prevention and control visualization module includes a region inspection subunit for:

[0033] According to the case probability density in the risk probability distribution, a set of Poisson distribution parameters is constructed to determine the generation parameters of the random case distribution;

[0034] According to the segmentation of the preset total number of Monte Carlo simulations into multiple parallel subtasks, and through the graphics processing unit, the case distribution simulation of each subtask is performed;

[0035] Based on the generation parameters of the random case distribution, a spatial point process simulation is performed through a Monte Carlo simulation engine to generate a simulated case data set conforming to the risk probability distribution;

[0036] For the simulated case data set of each subtask, a Gaussian penalty factor is calculated according to the deviation of temperature, precipitation and vegetation index in the environmental data from the historical mean;

[0037] For the regions in each simulated case data set in which the expected number of cases is less than 1, a risk adjustment term is obtained by a gamma distribution compensation algorithm;

[0038] In combination with the initial risk value, the Gaussian penalty factor and the gamma function adjustment term, a simulated joint likelihood ratio is generated, and the simulated joint likelihood ratios of each subtask are counted to generate a global simulated joint likelihood ratio set;

[0039] According to the actual joint likelihood ratio in the global simulation joint likelihood ratio set and the risk probability distribution, a corrected p-value calculation process is performed to obtain a corrected p-value;

[0040] If the corrected p-value is less than a preset threshold, it is determined that the corresponding scanning area is a significant disease aggregation area.

[0041] In one embodiment, the hierarchical prevention and control visualization module includes a prevention and control heat map generation subunit for:

[0042] According to the historical infection rate, the species susceptibility weight is determined, and the frequency statistics processing is performed based on the historical transportation record data to obtain the animal transportation frequency;

[0043] The actual joint likelihood ratio, the species susceptibility weight and the animal transportation frequency are linearly superimposed to obtain a comprehensive risk index;

[0044] Based on the comprehensive risk index, Sigmoid function mapping processing is performed to obtain a standardized risk value;

[0045] According to the standardized risk value, a grade threshold division process is performed to generate a hierarchical prevention and control heat map, which includes a high-risk red area, a medium-risk orange area and a low-risk yellow area.

[0046] In a second aspect, the application also provides an animal disease monitoring data statistical analysis method, which comprises:

[0047] Based on the input case data, environmental data and population data, standardized fusion processing is performed to obtain a spatio-temporal joint data matrix, the case data including animal case geographic coordinates, timestamps and species codes, the environmental data including temperature, precipitation and vegetation index, and the population data including farm density and animal stock;

[0048] The geographic feature data and population distribution data in the spatio-temporal joint data matrix are input into a dynamic window prediction model for regional scanning shape prediction, and a dynamic scanning area set is output;

[0049] According to the dynamic scanning area set, environmental correction and small sample stabilization processing are performed to obtain a risk probability distribution, the risk probability distribution including a basic risk parameter and an actual joint likelihood ratio, the basic risk parameter including an initial risk value, an environmental correction factor and a small sample correction term;

[0050] Based on the risk probability distribution, spatial aggregation test is performed by a graphics processing unit to output a significant disease aggregation area;

[0051] According to the significant disease aggregation area and the environmental correction factor, weight superposition processing is performed to generate a hierarchical prevention and control heat map, and the environmental correction factor includes a temperature deviation coefficient, a precipitation cumulative deviation value and a vegetation index abnormality degree.

[0052] In a third aspect, the present application also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the first aspect when executing the computer program.

[0053] In a fourth aspect, the present application also provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of the first aspect.

[0054] The animal disease monitoring data statistical analysis system can perform standardized fusion processing on case data, environmental data and population data through the multi-source data fusion module to form a time-space joint data matrix, thereby providing comprehensive data support for subsequent risk prediction. The risk probability prediction module can use a dynamic window prediction model to perform regional scanning shape prediction according to geographic feature data and population distribution data, output a dynamic scanning region set, and combine environmental correction to suppress abnormal region misjudgment, small sample stabilization to reduce risk fluctuation, thereby further improving the accuracy of the risk probability distribution. The hierarchical prevention and control visualization module uses a graphics processor to accelerate spatial aggregation test, accurately locates a significant disease aggregation area, and generates a hierarchical prevention and control heat map in combination with environmental correction factors, thereby providing a scientific basis for the development of prevention and control measures.

[0055] Through the cooperative operation of the above modules, the system significantly improves the accuracy and timeliness of disease monitoring, enhances the scientificity of risk prediction and the visualization effect of prevention and control measures, and provides a more precise and intelligent tool for animal disease monitoring and prevention and control. BRIEF DESCRIPTION OF DRAWINGS

[0056] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related art, the drawings needed to be used in the embodiment or related art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0057] Figure 1 A structure schematic diagram of an animal disease monitoring data statistical analysis system is provided for an exemplary embodiment of the present application.

[0058] Figure 2 A flowchart of an animal disease monitoring data statistical analysis method is provided for an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0059] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not to limit the present application.

[0060] In one embodiment, as shown in Figure 1 A disease monitoring data statistical analysis system 100 is provided, and the present embodiment is exemplified by the system applied to a terminal. It should be understood that the system can also be applied to a server, and can also be applied to a system including a terminal and a server, and can be realized through the interaction of the terminal and the server. In the present embodiment, the system includes:

[0061] A multi-source data fusion module 101 is configured to perform standardized fusion processing based on inputted case data, environmental data and population data, to obtain a spatio-temporal joint data matrix. The case data includes geographical coordinates, time stamps and species codes of animal cases. The environmental data includes temperature, precipitation and vegetation index. The population data includes farm density and animal stock.

[0062] Specifically, the geographical coordinates of animal cases accurately identify the geographical location of the disease, providing a basis for subsequent spatial analysis. The time stamp records the time when the case occurs, which helps to track the development process and transmission trend of the disease. The species code can clearly identify the type of the diseased animal. In addition, temperature, precipitation and vegetation index in the environmental data have a significant impact on the spread of animal diseases. For example, high temperature and high humidity environment may promote the breeding and spread of certain pathogens, while specific vegetation conditions may affect the distribution and range of host animals. In addition, the farm density in the population data reflects the degree of animal aggregation; the animal stock is an important basis for evaluating the potential impact of the disease. The multi-source data fusion module 101 performs standardized fusion processing on these data, that is, it can adopt corresponding standardization methods according to the characteristics of different types of data. For example, for geographical coordinate data, the coordinate system and precision standard can be unified. Finally, the standardized data of various types are fused to generate a spatio-temporal joint data matrix, so that each data dimension can work together to analyze the disease, and provide a comprehensive and unified data basis for the subsequent modules.

[0063] A risk probability prediction module 102 is configured to:

[0064] Input the geographical feature data and population distribution data in the spatio-temporal joint data matrix into a dynamic window prediction model to perform regional scanning shape prediction, and output a dynamic scanning region set;

[0065] Based on the dynamic scanning area set, environmental correction and small sample stabilization processing are performed to obtain the risk probability distribution. The risk probability distribution includes basic risk parameters and actual joint likelihood ratios. The basic risk parameters include initial risk values, environmental correction factors and small sample correction items.

[0066] Specifically, the dynamic window prediction model is a core tool for regional scanning and shape prediction based on a prediction algorithm, based on a spatiotemporal joint data matrix. Furthermore, the dynamic window prediction model continuously adjusts the scanning window based on spatiotemporal trends in the data to optimize the detection and assessment of risks in different regions, thereby ensuring the model's sensitivity to changes in regional risks. The model uses geographic feature data and population distribution data from the spatiotemporal joint data matrix to perform regional scanning shape prediction, ultimately generating a dynamic scanning region set. This set represents the high-risk areas predicted by the model within a certain temporal and spatial range.

[0067] Specifically, after obtaining the set of dynamic scanning areas, the module also needs to perform environmental correction and small sample stabilization processing. Environmental correction is intended to account for the impact of external environmental changes on risk. By correcting the original prediction results, the risk prediction can be made more consistent with the actual environment. Furthermore, in actual applications, there may be insufficient or biased sample data. Therefore, small sample stabilization processing can be performed to smooth the data using statistical methods, reducing the uncertainty caused by small samples in the data, thereby improving the stability and reliability of risk prediction and outputting a more accurate risk probability distribution. This risk probability distribution is a multi-dimensional set of probability values ​​that can clearly display the risk levels of different regions and different risk factors.

[0068] The hierarchical prevention and control visualization module 103 is used to:

[0069] Based on the risk probability distribution, a spatial clustering test is performed through a graphics processor to output significant disease clusters;

[0070] Based on the significant epidemic clusters and environmental correction factors, weighted superposition processing is performed to generate a graded prevention and control heat map. The environmental correction factors include temperature deviation coefficient, precipitation cumulative deviation value and vegetation index anomaly.

[0071] Specifically, the module first performs spatial aggregation test through the graphics processor to identify significant disease aggregation areas. Spatial aggregation test is a statistical method that can detect the degree of spatial aggregation of animal diseases, thereby determining high-risk areas of the disease. And through the graphics processor, the process of spatial aggregation test can be accelerated, enabling it to quickly process large amounts of data and output significant disease aggregation areas. After obtaining the significant disease aggregation areas, the module further combines environmental correction factors for weight superposition processing. Among them, the environmental correction factors include temperature deviation coefficient, precipitation cumulative deviation value and vegetation index abnormality, which reflect the influence of environmental conditions on disease transmission. By combining the environmental correction factors with the significant disease aggregation areas, the module can more accurately assess the disease risk level in different regions and generate a hierarchical prevention and control heat map. The heat map can display the disease risk level in different regions through color changes. Through this process, not only the efficiency of decision-making is improved, but also the pertinence and implementation effect of prevention and control measures are enhanced, providing scientific and intuitive decision support for animal disease monitoring and prevention.

[0072] In the above-mentioned animal disease monitoring data statistical analysis system 100, the multi-source data fusion module 101 standardizes and fuses case data, environmental data and population data to obtain a spatio-temporal joint data matrix, so that the geographical coordinates, time stamps of cases, temperature, precipitation of environment, and breeding field density, animal stock of population, etc. multi-dimensional information are organically combined, providing a comprehensive data basis for subsequent disease analysis. The risk probability prediction module 102 inputs the geographical feature data and population distribution data in the matrix into a dynamic window prediction model, which can output a dynamic scanning area set that is more in line with the actual disease transmission situation, combined with environmental correction processing, which suppresses the risk misjudgment of environmental abnormal areas, and small sample stabilization processing reduces the risk fluctuation of small sample areas, thereby improving the accuracy of risk probability distribution. The hierarchical prevention and control visualization module 103 can use the graphics processor to perform spatial aggregation test to quickly and accurately locate significant disease aggregation areas, and perform weight superposition processing on the significant disease aggregation areas and environmental correction factors to generate a hierarchical prevention and control heat map, realizing the visualization classification of disease prevention and control areas and providing intuitive basis for prevention and control decision-making.

[0073] Compared with traditional animal disease monitoring methods, this system can not only more comprehensively integrate multi-source heterogeneous data, but also significantly improve the accuracy of disease risk prediction and the precision of aggregation area identification through multi-source data fusion, dynamic area accurate prediction, environmental and small sample fine processing, and spatial aggregation test based on graphics processor acceleration. In addition, by generating a visual hierarchical prevention and control heat map, it provides a scientific and intuitive basis for prevention and control decision-making, which helps to promote the development of animal disease monitoring and prevention work towards more accurate and intelligent direction.

[0074] In one embodiment, the multi-source data fusion module 101 comprises:

[0075] a data processing subunit, configured to:

[0076] perform inverse distance weighted interpolation processing on missing geographic coordinates in the input case data to obtain case data after interpolation completion;

[0077] normalize temperature, precipitation and vegetation index in the environmental data to obtain normalized environmental data;

[0078] perform logarithmic transformation processing on the breeding farm density and the animal stock in the population data to obtain population data after logarithmic transformation;

[0079] a data fusion subunit, configured to perform weight fusion processing on the case data after interpolation completion, the normalized environmental data and the population data after logarithmic transformation to obtain a spatio-temporal joint data matrix.

[0080] Specifically, in actual animal disease monitoring, case data may have missing geographic coordinates. Therefore, the data processing subunit can use the inverse distance weighted interpolation processing method. For each case with missing geographic coordinates, the distances of the surrounding known coordinate points are calculated, and the corresponding weights are assigned according to the reciprocal of the distance, and then the coordinates of these known points are weighted and averaged to obtain the estimated value of the missing coordinates. After inverse distance weighted interpolation processing, the geographic coordinates in the case data are completed, and then the case data after interpolation completion is formed. In order to eliminate the influence of the dimensional difference in the environmental data, the linear transformation method can be used to map the value of each variable to a specific interval for normalization, converting it into a numerical value with the same scale. The breeding farm density and the animal stock in the population data usually have a large numerical range and may present a skewed distribution. However, this distribution characteristic may have an adverse effect on subsequent data analysis and model training. Therefore, the data processing subunit performs logarithmic transformation processing on the breeding farm density and the animal stock, compressing the large numerical range into a smaller interval and making the data distribution more uniform, so as to facilitate subsequent analysis and modeling.

[0081] Since different types of data have different importance and influence in animal disease monitoring. For example, case data directly reflects the occurrence of the disease, environmental data may affect the spread and spread of the disease, and population data is closely related to the risk of disease transmission and the potential impact of the scale. Therefore, the data fusion sub-unit can assign appropriate weights to the interpolated and completed case data, normalized environmental data and log-transformed population data based on various factors such as analysis of historical data, expert experience and judgment, etc. to generate a spatio-temporal joint data matrix. This matrix not only contains information such as geographical coordinates, time stamp, species code of animal cases, but also integrates environmental data and population data, providing a comprehensive and unified data basis for subsequent disease monitoring analysis, enabling the system to conduct in-depth analysis and prediction of animal diseases from multiple dimensions.

[0082] In one embodiment, the risk probability prediction module 102 includes a dynamic window prediction sub-unit for:

[0083] Based on the river buffer and road network topology in the geographical feature data, the feature extraction is performed to obtain the passability matrix, and the river buffer is the area within 500 meters from the river;

[0084] The density gradient in the population data and the passability matrix are subjected to feature fusion processing to obtain an input feature matrix, and the density gradient is generated by calculating the difference in the number of animals in the adjacent grid;

[0085] The input feature matrix is input into the dynamic window prediction model based on the convolutional neural network to perform sliding window probability prediction, and a probability distribution map is output;

[0086] According to the probability distribution map, threshold truncation processing is performed to obtain a dynamic scanning area set.

[0087] Specifically, river buffer zones, defined as areas within 500 meters of a river, are critical habitats for animals seeking water and roosting, and are also highly likely to serve as sites of disease transmission. This subunit utilizes geographic information system technology to accurately map river buffer zones, quantify their geometric parameters such as area and perimeter, and assess the potential role of water flow in driving disease transmission, taking into account factors such as river flow direction and velocity. Road network topology reflects the layout of roads within a region, including connectivity, intersection locations, and the capacity of different road sections. Animal transport frequently relies on road networks, and diseases can easily spread through the movement of animals during transport. This subunit utilizes graph theory and other algorithms to abstract the road network into a graph structure composed of nodes and edges. Nodes represent road intersections or important transportation hubs, while edges represent road sections connecting these nodes. Each edge is assigned a weight to further measure road accessibility. Through comprehensive analysis and feature extraction of river buffer zones and road network topology, an accessibility matrix is ​​constructed that reflects the geographic accessibility of the region.

[0088] Specifically, this subunit generates a density gradient by calculating the difference in animal populations between adjacent grids. First, the entire monitoring area can be divided into multiple uniform grids, with each grid serving as a basic analysis unit. Second, the animal population within each grid is counted. Finally, the difference in animal populations between adjacent grids is calculated, and the density gradient is determined based on the magnitude and direction of this difference. The direction and magnitude of this density gradient visually demonstrate the mobility trends and concentration of animal populations, which is important for assessing disease transmission risk. The subunit then combines the density gradient with the previously generated accessibility matrix, comprehensively considering the interrelationship between the two, to perform feature fusion processing. For example, if a high density gradient is also present in an area with high accessibility, it indicates high animal mobility and uneven population distribution, potentially significantly increasing the risk of disease transmission in that area. Finally, an input feature matrix is ​​generated. This matrix integrates geographic accessibility and population distribution characteristics, providing comprehensive and more targeted input data for the subsequent dynamic window prediction model.

[0089] The convolutional neural network has strong feature learning ability and can automatically extract complex feature patterns from a large amount of data. In this embodiment, the input feature matrix is input into the dynamic window prediction model based on the convolutional neural network, which can predict each window on the input feature matrix in a sliding window manner, where each window corresponds to a specific area, and then output the probability value of the occurrence of the epidemic in the area to generate a probability distribution map. The probability distribution map can show the possibility of the occurrence of the epidemic in different positions in the entire monitoring area in a graphical manner. In addition, after obtaining the probability distribution map, the subunit can further determine which areas are high-risk areas that need to be focused on through threshold truncation processing. Illustratively, a suitable probability threshold can be set according to historical epidemic data, expert experience and actual monitoring needs. For example, the threshold is set to 0.6, which means that when the probability of the occurrence of the epidemic in a certain area is greater than or equal to 0.6, the area is determined as a high-risk area. Then, the probability value of each area in the probability distribution map is compared with the threshold, and the area with a probability value greater than or equal to the threshold is marked as a dynamic scanning area. Finally, all the marked dynamic scanning areas are summarized to form a dynamic scanning area set. The areas in the set are high-risk areas selected according to the model prediction results, which greatly improves the pertinence and efficiency of subsequent epidemic monitoring and prevention and control work.

[0090] In one embodiment, the risk probability prediction module further comprises a risk probability synthesis subunit, configured to:

[0091] According to the actual case number and the expected case number in the dynamic scanning area set, perform a basic likelihood ratio calculation process to obtain an initial risk value, and the expected case number is generated by regression prediction based on historical epidemic data and animal inventory;

[0092] According to the deviation of temperature, precipitation and vegetation index in the environmental data from the historical mean value, perform a Gaussian penalty process to obtain an environmental correction factor;

[0093] According to the region in the dynamic scanning area set with an expected case number less than 1, perform a gamma function stabilization process to obtain a small sample correction term;

[0094] Combine the initial risk value, the environmental correction factor and the small sample correction term to obtain a basic risk parameter, and perform a joint likelihood ratio synthesis process to generate an actual joint likelihood ratio;

[0095] Construct a risk probability distribution by the basic risk parameter and the actual joint likelihood ratio.

[0096] Specifically, for each dynamic scanning area in the dynamic scanning area set, the actual case number is the actual number of sick animals in the area in the current epidemic monitoring, which can directly reflect the actual situation of the epidemic in the area. And through in-depth analysis of historical epidemic data, the rules and trends of the occurrence of the epidemic can be mined, combined with the current animal inventory information, and the regression prediction model is used to calculate the expected case number. Among them, the regression prediction model can establish a mathematical relationship between the expected case number according to multiple related factors in the historical data such as the epidemic incidence in different time periods in the past, the change of animal inventory, etc. Using the regression prediction model, inputting the current animal inventory and other related data, the expected case number that may occur in the dynamic scanning area under normal circumstances can be predicted. After obtaining the actual case number and the expected case number, the base likelihood ratio is calculated to obtain the initial risk value. The initial risk value preliminarily reflects the relative risk degree of the occurrence of the epidemic in the area. Among them, the base likelihood ratio is an important indicator to measure the deviation degree of the actual observation result from the expected result.

[0097] Specifically, environmental factors have an important influence on the spread and occurrence of animal epidemics. For temperature data, the difference between the current temperature and the historical average, that is, the temperature deviation value, can be calculated. If the current temperature is significantly higher or lower than the historical average, it may affect the survival, reproduction of the pathogen, and the immunity and behavior pattern of the animal, thereby changing the epidemic transmission risk. Similarly, the precipitation cumulative deviation value, that is, the difference between the total amount of precipitation in the current period and the historical average of the same period, can be calculated. In addition, the deviation of the vegetation index from the historical average can also be calculated to quantify the influence of the vegetation condition on the spread of the epidemic.

[0098] In one embodiment, the calculation formula of the environmental correction factor is:

[0099]

[0100] wherein L is the environmental correction factor, e k represents the current environmental factor, including temperature, precipitation, vegetation index, μ k is the historical average of the environmental factor, σ k is the historical standard deviation of the environmental factor, and k is the total number of environmental factors.

[0101] Specifically, in the above formula, when the deviation degree is small, L is close to 1, indicating that the current environmental condition is close to the historical average. When the deviation degree is large, L is close to 0, indicating that the current environmental condition deviates significantly from the historical average.

[0102] In the dynamic scanning area set, some areas may have less than one expected case due to fewer animals or other reasons. For these small sample areas, a gamma function stabilization process can be used to obtain a small sample correction term. Specifically, the gamma distribution has two parameters, shape parameter and scale parameter, which can be set based on historical data fitting or prior knowledge. For areas with less than one expected case, the adjusted risk value, i.e. the small sample correction term, can be calculated according to the probability density function of the gamma distribution, combined with the actual case number and the expected case number, so that the risk assessment result is more stable and reliable in the case of small sample, avoiding risk misjudgment due to small sample size. The initial risk value, the environmental correction factor and the small sample correction term obtained above can be combined to obtain the basic risk parameter. And the factors in the basic risk parameter can be combined to obtain the actual joint likelihood ratio by multiplication. Compared with the initial risk value, the actual joint likelihood ratio can more accurately reflect the real disease risk level of the corresponding area after considering environmental factors and small sample cases. Finally, the risk probability distribution can be constructed based on the basic risk parameter and the actual joint likelihood ratio. The risk probability distribution describes the likelihood of disease occurrence at different risk levels.

[0103] In one embodiment, the hierarchical prevention and control visualization module includes a region inspection subunit for:

[0104] According to the case probability density in the risk probability distribution, a set of Poisson distribution parameters is constructed to determine the generation parameters of the random case distribution;

[0105] According to the preset total number of Monte Carlo simulations divided into multiple parallel subtasks, and through the graphics processor, the case distribution simulation of each subtask is simulated;

[0106] Based on the generation parameters of the random case distribution, a spatial point process simulation is performed through a Monte Carlo simulation engine to generate a simulated case data set that conforms to the risk probability distribution;

[0107] For the simulated case data set of each subtask, a Gaussian penalty factor is calculated according to the deviation of temperature, precipitation and vegetation index from the historical mean in the environmental data;

[0108] For areas with less than one expected case in each simulated case data set, a gamma distribution compensation algorithm is used for correction to obtain a risk adjustment term;

[0109] Combined with the initial risk value, the Gaussian penalty factor and the gamma function adjustment term, a simulated joint likelihood ratio is generated, and the simulated joint likelihood ratios of each subtask are counted to generate a global simulated joint likelihood ratio set;

[0110] According to the actual joint likelihood ratio in the global simulation joint likelihood ratio set and the risk probability distribution, a corrected p-value calculation process is performed to obtain a corrected p-value;

[0111] If the corrected p-value is less than a preset threshold, it is determined that the corresponding scanning area is a significant disease aggregation area.

[0112] Specifically, the case probability density reflects the possibility of case occurrence in different regions. Poisson distribution is often used to describe the number of random events occurring within a certain time or space range. This subunit can calculate the parameter λ of Poisson distribution, i.e., the average number of events occurring per unit time or unit area, by analyzing the case probability density, combining the area of the region, the number of animals in stock, and other related factors, and then construct a set of Poisson distribution parameters for each dynamic scanning area to determine the generation parameters of random case distribution. Monte Carlo simulation is a method of simulating the behavior of complex systems through random sampling. In this embodiment, Monte Carlo simulation can simulate the distribution of cases under different conditions to assess whether the actual observed case distribution has statistical significance of aggregation.

[0113] Illustratively, assuming that the preset total number of Monte Carlo simulations is N, the subunit uniformly distributes N simulations into multiple parallel subtasks. For example, it is divided into M subtasks, and each subtask undertakes N / M simulations. Subsequently, the case distribution simulation of each subtask can be performed with the parallel computing capability of a graphics processing unit, which can greatly shorten the time required for simulation compared to a traditional central processing unit. In each subtask, a series of random case distribution samples conforming to Poisson distribution can be generated by a random number generator according to the determined random case distribution generation parameters, and each sample represents a simulated result of a possible case distribution in the region. Spatial point process simulation takes into account the characteristics of geographic space and the mutual relationship between different positions, and can simulate the occurrence position of random events such as animal disease cases in the spatial domain. During the simulation process, the Monte Carlo simulation engine can randomly generate the position coordinates of cases in each dynamic scanning area according to the set of Poisson distribution parameters, and combine geographic feature data such as river buffer, road network topology, etc., so that the generated case positions are more consistent with the actual epidemic transmission scenario. Through multiple simulations, a large number of simulated case data sets conforming to the risk probability distribution can be generated. The simulated case data set contains the spatial distribution information of cases under different simulation conditions, providing rich data samples for subsequent analysis.

[0114] Further, for the simulation case data set generated for each subtask, the subunit also needs to consider the impact of environmental factors on the spread of the epidemic, that is, the deviation of temperature, precipitation, and vegetation index in the environmental data from the historical mean can be used to calculate a Gaussian penalty factor. The Gaussian penalty factor reflects the degree of influence of environmental abnormalities on the distribution of simulated cases. Similarly, in each simulated case data set, the risk value of the small sample area can be modified by the gamma distribution compensation algorithm to obtain a risk adjustment term for the small sample area, which is used for subsequent simulation joint likelihood ratio calculation to more accurately reflect the epidemic risk of these areas. Combined with the initial risk value, the Gaussian penalty factor, and the gamma function adjustment term, the simulation joint likelihood ratio is generated, and the global simulation joint likelihood ratio set is obtained by summarizing. Finally, according to the global simulation joint likelihood ratio set and the actual joint likelihood ratio in the risk probability distribution, the corrected p-value calculation process is performed. The corrected p-value can measure whether the actual observed case distribution is significantly different from the numerous random case distributions obtained by Monte Carlo simulation. Illustratively, the p-value of the actual joint likelihood ratio can be determined by comparing its position in the global simulation joint likelihood ratio set, and comparing it with a preset threshold. The preset threshold can be determined according to the actual monitoring requirements, historical data, and statistical significance level, etc. For example, a common preset threshold is 0.05. If the corrected p-value is less than the preset threshold, it indicates that the actual observed case distribution is significantly different from the randomly simulated case distribution, that is, the corresponding scanning area is most likely a significant epidemic aggregation area, which needs to be paid attention to and corresponding prevention and control measures need to be taken; otherwise, if the corrected p-value is greater than or equal to the preset threshold, the area cannot be determined as a significant epidemic aggregation area, and the distribution of the epidemic in the area may be random fluctuation rather than significant aggregation.

[0115] In one embodiment, the hierarchical prevention and control visualization module includes a prevention and control heat map generation subunit for:

[0116] According to the historical infection rate, the species susceptibility weight is determined, and the frequency statistics processing is performed based on the historical transportation record data to obtain the animal transportation frequency;

[0117] The actual joint likelihood ratio, the species susceptibility weight, and the animal transportation frequency are linearly superimposed to obtain a comprehensive risk index;

[0118] Based on the comprehensive risk index, Sigmoid function mapping processing is performed to obtain a standardized risk value;

[0119] According to the standardized risk value, a hierarchical threshold division processing is performed to generate a hierarchical prevention and control heat map, which includes a high-risk red area, a medium-risk orange area, and a low-risk yellow area.

[0120] Specifically, the historical infection rate reflects the proportion of different species infected with a specific disease in the past period of time. This subunit can use statistical analysis methods, such as calculating the mean, variance, and other statistical quantities of the infection rate of each species, to evaluate the stability of its infection risk based on a large amount of historical infection rate data of different species in various disease scenarios. For species with high and volatile infection rates, a higher species susceptibility weight is assigned, and then the historical infection rate is converted into a weight value that reflects the difference in species susceptibility to the disease. Animal transportation is one of the important ways of disease transmission. The historical transportation records cover the transportation information between different regions in different time periods, including the starting location, destination, species and quantity of transported animals, etc. This subunit can obtain the animal transportation frequency value of each region by counting the number of animal transportation in different regions within a certain period of time. This value reflects the activity level of the region in the animal transportation network, and the higher the transportation frequency, the higher the risk of disease transmission through animal transportation in that region. Linear superposition of the actual joint likelihood ratio, species susceptibility weight, and animal transportation frequency can obtain a comprehensive risk index, which can comprehensively reflect the overall risk level of animal diseases in a certain region.

[0121] In addition, in the present embodiment, the value range of the comprehensive risk index can be relatively wide, and it is difficult to directly compare the comprehensive risk indexes of different regions. Therefore, the comprehensive risk index can be converted into a standardized risk value of a unified scale by Sigmoid function mapping, so that the risk levels of different regions are comparable. And according to the actual prevention and control needs and historical data characteristics, the threshold range of different risk levels can be determined. By comparing the standardized risk value of each region with the set threshold, the risk level to which the region belongs can be determined. Finally, geographic information system technology can be used to visualize the regions of different risk levels on the map. Among them, high-risk areas are marked in red to attract the attention of prevention and control personnel and remind them that the risk of disease transmission in this region is high and strict prevention and control measures need to be taken, such as strengthening epidemic monitoring, limiting animal flow, and conducting comprehensive disinfection; medium-risk areas are represented in orange to indicate that prevention and control personnel need to pay close attention to these areas and appropriately strengthen prevention and control efforts, such as increasing monitoring frequency and carrying out targeted prevention and publicity; low-risk areas are displayed in yellow, indicating that although the risk is relatively low, continuous monitoring and daily prevention and control work are still needed to prevent the risk of disease transmission from rising. Through this hierarchical prevention and control heat map, prevention and control personnel can understand the risk distribution of animal diseases in the entire monitoring area, providing intuitive and accurate basis for formulating scientific and reasonable prevention and control strategies, and greatly improving the efficiency and pertinence of animal disease prevention and control work.

[0122] Based on the same inventive concept, as Figure 2As shown, the embodiments of the present application also provide an animal epidemic disease monitoring data statistical analysis method. The implementation scheme for solving the problem provided by the method is similar to the implementation scheme described in the above system. Therefore, the specific limitations in one or more embodiments of the animal epidemic disease monitoring data statistical analysis method provided below can be referred to the limitations of the animal epidemic disease monitoring data statistical analysis system described above, which will not be repeated here. The method comprises:

[0123] S201: Based on the input case data, environmental data and population data, standardized fusion processing is performed to obtain a spatio-temporal joint data matrix. The case data includes geographical coordinates, time stamp and species code of animal cases. The environmental data includes temperature, precipitation and vegetation index. The population data includes farm density and animal inventory;

[0124] S202: The geographical feature data and population distribution data in the spatio-temporal joint data matrix are input into a dynamic window prediction model for regional scanning shape prediction, and a dynamic scanning region set is output.

[0125] S203: According to the dynamic scanning region set, environmental correction and small sample stabilization processing are performed to obtain a risk probability distribution. The risk probability distribution includes a basic risk parameter and an actual joint likelihood ratio. The basic risk parameter includes an initial risk value, an environmental correction factor and a small sample correction term.

[0126] S204: Based on the risk probability distribution, spatial clustering test is performed by a graphics processor to output a significant epidemic disease clustering area. Weight superposition processing is performed in combination with the environmental correction factor to generate a hierarchical prevention and control heat map. The environmental correction factor includes a temperature deviation coefficient, a precipitation cumulative deviation value and a vegetation index abnormality degree.

[0127] In the above animal epidemic disease monitoring data statistical analysis method, the input case data, environmental data and population data are standardized and fused to form a spatio-temporal joint data matrix, effectively solving the information island problem and providing comprehensive data support for subsequent risk assessment and prevention and control measures. Secondly, the geographical feature data and population distribution data in the spatio-temporal joint data matrix are input into a dynamic window prediction model for regional scanning shape prediction, fully considering the restriction of geographical features on the spread of epidemic diseases and the characteristics of population distribution. Compared with the traditional preset fixed shape window method, it can better fit the spread path and potential risk area distribution of epidemic diseases in the real complex environment, significantly improving the accuracy and dynamic adaptability of epidemic disease risk area prediction. Moreover, according to the dynamic scanning region set, environmental correction and small sample stabilization processing are performed, which can effectively suppress the risk misjudgment that may be caused by environmental abnormal areas, reduce risk fluctuations, and make the generated risk probability distribution more accurately reflect the actual risk situation of epidemic diseases, providing a reliable risk assessment basis for subsequent epidemic disease prevention and control.

[0128] Finally, based on the risk probability distribution, the spatial aggregation test is performed by a graphics processor, and the significant disease aggregation area is output. The environmental correction factor is combined to perform weight superposition processing to generate a hierarchical prevention and control heat map. In this process, the operation efficiency of the spatial aggregation test is significantly improved by the graphics processor, which can quickly and accurately identify the disease aggregation area. The hierarchical prevention and control heat map helps the prevention and control personnel to grasp the key prevention and control area and the risk degree of different areas at a glance, which provides intuitive and scientific decision support for formulating targeted and efficient prevention and control strategies, and greatly improves the accuracy and effectiveness of animal disease prevention and control work.

[0129] In an exemplary embodiment, the present application also provides a computer device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the steps of the animal disease monitoring data statistical analysis system of the present application when executing the computer program. Preferably, a multi-core processor is used to improve the parallel processing capability of the system. The memory provides sufficient temporary storage space to support the running of the program and the processing of data. The memory capacity should be large enough to accommodate a large amount of supply information and computing tasks.

[0130] In an exemplary embodiment, the present application also provides a computer readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the steps of the animal disease monitoring data statistical analysis system of the present application. The computer readable storage medium can include a read-only memory (ROM), a random access memory (RAM), a solid state disk (SSD), or an optical disk. Among them, the random access memory can include a resistance random access memory (ReRAM) and a dynamic random access memory (DRAM).

[0131] The above-described embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent of the application. It should be noted that for ordinary skilled persons in the art, without departing from the concept of the present application, several modifications and improvements can be made, which are all within the protection scope of the present application.

Claims

1. A statistical analysis system for animal disease monitoring data, characterized in that: The system comprises: A multi-source data fusion module is used to perform standardized fusion processing based on the input case data, environmental data, and population data to obtain a spatiotemporal joint data matrix. The case data includes the geographic coordinates, timestamp, and species code of the animal case; the environmental data includes temperature, precipitation, and vegetation index; and the population data includes farm density and animal inventory; Risk probability prediction module, used for: Inputting the geographic feature data and population distribution data in the spatiotemporal joint data matrix into a dynamic window prediction model to perform regional scanning shape prediction, and outputting a dynamic scanning area set; Performing environmental correction and small sample stabilization processing on the set of dynamic scanning areas to obtain a risk probability distribution, wherein the risk probability distribution includes a basic risk parameter and an actual joint likelihood ratio, wherein the basic risk parameter includes an initial risk value, an environmental correction factor, and a small sample correction term; Hierarchical prevention and control visualization module, used for: Based on the risk probability distribution, a spatial clustering test is performed by a graphics processor to output significant disease cluster areas; According to the significant epidemic cluster areas and the environmental correction factors, weighted superposition processing is performed to generate a graded prevention and control heat map. The environmental correction factors include temperature deviation coefficient, precipitation cumulative deviation value and vegetation index anomaly.

2. The method according to claim 1, characterized in that The multi-source data fusion module includes: A data processing subunit, configured to: Performing inverse distance weighted interpolation processing based on the missing geographic coordinates in the input case data to obtain interpolated and completed case data; Normalizing the temperature, precipitation, and vegetation index in the environmental data to obtain normalized environmental data; Performing logarithmic transformation based on the farm density and animal inventory in the population data to obtain logarithmic transformed population data; The data fusion subunit is used to perform weighted fusion processing based on the interpolated case data, the normalized environmental data and the logarithmically transformed population data to obtain the spatiotemporal joint data matrix.

3. The method according to claim 1, characterized in that The risk probability prediction module includes a dynamic window prediction subunit, which is used to: Extract features based on the river buffer zone and road network topology in the geographic feature data to obtain a traversability matrix, wherein the river buffer zone is an area within 500 meters from the river; Performing feature fusion processing on the density gradient in the population data and the accessibility matrix to obtain an input feature matrix, wherein the density gradient is generated by calculating the difference in the animal inventory of adjacent grids; Inputting the input feature matrix into a dynamic window prediction model based on a convolutional neural network to perform sliding window probability prediction and output a probability distribution graph; According to the probability distribution graph, a threshold truncation process is performed to obtain the dynamic scanning area set.

4. The method according to claim 1, wherein The risk probability prediction module further includes a risk probability synthesis subunit, which is used to: Performing a basic likelihood ratio calculation based on the actual number of cases and the expected number of cases within the dynamic scanning area set to obtain the initial risk value, wherein the expected number of cases is generated by regression prediction based on historical epidemic data and animal inventory; Gaussian penalty processing is performed based on the degree of deviation of the temperature, precipitation and vegetation index in the environmental data from the historical mean to obtain the environmental correction factor; performing gamma function stabilization processing on the area where the expected number of cases is less than 1 within the dynamic scanning area set to obtain the small sample correction term; Combining the initial risk value, the environmental correction factor, and the small sample correction term to obtain the basic risk parameter, and performing joint likelihood ratio synthesis processing to generate the actual joint likelihood ratio; The risk probability distribution is constructed using the basic risk parameters and the actual joint likelihood ratio.

5. The method according to claim 4, characterized in that The calculation formula of the environmental correction factor is: Wherein, L is the environmental correction factor, e k Represents the current environmental factors, including temperature, precipitation, vegetation index, μ k is the historical mean of environmental factors, σ k is the historical standard deviation of environmental factors, and k is the total number of environmental factors.

6. The method according to claim 1, characterized in that The hierarchical prevention and control visualization module includes a regional inspection subunit, which is used to: Constructing a Poisson distribution parameter set based on the case probability density in the risk probability distribution and determining the generation parameters of the random case distribution; Dividing the total number of Monte Carlo simulations into multiple parallel subtasks according to a preset number of times, and performing case distribution simulation on each of the subtasks through the graphics processor; Based on the generation parameters of the random case distribution, a spatial point process simulation is performed through a Monte Carlo simulation engine to generate a simulated case data set that conforms to the risk probability distribution; For the simulated case data set of each subtask, a Gaussian penalty factor is calculated according to the degree of deviation of the temperature, the precipitation, and the vegetation index in the environmental data from the historical mean; For each of the simulated case data sets, regions where the expected number of cases is less than 1 are corrected using a gamma distribution compensation algorithm to obtain a risk adjustment term; Combining the initial risk value, the Gaussian penalty factor, and the gamma function adjustment term to generate a simulated joint likelihood ratio, and calculating the simulated joint likelihood ratios of each of the subtasks to generate a global simulated joint likelihood ratio set; Performing a correction p-value calculation process based on the global simulated joint likelihood ratio set and the actual joint likelihood ratio in the risk probability distribution to obtain a corrected p-value; If the corrected p-value is less than a preset threshold, the corresponding scanning area is determined to be the significant disease cluster area.

7. The method according to claim 1, characterized in that The hierarchical prevention and control visualization module includes a prevention and control heat map generation subunit, which is used to: Based on historical infection rates, species susceptibility weights were determined, and frequency statistics were performed based on historical transport record data to obtain animal transport frequencies; The actual joint likelihood ratio, species susceptibility weight and animal transportation frequency are linearly superimposed to obtain a comprehensive risk index; Based on the comprehensive risk index, a Sigmoid function mapping process is performed to obtain a standardized risk value; According to the standardized risk value, a level threshold division process is performed to generate a graded prevention and control heat map, which includes a high-risk red area, a medium-risk orange area and a low-risk yellow area.

8. A statistical analysis method for animal disease monitoring data, characterized in that: The method comprises: Based on the input case data, environmental data, and population data, standardized fusion processing is performed to obtain a spatiotemporal joint data matrix. The case data includes the geographic coordinates, timestamp, and species code of the animal case; the environmental data includes temperature, precipitation, and vegetation index; and the population data includes farm density and animal inventory. Inputting the geographic feature data and population distribution data in the spatiotemporal joint data matrix into a dynamic window prediction model to perform regional scanning shape prediction, and outputting a dynamic scanning area set; Performing environmental correction and small sample stabilization processing on the set of dynamic scanning areas to obtain a risk probability distribution, wherein the risk probability distribution includes a basic risk parameter and an actual joint likelihood ratio, wherein the basic risk parameter includes an initial risk value, an environmental correction factor, and a small sample correction term; Based on the risk probability distribution, a spatial clustering test is performed by a graphics processor to output significant disease cluster areas; According to the significant epidemic cluster areas and the environmental correction factors, weighted superposition processing is performed to generate a graded prevention and control heat map. The environmental correction factors include temperature deviation coefficient, precipitation cumulative deviation value and vegetation index anomaly.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the system according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the system according to any one of claims 1 to 7 are implemented.

Images