A fowl adenovirus time sequence early warning method and system based on multi-source data fusion

Abstract

The application discloses a kind of based on multi-source data fusion's avian adenovirus timing early warning method and system, specifically related to poultry infectious disease monitoring and early warning technical field, for solving the problem that non-specific pathogen detection fluctuation is identified as outbreak precursor and produces systematic false positive early warning;By extracting the pathogen that presents timing mutual exclusion characteristics in the outbreak window in multi-pathogen monitoring data as candidate pathogenic paving pathogen set, the outbreak probability difference of environmental data under different intervals is combined to identify conditional paving pathogen and independent paving pathogen, the coupling phase desynchronization index between intake and timing and daily weight gain timing is identified to identify immune suppression pathogen, according to the paving attribute classification of pathogen and current environmental stress level, determine the differential inhibition weight, after inhibiting processing to multi-pathogen monitoring data, input timing prediction model with environmental timing data together to output avian adenovirus outbreak early warning result, reduce false positive early warning repeatedly triggered due to conditional pathogen accompanied detection.

Description

Technical Field

[0001] This invention relates to the field of poultry infectious disease monitoring and early warning technology, and more specifically, to a method and system for time-series early warning of avian adenovirus based on multi-source data fusion. Background Technology

[0002] Avian adenovirus serotype 4 (FAdV-4) is the main pathogen causing hepatitis-pericardial effusion syndrome in chickens, spreading rapidly and causing high mortality in large-scale poultry farming. To identify outbreak risk before the appearance of obvious clinical symptoms in flocks, existing technologies typically employ a data-driven approach to construct time-series early warning models. This involves integrating viral virulence assessment results, farm environmental data, production data, and multi-pathogen monitoring data, using machine learning algorithms and time-series analysis to predict the probability of FADV-4 outbreaks. The core idea of ​​this approach is to improve the sensitivity and lead time of early warning by expanding the dimensions of data fusion. In existing early warning systems, multi-pathogen monitoring data is considered an important auxiliary input. The basic assumption is that flocks facing FADV-4 infection often show positive detections or abnormal indicators of other pathogens, and these accompanying signals can provide additional informational gain for early warning. However, in real-world poultry farming environments, the co-infection relationships among multiple pathogens are extremely complex. Some immunosuppressive pathogens can pre-infect and weaken the overall immunity of chicken flocks, creating susceptible conditions for subsequent FADV-4 invasion. There is a temporal causal relationship between the two. However, a large number of conditional pathogens or environmental microorganisms only reflect the current overall health fluctuations or sampling biases of the flock. They do not have a stable pathogenic causal relationship with the actual outbreak of FADV-4 and only constitute a statistically significant accompanying phenomenon.

[0003] Existing avian adenovirus time-series early warning methods, when integrating multi-pathogen surveillance data, cannot distinguish between the pathogenic causal time-series relationship between different pathogens and FAdV-4 outbreaks and the purely statistical co-occurrence relationship. This makes it easy for the model to identify fluctuations in the detection of non-specific pathogens as precursor features of FAdV-4 outbreaks, thus generating systematic false positive early warnings in the context of complex co-infections such as pre-existing infections with immunosuppressive pathogens. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method and system for avian adenovirus time-series early warning based on multi-source data fusion to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A time-series early warning method for avian adenovirus based on multi-source data fusion includes the following steps:

[0007] S1: Obtain environmental time-series data and multi-pathogen monitoring time-series data for the same chicken flock;

[0008] S2: Extract the pre-outbreak directional time window from the multi-pathogen surveillance time series data, extract the detection time series relationship between each other pathogen and the remaining pathogens, and screen out pathogens with time series mutual exclusion characteristics as a candidate pathogenic introductory pathogen set;

[0009] S3: Extract environmental time-series data for each pathogen in the candidate pathogenic causative pathogen set during the pathogen detection period and the outbreak period, compare the difference in the outbreak probability of FAdV-4 when the environmental data after the detection of the same pathogen is in different intervals, and identify conditional causative pathogens and independent causative pathogens.

[0010] S4: Obtain feed intake time-series data and daily weight gain time-series data of chicken flocks within the same time period after the detection of each conditional and independent ground pathogen. Calculate the coupling phase desynchronization index between feed intake time-series data and daily weight gain time-series data. Identify pathogens with coupling phase desynchronization index greater than the preset disorder threshold as immunosuppressive pathogens.

[0011] S5: Based on the classification results of immunosuppressive pathogens and their identification as conditional or independent caching pathogens, determine the differential inhibition weight of immunosuppressive pathogen detection signals in subsequent fusion warning.

[0012] S6: Based on the differentiated suppression weight, the multi-pathogen monitoring data is suppressed. The environmental time series data and the suppressed multi-pathogen monitoring data are input into the time series prediction model, and the FAdV-4 outbreak warning result is output.

[0013] Furthermore, S1 includes:

[0014] Environmental time-series data are obtained from the farm's sensor equipment at a fixed collection cycle, including time-series data on temperature and humidity and ammonia concentration.

[0015] Nucleic acid detection time series results of multiple pathogens obtained from samples sent from breeding farms at a fixed sampling period are used as time series data for multi-pathogen monitoring.

[0016] Furthermore, the multi-pathogen monitoring time series data records the positive or negative detection status of each pathogen at each sampling time point.

[0017] Furthermore, S2 includes:

[0018] Using the occurrence time of the FAdV-4 outbreak as the anchor point, a preset time frame is extracted to serve as the pre-outbreak directional time window;

[0019] Within the targeted time window before the outbreak, for each pathogen, the ratio of the intersection size to the union size of the positive detection time points with each other pathogen is calculated as the co-detection overlap.

[0020] Pathogen pairs with a co-detection overlap lower than a preset mutual exclusion threshold are identified as having temporally mutually exclusive characteristics.

[0021] Pathogens that exhibit at least one set of mutually exclusive temporal characteristics will be selected for inclusion in the candidate pathogenic precursor set.

[0022] Furthermore, S3 includes:

[0023] The pathogen detection time period is formed by extending a preset time forward and backward from the positive detection time point of each pathogen.

[0024] Extract time-series data of temperature and humidity and time-series data of ammonia concentration during the period of pathogen detection;

[0025] The environmental time series data within the pathogen detection period were divided into normal environmental intervals and environmental stress intervals.

[0026] The probability of FADV-4 outbreak when environmental time-series data after the detection of the same pathogen are within the normal range is used as the first outbreak probability.

[0027] The probability of FADV-4 outbreak when environmental time-series data after the detection of the same pathogen falls within the environmental stress range is used as the second outbreak probability.

[0028] By comparing the first outbreak probability with the second outbreak probability, pathogens whose difference between the second outbreak probability and the first outbreak probability exceeds a preset difference threshold are identified as conditional precipitating pathogens, while pathogens whose difference does not exceed the preset difference threshold are identified as independent precipitating pathogens.

[0029] Furthermore, S4 includes:

[0030] Extract time-series data of feed intake and daily weight gain within the same period after the detection of each conditional and independent bed pathogen from the farm's production records.

[0031] Phase extraction was performed on the time series data of feed intake and the time series data of daily weight gain to obtain the phase sequence of feed intake and the phase sequence of daily weight gain.

[0032] Calculate the phase difference sequence between the feed intake phase sequence and the daily weight gain phase sequence;

[0033] Determine the standard deviation of the phase difference sequence as the coupling phase desynchronization index;

[0034] Pathogens with a coupling phase desynchronization index greater than a preset disorder threshold are identified as immunosuppressive pathogens.

[0035] Furthermore, S5 includes:

[0036] For immunosuppressive pathogens that are identified as conditional cushioning pathogens, the suppression weight of their detection signal is set as the product of the basic suppression coefficient and the real-time environmental stress correction coefficient.

[0037] For immunosuppressive pathogens that are identified as independent caking pathogens, the inhibition weight of their detection signals is set to a fixed inhibition coefficient.

[0038] Furthermore, the basic suppression coefficient is determined based on the magnitude of the coupling phase desynchronization index exceeding the preset disorder threshold, and the real-time environmental stress correction coefficient is determined based on the degree to which the current environmental time series data deviates from the normal environmental range.

[0039] Furthermore, S6 includes:

[0040] Extract the detection signals corresponding to immunosuppressive pathogens from the time-series data of multi-pathogen surveillance;

[0041] The extracted immunosuppressive pathogen detection signals were weighted and suppressed based on the differential suppression weights to obtain the suppressed immunosuppressive pathogen detection signals.

[0042] The immunosuppressed pathogen detection signal was combined with the detection signals of other pathogens in the multi-pathogen monitoring time series data to obtain the multi-pathogen monitoring time series data after suppression treatment.

[0043] Environmental time-series data and multi-pathogen monitoring time-series data after suppression treatment are input into the time-series prediction model to output the FAdV-4 outbreak early warning result.

[0044] On the other hand, the present invention provides a time-series early warning system for avian adenovirus based on multi-source data fusion, comprising the following modules:

[0045] The data acquisition module is used to acquire environmental time-series data and multi-pathogen monitoring time-series data for the same flock of chickens.

[0046] The pathogen screening module is used to extract the pre-outbreak time window from the multi-pathogen monitoring time series data, extract the detection time series relationship between each other pathogen and the remaining pathogens, and screen out pathogens with time series mutual exclusion characteristics as a candidate pathogenic introductory pathogen set.

[0047] The pathogen classification module is used to extract environmental time-series data of each pathogen in the candidate pathogenic incubation pathogen set during the pathogen detection period and the outbreak period, compare the difference in the outbreak probability of FAdV-4 when the environmental data after the detection of the same pathogen is in different intervals, and identify conditional incubation pathogens and independent incubation pathogens.

[0048] The desynchronization identification module is used to acquire feed intake time-series data and daily weight gain time-series data of chicken flocks within the same time period after the detection of each conditional and independent bed pathogen, calculate the coupling phase desynchronization index between feed intake time-series and daily weight gain time-series, and identify pathogens with coupling phase desynchronization index greater than the preset disorder threshold as immunosuppressive pathogens.

[0049] The weight determination module is used to determine the differential suppression weight of the immunosuppressive pathogen detection signal in the subsequent fusion warning based on the classification results of the immunosuppressive pathogen and its classification as a conditional or independent caching pathogen.

[0050] The early warning output module is used to suppress multi-pathogen monitoring data based on differentiated suppression weights. It inputs environmental time series data and suppressed multi-pathogen monitoring data into the time series prediction model and outputs FAdV-4 outbreak early warning results.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] 1. By sequentially identifying the temporal mutual exclusion characteristics among multiple pathogens, distinguishing the differences in the probability of pathogen outbreaks under different environmental conditions, and assessing the degree of desynchronization of the coupling phase between feed intake and daily weight gain, the accompanying signals that are not causally related to avian adenovirus outbreaks are gradually stripped from the multi-pathogen monitoring data. This identifies immunosuppressive pathogens that substantially disrupt the metabolic order of chicken flocks. Instead of relying on the statistical correlation of all detected pathogens as the input for early warning, the hidden pathogenic grounding structure in the pathogen monitoring data is explicitly extracted. This eliminates the interference of non-specific pathogen detection fluctuations on the early warning model from the data source, ensuring that the multi-pathogen data received by the early warning model has been pre-calibrated for causal attributes and stripped of interference signals.

[0053] 2. Differentiated suppression weights are determined for identified immunosuppressive pathogens based on their pre-existing characteristics and the current level of environmental stress. This allows the early warning model to retain the early warning contribution of conditional pre-existing pathogens that are amplified in conjunction with environmental deterioration, while also compressing historical statistical noise in independent pre-existing pathogen signals that is unrelated to the current environmental state. After the multi-pathogen monitoring data after suppression treatment is fused with environmental time-series data, the outbreak early warning results output by the time-series prediction model can more accurately reflect the true immunosuppressive pre-existing state of the flock and reduce false positive early warnings repeatedly triggered by the detection of conditional pathogens. Attached Figure Description

[0054] Figure 1 This is a flowchart of a time-series early warning method for avian adenovirus based on multi-source data fusion according to the present invention;

[0055] Figure 2 This is a schematic diagram of the structure of a time-series early warning system for avian adenovirus based on multi-source data fusion according to the present invention. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0057] Example 1: Figure 1 This invention presents a time-series early warning method for avian adenovirus based on multi-source data fusion, comprising the following steps:

[0058] S1: Obtain environmental time-series data and multi-pathogen monitoring time-series data for the same chicken flock;

[0059] S2: Extract the pre-outbreak directional time window from the multi-pathogen surveillance time series data, extract the detection time series relationship between each other pathogen and the remaining pathogens, and screen out pathogens with time series mutual exclusion characteristics as a candidate pathogenic introductory pathogen set;

[0060] S3: Extract environmental time-series data for each pathogen in the candidate pathogenic causative pathogen set during the pathogen detection period and the outbreak period, compare the difference in the outbreak probability of FAdV-4 when the environmental data after the detection of the same pathogen is in different intervals, and identify conditional causative pathogens and independent causative pathogens.

[0061] S4: Obtain feed intake time-series data and daily weight gain time-series data of chicken flocks within the same time period after the detection of each conditional and independent ground pathogen. Calculate the coupling phase desynchronization index between feed intake time-series data and daily weight gain time-series data. Identify pathogens with coupling phase desynchronization index greater than the preset disorder threshold as immunosuppressive pathogens.

[0062] S5: Based on the classification results of immunosuppressive pathogens and their identification as conditional or independent caching pathogens, determine the differential inhibition weight of immunosuppressive pathogen detection signals in subsequent fusion warning.

[0063] S6: Based on the differentiated suppression weight, the multi-pathogen monitoring data is suppressed. The environmental time series data and the suppressed multi-pathogen monitoring data are input into the time series prediction model, and the FAdV-4 outbreak warning result is output.

[0064] S1: Obtain environmental time-series data and multi-pathogen surveillance time-series data for the same chicken flock, implemented as follows:

[0065] In large-scale avian adenovirus farms, temperature and humidity sensors and ammonia concentration sensors are deployed. These sensors record environmental parameters within the chicken house at fixed collection intervals. The temperature and humidity sensors are installed along the longitudinal central axis of the chicken house, at a height of 1.5 to 2.0 meters above the ground. Each chicken house has at least three temperature and humidity sampling points, located at the front, middle, and rear sections of the house. The ammonia concentration sensors are installed inside the exhaust vents of the chicken house, and their locations correspond to those of the temperature and humidity sensors. The fixed data collection cycle is set to once every 15 minutes. Each collection outputs a timestamp, a temperature value, a humidity value, and an ammonia concentration value. The temperature sequence, sorted by timestamp, formed by continuous collection constitutes the temperature sequence in the temperature and humidity time-series data; the humidity sequence, sorted by timestamp, forms the humidity sequence; and the ammonia concentration sequence, sorted by timestamp, forms the ammonia concentration time-series data. The temperature and humidity time-series data and the ammonia concentration time-series data together constitute the environmental time-series data. "Same flock" refers to a group of broilers or laying hens raised in the same chicken house and introduced in the same batch. The environmental time-series data is linked to the same flock, and this link is uniquely identified by the chicken house number and batch number.

[0066] The acquisition of multi-pathogen surveillance time-series data is achieved through a regular testing procedure at poultry farms. Farms collect throat or cloacal swabs from the same flock at a fixed sampling cycle, once a week. The sampling time corresponds one-to-one with the time the samples are delivered to the testing institution for nucleic acid testing. For each sampling, a predetermined number of chickens are randomly selected from the same flock. The predetermined number is determined based on the statistical sampling ratio of the total flock size; for example, when the total flock size is over 10,000, 20 to 30 chickens are selected. Swabs are collected from each individual chicken and then combined into a single pooled sample. This pooled sample represents the pathogen-carrying status of the same flock at the current sampling time. After the pooled sample is delivered to the testing institution, the institution performs nucleic acid testing on a predetermined multi-pathogen detection combination, which includes avian adenovirus serotype 4, avian infectious anemia virus, Marek's disease virus, infectious bursal disease virus, avian reovirus, avian leukosis virus, Salmonella, Escherichia coli, and Mycoplasma. The nucleic acid detection method employs multiplex real-time quantitative PCR. Each test outputs a qualitative result of positive or negative for each pathogen in the multi-pathogen detection combination. The positive or negative status of all pathogens at each sampling time point constitutes a record, associated with the sampling timestamp, chicken house number, batch number, and pathogen name. The accumulated records from multiple consecutive sampling cycles, sorted by sampling timestamp, constitute the multi-pathogen monitoring time-series data. The binding relationship between the multi-pathogen monitoring time-series data and the same chicken flock is maintained through the binding relationship between chicken house number and batch number and environmental time-series data, ensuring that the environmental time-series data and the multi-pathogen monitoring time-series data correspond to the same batch group at the chicken flock level.

[0067] The representative temperature value in the environmental time-series data is determined as follows: At fixed 15-minute intervals, the temperature and humidity sensors output Celsius temperatures, rounded to one decimal place. Each of the three temperature and humidity sensors outputs a Celsius temperature value within each fixed interval. The arithmetic mean of these three values ​​is used as the representative temperature value for the same flock at that time. The representative humidity value in the environmental time-series data is determined as follows: The temperature and humidity sensors output a relative humidity percentage value, rounded to the nearest integer (%). Each of the three temperature and humidity sensors outputs a relative humidity percentage value within each fixed interval. The arithmetic mean of these three relative humidity percentage values ​​is used as the representative humidity value for the same flock at that time. The representative value of ammonia concentration is determined as follows: the ammonia concentration value output by the ammonia concentration sensor is in parts per million (ppm), and the ammonia concentration value is rounded to the nearest integer. Each of the three ammonia concentration collection points outputs an ammonia concentration value within each fixed collection period. The arithmetic mean of the three ammonia concentration values ​​is taken as the representative ammonia concentration value for the same flock at that timestamp. The representative values ​​of temperature, humidity, and ammonia concentration are aligned and merged according to the timestamp to form the environmental time-series data for the same flock. The environmental time-series data is stored in tabular form, with columns including timestamp, temperature, humidity, and ammonia concentration. The timestamp format is year-month-day hour:minute, for example, 2024-03-15 08:00.

[0068] The determination of positive or negative detection status in the multi-pathogen surveillance time series data is as follows: In multiplex real-time quantitative PCR detection, specific primers and TaqMan probes are set for each pathogen. The sequences of the specific primers and TaqMan probes are designed based on the conserved gene regions of each pathogen. After the reaction is completed, the cycle threshold is read. The cycle threshold is the number of amplification cycles corresponding to when the amplification signal reaches the preset fluorescence threshold. If the cycle threshold is less than or equal to the preset positive judgment threshold, it is judged as a positive detection. If the cycle threshold is greater than the preset positive judgment threshold or no amplification signal is detected, it is judged as a negative detection. The preset positive judgment threshold is determined according to the positive judgment criteria set in the test kit instructions. Usually, a cycle threshold of 35 is taken as the judgment threshold. The multi-pathogen surveillance time series data is stored in tabular form. The table columns include sampling timestamp, pathogen name, and detection status. The detection status is either positive or negative. The sampling timestamp format is year-month-day, for example, 2024-03-11. In multi-pathogen surveillance time-series data, the detection status of each pathogen is arranged in ascending order by sampling timestamp to form the detection time-series sequence for that pathogen. Sampling times where the detection status of the same pathogen remains negative, changes from negative to positive, or from positive to negative between two adjacent sampling times constitute the detection status change events for that pathogen. The detection time-series sequences of all pathogens in the multi-pathogen surveillance time-series data are merged to form a complete multi-pathogen surveillance time-series dataset for the same chicken flock.

[0069] S2: Extract a targeted time window before the outbreak from the multi-pathogen surveillance time series data, extract the detection time series relationship between each other pathogen and the remaining pathogens, and screen out pathogens with mutually exclusive time series characteristics as a candidate pathogenic introductory pathogen set. This is implemented as follows:

[0070] The timing of FAdV-4 outbreaks was determined by combining clinical diagnostic results recorded by farm veterinarians with laboratory confirmation results. Clinical diagnostic results refer to events within the same flock exhibiting pericardial effusion, hepatomegaly and necrosis, accompanied by a rising daily mortality rate, over a specific time period. Laboratory confirmation results refer to positive FAdV-4-specific nucleic acid tests on tissue samples from affected chickens. The sampling timestamps corresponding to the laboratory confirmation results were traced back to the first point in the clinical diagnostic results where the daily mortality rate began to rise continuously; this point was taken as the timing of the FAdV-4 outbreak. The timing of the FAdV-4 outbreak was recorded in timestamp format with a daily precision. Using the timing of the FAdV-4 outbreak as an anchor point, a preset duration was extracted as a pre-outbreak directional time window. The preset duration was determined based on the average incubation period of FAdV-4 in the flock from infection to the appearance of typical clinical symptoms and the typical pre-infection period of immunosuppressive pathogens. For example, in broiler farming, the average incubation period for FAdV-4 is 3 to 7 days, while the pre-infection cycle for immunosuppressive pathogens such as avian infectious anemia virus (AIV) is 7 to 14 days. The preset duration can be set to 21 days before the outbreak, meaning a 21-day window prior to the occurrence of the FAdV-4 outbreak can be used as the pre-outbreak directional time window. The format of the pre-outbreak directional time window is from the start timestamp to the end timestamp, where the end timestamp is the occurrence time of the FAdV-4 outbreak, and the start timestamp is the date corresponding to the end timestamp shifted forward by the preset duration.

[0071] Within the pre-outbreak targeted time window, records from the multi-pathogen surveillance time series data whose sampling timestamps fall between the start and end timestamps of the pre-outbreak targeted time window are extracted, forming a subset of the pre-outbreak multi-pathogen surveillance time series data. For each pathogen recorded in the pre-outbreak multi-pathogen surveillance time series data subset, the detection status of the pathogen at each sampling time point is sequentially marked as positive or negative. All sampling time points with positive detections are extracted to form the set of positive detection time points for the pathogen within the pre-outbreak targeted time window. For each pathogen recorded in the pre-outbreak multi-pathogen surveillance time series data subset, a pairwise comparison is performed with each other pathogen recorded in the pre-outbreak multi-pathogen surveillance time series data subset. During the pairwise comparison, the intersection size of the current pathogen's positive detection time point set and the positive detection time point set of the comparison target pathogen is calculated. The intersection size refers to the number of overlapping time points in the two positive detection time point sets. Next, the size of the union of the current pathogen's positive detection time point set and the comparison pathogen's positive detection time point set is calculated. The union size refers to the number of all non-repeating sampling time points included in the merged positive detection time point sets. The ratio of the intersection size to the union size is taken as the co-detection overlap between the current pathogen and the comparison pathogen. The co-detection overlap value is greater than or equal to 0 and less than or equal to 1. A value of 0 indicates that the current pathogen and the comparison pathogen were never detected positive at the same sampling time point within the pre-outbreak targeted time window, while a value of 1 indicates that every positive detection of the current pathogen and the comparison pathogen within the pre-outbreak targeted time window occurred at the exact same sampling time point.

[0072] The preset mutual exclusion threshold is set as follows: The overall distribution of co-detection overlap in the same flock or other flocks raised during the same period within a historical breeding cycle is statistically analyzed. The lower quartile of this distribution is taken, or adjusted based on the farm's tolerance for false positive warnings. For example, within a complete breeding cycle, the co-detection overlap value sequence between all pathogen pairs is obtained. After arranging the co-detection overlap value sequence in ascending order, the value at the 25th percentile is taken as the preset mutual exclusion threshold. The preset mutual exclusion threshold ranges from 0.1 to 0.3. The calculated co-detection overlap is compared with the preset mutual exclusion threshold. If the co-detection overlap of a pathogen pair is less than the preset mutual exclusion threshold, the pathogen pair is determined to have temporally mutually exclusive characteristics. The existence of temporally mutually exclusive characteristics indicates that the positive detection times of the two pathogens within the pre-outbreak directional time window are mutually exclusive in time; that is, the two pathogens are rarely detected positive simultaneously, exhibiting an alternating dominance or one suppressing the other. Each pathogen recorded in the pre-outbreak multi-pathogen surveillance time-series data subset is compared pairwise with all other pathogens, and the number of pathogen pairs identified as having mutually exclusive temporal characteristics for each pathogen is counted. For any pathogen, if it forms a mutually exclusive pathogen pair with at least one other pathogen, the pathogen is selected for inclusion in the candidate pathogenic incubation pathogen set. Each pathogen in the candidate pathogenic incubation pathogen set satisfies a mutually exclusive detection time sequence with at least one other pathogen within the pre-outbreak targeted time window. Other pathogens in the pre-outbreak multi-pathogen surveillance time-series data subset that are not selected for inclusion in the candidate pathogenic incubation pathogen set do not proceed to the subsequent pathogenic incubation attribute identification process because they do not exhibit mutually exclusive temporal characteristics with any other pathogen.

[0073] S3: Extract environmental time-series data for each pathogen in the candidate pathogenic incubation pathogen set during the pathogen detection period and the outbreak period. Compare the outbreak probability differences of FAdV-4 when the environmental data after the detection of the same pathogen are in different intervals to identify conditional incubation pathogens and independent incubation pathogens. This is implemented as follows:

[0074] For each pathogen in the candidate pathogenic incubation set, the positive detection time point is determined. The positive detection time points are derived from the sampling timestamps within the set of positive detection time points of pathogens in the pre-outbreak multi-pathogen surveillance time-series data subset. Taking each positive detection time point as the center, a first preset duration is extended forward, and a second preset duration is extended backward, forming a single pathogen detection period corresponding to the positive detection time point. The first and second preset durations are set based on the average window period from infection to detection by nucleic acid testing in the chicken population. The first preset duration is the lower limit of the average window period, and the second preset duration is the upper limit. For example, the average window period from infection to positive nucleic acid testing in pharyngeal or cloacal swabs for chicken infectious anemia virus is 3 to 10 days. Therefore, the first preset duration is set to 3 days, and the second preset duration is set to 10 days, thus covering the time range from 3 days before to 10 days after the positive detection time point. If a pathogen has multiple positive detection time points within a targeted time window before an outbreak, then a corresponding single pathogen detection period is constructed with each positive detection time point as the center, and the union of all single pathogen detection periods constitutes the pathogen detection period of the pathogen.

[0075] For each pathogen in the candidate pathogenic incubation pathogen set, environmental time-series data within the pathogen detection period are extracted. The extraction method is as follows: using the various time intervals included in the pathogen detection period as filtering criteria, all records with timestamps falling within any one time interval are extracted from the environmental time-series data of the same flock, obtaining temperature and humidity time-series data and ammonia concentration time-series data within the pathogen detection period. The temperature and humidity time-series data within the pathogen detection period includes representative temperature and humidity value sequences, and the ammonia concentration time-series data within the pathogen detection period includes representative ammonia concentration value sequences. Simultaneously, environmental time-series data within the outbreak period corresponding to each pathogen in the candidate pathogenic incubation pathogen set are extracted. The outbreak period refers to the time interval centered on the occurrence time of the FAdV-4 outbreak event, extending before and after a third preset duration. The third preset duration is set based on the typical duration of the acute onset period of FAdV-4; for example, the outbreak period is taken as 3 days before the occurrence time of the FAdV-4 outbreak event to 3 days after the occurrence time of the FAdV-4 outbreak event. The outbreak period is used to confirm whether the chicken house environment was under stress when the FAdV-4 outbreak occurred, but the core basis for subsequent probability comparison is the correlation between the environmental data range within the pathogen detection period and whether an outbreak eventually occurred.

[0076] Environmental time-series data within the pathogen detection period were divided into normal environmental ranges and environmental stress ranges. The division method involved evaluating the numerical combinations corresponding to each timestamp in the representative temperature, humidity, and ammonia concentration sequences within the pathogen detection period. If the temperature, humidity, and ammonia concentration values ​​were all within the preset comfortable temperature, humidity, and ammonia concentration ranges, the environmental data for that timestamp was considered to be within the normal environmental range; otherwise, it was considered to be within the environmental stress range. The preset comfortable temperature, humidity, and ammonia concentration ranges were determined based on the physiological comfort zone parameters of broiler or layer chicken breeds. For example, the preset comfortable temperature range was 18°C ​​to 25°C, the preset comfortable humidity range was 50% to 70%, and the preset safe ammonia concentration range was less than or equal to 20 ppm. After completing the point-by-point judgment, the overall environment of each pathogen detection event is classified: if the number of time points in the normal environmental range exceeds 50% of the total number of time points in the single pathogen detection period corresponding to a pathogen detection event, then the environmental data of the pathogen detection event is classified as being in the normal environmental range; if the number of time points in the environmental stress range is greater than or equal to 50%, then the environmental data of the pathogen detection event is classified as being in the environmental stress range.

[0077] The probability of FAdV-4 outbreak when environmental time-series data after the detection of the same pathogen falls within the normal environmental range is used as the first outbreak probability. The statistical method is as follows: For a pathogen in the candidate pathogenic precursor pathogen set, historical batch records of the same chicken flock and historical records of other chicken flocks under the same farming conditions are retrospectively analyzed to extract all historical detection events where the pathogen was detected positive and the environmental data was classified as falling within the normal environmental range. Among these historical detection events, the number of FAdV-4 outbreak events occurring within a preset observation window after pathogen detection is counted. The preset observation window duration is consistent with the preset duration of the pre-outbreak directional time window, for example, set to 21 days. The number of FAdV-4 outbreak events is divided by the total number of historical detection events to obtain the first outbreak probability. The probability of FAdV-4 outbreak when environmental time-series data after the detection of the same pathogen falls within the environmental stress range is used as the second outbreak probability. The statistical method is as follows: extract all historical detection events in which the pathogen was detected positive and the environmental data was classified as being in the environmental stress range, count the number of FAdV-4 outbreak events that occurred within the preset observation window after pathogen detection, divide the number of FAdV-4 outbreak events by the total number of corresponding historical detection events to obtain the second outbreak probability.

[0078] The method for setting the preset difference threshold is determined based on the farm's need to balance false positive and false negative warnings. For example, the optimal dividing point can be determined by using the maximum inter-class variance in historical statistical data, or the corresponding difference threshold can be derived by back-calculating the upper limit of the acceptable underreporting rate for the farm. One setting method is as follows: Arrange the differences between the first and second outbreak probabilities of all pathogens in the candidate pathogenic precursor pathogen set in ascending order, and take the median or upper quartile of the difference sequence as the preset difference threshold, for example, setting the preset difference threshold to 0.2. Compare the second outbreak probability with the first outbreak probability for each pathogen, and calculate the difference between the second and first outbreak probabilities. If the difference is greater than the preset difference threshold, it is determined that the pathogenic precursor effect is synergistically amplified by environmental factors, and the outbreak probability is significantly higher under environmental stress conditions than under normal environmental conditions; the pathogen is then identified as a conditional precursor pathogen. If the difference is less than or equal to the preset difference threshold, it is determined that the pathogenic precursor effect is independent of environmental fluctuations; the pathogen is then identified as an independent precursor pathogen. The identified conditional and independent pathogens and their classification results constitute the processing objects for subsequent steps.

[0079] S4: Obtain feed intake time-series data and daily weight gain time-series data of the flock within the same time period after the detection of each conditional and independent bed pathogen. Calculate the coupling phase desynchronization index between the feed intake time-series and the daily weight gain time-series. Identify pathogens with a coupling phase desynchronization index greater than a preset disorder threshold as immunosuppressive pathogens. The implementation is as follows:

[0080] Time-series data on feed intake and daily weight gain were extracted from the farm's production records for the same period following the detection of each conditional and independent bed pathogen. These production records, stored in the farm management database, include daily records of total feed intake and flock weight. Total feed intake was obtained by measuring the difference between the cumulative feed input and remaining feed at fixed times each day using a feed tower weighing sensor or feed line flow meter. Flock weight was obtained automatically every three days using a chicken scale platform installed in the chicken house or through manual sampling. For each conditional and independent bed pathogen, time-series data on feed intake and daily weight gain were extracted from the farm's production records for the same period corresponding to the pathogen's detection time. The same period refers to a continuous time interval starting from the last positive detection time of the pathogen in the candidate pathogenic bed pathogen set and ending at the time of the FAdV-4 outbreak. The extraction method for feed intake time series data is as follows: within the same time period, the total feed intake of the flock read at fixed time points each day is arranged in ascending order by timestamp to form a daily feed intake sequence, with the unit of daily feed intake being kilograms. The extraction method for daily weight gain time series data is as follows: within the same time period, the difference between two adjacent flock weight weighing data points is divided by the number of days between the two weighings to obtain the estimated daily weight gain value for each chicken. The estimated daily weight gain values ​​corresponding to each weighing cycle are arranged in ascending order by timestamp to form a daily weight gain sequence, with the unit of daily weight gain being grams per chicken per day. The daily feed intake sequence and the daily weight gain sequence together constitute the feed intake time series data and the daily weight gain time series data.

[0081] Phase extraction was performed on the time-series data of feed intake and daily weight gain, respectively, to obtain the feed intake phase sequence and the daily weight gain phase sequence. Phase extraction was implemented using Hilbert transform. The arithmetic mean of all values ​​in the daily feed intake sequence was subtracted from each value in the daily feed intake sequence to obtain the mean-free feed intake sequence. A Hilbert transform was then applied to the mean-free feed intake sequence to obtain an analytical signal sequence of the same length. Each element in the analytical signal sequence is a complex number; the real part of the complex number represents the value at the corresponding time in the mean-free feed intake sequence, and the imaginary part is the result of the Hilbert transform. The phase angle was extracted from each complex element in the analytical signal sequence. The phase angle was obtained by calculating the arctangent of the ratio of the imaginary part of the complex number to its real part, with the unit being radians. The phase angles at each time point were arranged in ascending order of timestamps to form the feed intake phase sequence. The daily weight gain phase sequence is generated as follows: the daily weight gain sequence is de-meaned, and a Hilbert transform is performed on the de-meaned daily weight gain sequence to obtain an analytical signal sequence. The phase angle of each time point is extracted from the analytical signal sequence, and the sequences are arranged in ascending order of timestamps to form the daily weight gain phase sequence. Before de-meaning, if the time resolution of the daily weight gain sequence is lower than that of the feed intake time series data, linear interpolation is used to fill in the daily weight gain sequence to form a daily sequence, and then the same process of de-meaning and Hilbert transform is performed.

[0082] Calculate the phase difference sequence between the feed intake phase sequence and the daily weight gain phase sequence. The timestamps of the feed intake and daily weight gain phase sequences have been aligned to the same time period in the previous steps. For each timestamp, subtract the phase angle corresponding to that timestamp in the feed intake phase sequence from the phase angle corresponding to that timestamp in the daily weight gain phase sequence to obtain the phase difference value for that timestamp. The unit of the phase difference value is radians. If the phase angle in either the feed intake or daily weight gain phase sequence jumps from near positive π radians to near negative π radians at a certain timestamp while the other has not yet completed the jump, then the phase angle that has completed the jump is compensated with 2π radians before calculating the difference value to eliminate the interference of phase entanglement on the difference calculation. The phase difference values ​​of all timestamps are arranged in ascending order of timestamp to form the phase difference sequence. The length of the phase difference sequence is equal to the length of the feed intake and daily weight gain phase sequences.

[0083] The standard deviation of the phase difference sequence is determined as the coupling phase desynchronization index. The standard deviation of the phase difference sequence is calculated as follows: calculate the arithmetic mean of all phase differences in the sequence; calculate the deviation of each phase difference from the arithmetic mean; sum the squares of each deviation and divide by the length of the phase difference sequence minus 1 to obtain the sample variance; the square root of the sample variance is the standard deviation of the phase difference sequence. The physical meaning of the coupling phase desynchronization index is the degree of fluctuation in the phase coupling relationship between feed intake time series and daily weight gain time series. A larger coupling phase desynchronization index indicates a more disordered phase synchronization relationship between feed intake and daily weight gain.

[0084] The preset disorder threshold is set based on the baseline fluctuation level of the phase coupling relationship between feed intake and daily weight gain in the same flock during a historical normal feeding cycle. Several consecutive time periods during which the same flock did not experience an outbreak of FADV-4 are selected. For each consecutive time period, the standard deviation of the phase difference sequence between the feed intake phase sequence and the daily weight gain phase sequence is calculated. The arithmetic mean of the multiple standard deviations is taken as the baseline fluctuation level. The preset disorder threshold is set as the baseline fluctuation level multiplied by a preset multiple; for example, if the preset multiple is 1.5, then the preset disorder threshold is equal to the baseline fluctuation level multiplied by 1.5. The coupled phase desynchronization index is compared with the preset disorder threshold. If the coupled phase desynchronization index corresponding to a certain conditional or independent precipitating pathogen is greater than the preset disorder threshold, the pathogen is identified as an immunosuppressive pathogen.

[0085] S5: Based on the classification results of immunosuppressive pathogens and their determination as conditional or independent caching pathogens, determine the differential suppression weight of immunosuppressive pathogen detection signals in subsequent fusion early warning, and implement it as follows:

[0086] For immunosuppressive pathogens identified as conditional precipitating pathogens, the suppression weight of their detection signals is set as the product of the baseline suppression coefficient and the real-time environmental stress correction coefficient. The baseline suppression coefficient is determined based on the magnitude by which the coupled-phase desynchronization index exceeds a preset disorder threshold. The determination method is as follows: calculate the ratio of the coupled-phase desynchronization index to the preset disorder threshold, and multiply this ratio by a preset baseline suppression base to obtain the baseline suppression coefficient. The preset baseline suppression base is determined based on the lower limit of the distribution of the magnitude by which the coupled-phase desynchronization index of the immunosuppressive pathogen exceeds the preset disorder threshold when a true positive event occurs in the farm's historical early warning data. For example, if the preset baseline suppression base is set to 0.3, when the ratio of the coupled-phase desynchronization index to the preset disorder threshold is 2, the baseline suppression coefficient is 0.6. The value range of the baseline suppression coefficient is controlled between 0 and 1; when the calculated result of the baseline suppression coefficient exceeds 1, it is taken as 1. The real-time environmental stress correction coefficient is determined based on the degree to which the current environmental time-series data deviates from the normal environmental range. The current environmental time-series data refers to the representative values ​​of temperature, humidity, and ammonia concentration in the latest environmental time-series data collected within the current warning period. The degree of deviation is quantified as follows: the temperature deviation value between the current representative temperature value and the median of the preset comfortable temperature range, the humidity deviation value between the current representative humidity value and the median of the preset comfortable humidity range, and the ammonia deviation value between the current representative ammonia concentration value and the upper limit of the preset safe ammonia concentration range are calculated separately. The temperature deviation value is divided by the preset normalized temperature deviation benchmark to obtain the temperature stress component, the humidity deviation value is divided by the preset normalized humidity deviation benchmark to obtain the humidity stress component, and the ammonia deviation value is divided by the preset normalized ammonia deviation benchmark to obtain the ammonia stress component. The maximum value among the temperature stress component, humidity stress component, and ammonia stress component is taken as the real-time environmental stress correction coefficient. The minimum value of the real-time environmental stress correction coefficient is limited to 1.0. The preset temperature deviation normalization benchmark is taken as half the width of the preset temperature comfort range; the preset humidity deviation normalization benchmark is taken as half the width of the preset humidity comfort range; and the preset ammonia deviation normalization benchmark is taken as the upper limit of the preset ammonia concentration safe range. The suppression weight of conditionally triggered pathogens is obtained by multiplying the base suppression coefficient by the real-time environmental stress correction coefficient. If the result of the multiplication exceeds 1, it is taken as 1 as the suppression weight. The suppression weight of conditionally triggered pathogens characterizes the degree of suppression of pathogen detection signals under the synergistic amplification effect of the current environmental stress level. The stronger the environmental stress, the larger the real-time environmental stress correction coefficient, and the greater the suppression weight, resulting in a higher degree of retention of the pathogen detection signal in subsequent fusion warnings.

[0087] For pathogens identified as independent bedridden pathogens among immunosuppressive pathogens, a fixed inhibition coefficient is set for the inhibition weight of their detection signals. The fixed inhibition coefficient is set as follows: The ratio distribution of the coupling phase desynchronization index to a preset disorder threshold corresponding to all immunosuppressive pathogens identified as independent bedridden pathogens in historical breeding batches of the same flock is statistically analyzed. The arithmetic mean of this ratio distribution is multiplied by a preset baseline coefficient to obtain the fixed inhibition coefficient. The preset baseline coefficient is determined based on the farm's overall requirement for preserving independent bedridden pathogen signals; for example, the fixed inhibition coefficient is set to 0.4. The inhibition weight of independent bedridden pathogens does not change with fluctuations in current environmental time-series data.

[0088] The determination of differential inhibition weights allows for different types of inhibition processing logics for conditional and independent bed pathogens: the inhibition weight for conditional bed pathogens dynamically changes with the current environmental stress level; when the chicken house environment suddenly deteriorates, the real-time environmental stress correction coefficient increases, and the retention rate of the conditional bed pathogen detection signal in the fusion warning increases accordingly. The inhibition weight for independent bed pathogens remains constant, relying solely on a fixed proportion of signal compression based on its historical pathogenic bed intensity. The values ​​of the differential inhibition weights are all between 0 and 1, with a weight of 0 indicating complete inhibition of the pathogen detection signal and a weight of 1 indicating complete retention of the pathogen detection signal. These differential inhibition weights are then passed to subsequent steps for inhibiting multi-pathogen monitoring data.

[0089] S6: Based on the differentiated suppression weights, the multi-pathogen surveillance data is suppressed. The environmental time-series data and the suppressed multi-pathogen surveillance data are input into the time-series prediction model, and the FAdV-4 outbreak warning result is output. The implementation is as follows:

[0090] The detection signals corresponding to immunosuppressive pathogens in the multi-pathogen surveillance time-series data were extracted from the same flock. The extraction method was as follows: For each pathogen name recorded in the multi-pathogen surveillance time-series data, if the pathogen name matched the name of an immunosuppressive pathogen, the detection status records for all sampling time points corresponding to that pathogen name were extracted. These extracted detection status records constituted a set of immunosuppressive pathogen detection signals. Each set of immunosuppressive pathogen detection signals contained multiple detection status records arranged in ascending order of sampling timestamps, with each record showing a value of either positive or negative. The detection signals corresponding to other pathogens besides immunosuppressive pathogens in the multi-pathogen surveillance time-series data were retained in the original multi-pathogen surveillance time-series data during the extraction of the immunosuppressive pathogen detection signals.

[0091] The extracted immunosuppressive pathogen detection signals are weighted and suppressed based on differential suppression weights. The weighted suppression process is as follows: for each detection status record in each group of immunosuppressive pathogen detection signals, the original value corresponding to the detection status record is multiplied by the differential suppression weight corresponding to the immunosuppressive pathogen to obtain the suppressed immunosuppressive pathogen detection signal. The conversion rules for the original values ​​are: negative detection is converted to a value of 0, and positive detection is converted to a value of 1. If the original value of a detection status record is 1 and the corresponding differential suppression weight is 0, then the value of the corresponding record after multiplication becomes 0; if the original value of a detection status record is 1 and the corresponding differential suppression weight is 1, then the value of the corresponding record after multiplication remains 1; if the original value of a detection status record is 1 and the corresponding differential suppression weight is greater than 0 and less than 1, then the value of the corresponding record after multiplication is a floating-point number greater than 0 and less than 1. After processing all detection status records in the same group of immunosuppressive pathogen detection signals one by one, the suppressed immunosuppressive pathogen detection signal corresponding to the immunosuppressive pathogen is obtained.

[0092] The suppressed immunosuppressed pathogen detection signals were merged with the detection signals of other pathogens in the multi-pathogen surveillance time series data. The merging method was as follows: each group of suppressed immunosuppressed pathogen detection signals was placed back into its original position in the multi-pathogen surveillance time series data according to the pathogen name corresponding to the immunosuppressed pathogen, replacing the original detection status record without suppression treatment. The detection signals of the other pathogens in the multi-pathogen surveillance time series data that were not replaced retained their original detection status records; negative detections were still recorded as value 0, and positive detections were still recorded as value 1. After merging, a data table with the same structure as the original multi-pathogen surveillance time series data was obtained; this data table represents the suppressed multi-pathogen surveillance time series data.

[0093] Environmental time-series data and suppressed multi-pathogen monitoring time-series data are jointly input into the time-series prediction model. The time-series prediction model adopts a sequence classification model based on a long short-term memory network. The structure of the sequence classification model based on a long short-term memory network includes an input layer, several long short-term memory layers, a fully connected layer, and an output layer. The input feature vector received by the input layer at each time step is composed of the representative values ​​of temperature, humidity, and ammonia concentration in the environmental time-series data for the corresponding time step, and the signal values ​​of each pathogen in the suppressed multi-pathogen monitoring time-series data for the corresponding time step. When the temporal resolution of the multi-pathogen monitoring time-series data is inconsistent with that of the environmental time-series data, the multi-pathogen monitoring time-series data is padded backward according to the sampling timestamp to align with the timestamp of the environmental time-series data. That is, the pathogen signal value corresponding to each time step of the environmental time-series data is the pathogen signal value corresponding to the most recent sampling timestamp of the multi-pathogen monitoring. Before inputting the feature vectors into the Long Short-Term Memory (LSTM) network, the signal values ​​corresponding to each pathogen are normalized. The normalization process involves calculating the maximum and minimum values ​​for the signal value sequence corresponding to each pathogen, subtracting the minimum value from each signal value, and then dividing by the difference between the maximum and minimum values. When the normalization denominator is zero, all signal values ​​are uniformly set to 0. The sequence classification model based on the LSM network is trained using historical batches of chicken flock data. The training samples consist of labeled information from historical outbreak events, including outbreak labels and outbreak time windows. The outbreak label is either "outbreak" or "no outbreak."

[0094] The time-series prediction model outputs a FAdV-4 outbreak warning once per warning period. The warning period is set to once daily, and the data window length input to the time-series prediction model is consistent with the preset duration of the pre-outbreak directional time window, for example, a data window length of 21 days. The output layer of the time-series prediction model outputs an outbreak risk probability value between 0 and 1. This outbreak risk probability value is compared with a preset warning threshold. If the outbreak risk probability value is greater than or equal to the preset warning threshold, a warning result indicating a FAdV-4 outbreak risk is output; if the outbreak risk probability value is less than the preset warning threshold, a warning result indicating no FAdV-4 outbreak risk is detected is output. The FAdV-4 outbreak warning result is pushed to the farm management terminal in the form of a warning message, which includes the warning generation time, warning risk level, and flock identification information.

[0095] The preset warning threshold is determined based on the balance point between the true positive rate and the false positive rate in historical warning backtesting. Backtesting of the time-series prediction model is performed on multiple batches of historical chicken flock data. The outbreak risk probability values ​​output in each warning period are sorted in ascending order and used as candidate thresholds. The true positive rate and false positive rate corresponding to each candidate threshold are calculated. The candidate threshold corresponding to the maximum difference between the true positive rate and 1 minus the false positive rate is used as the preset warning threshold. In scenarios where the true positive rate and false positive rate are equally important, the preset warning threshold typically falls within the range of 0.4 to 0.6, for example, set to 0.5.

[0096] Example 2: Figure 2 A schematic diagram of a time-series early warning system for avian adenovirus based on multi-source data fusion is provided. This system includes the following modules:

[0097] The data acquisition module is used to acquire environmental time-series data and multi-pathogen monitoring time-series data for the same flock of chickens.

[0098] The pathogen screening module is used to extract the pre-outbreak time window from the multi-pathogen monitoring time series data, extract the detection time series relationship between each other pathogen and the remaining pathogens, and screen out pathogens with time series mutual exclusion characteristics as a candidate pathogenic introductory pathogen set.

[0099] The pathogen classification module is used to extract environmental time-series data of each pathogen in the candidate pathogenic incubation pathogen set during the pathogen detection period and the outbreak period, compare the difference in the outbreak probability of FAdV-4 when the environmental data after the detection of the same pathogen is in different intervals, and identify conditional incubation pathogens and independent incubation pathogens.

[0100] The desynchronization identification module is used to acquire feed intake time-series data and daily weight gain time-series data of chicken flocks within the same time period after the detection of each conditional and independent bed pathogen, calculate the coupling phase desynchronization index between feed intake time-series and daily weight gain time-series, and identify pathogens with coupling phase desynchronization index greater than the preset disorder threshold as immunosuppressive pathogens.

[0101] The weight determination module is used to determine the differential suppression weight of the immunosuppressive pathogen detection signal in the subsequent fusion warning based on the classification results of the immunosuppressive pathogen and its classification as a conditional or independent caching pathogen.

[0102] The early warning output module is used to suppress multi-pathogen monitoring data based on differentiated suppression weights. It inputs environmental time series data and suppressed multi-pathogen monitoring data into the time series prediction model and outputs FAdV-4 outbreak early warning results.

[0103] The calculations involved in the embodiments are all dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.

[0104] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0105] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and inventive constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0106] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0107] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0108] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0109] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A time-series early warning method for avian adenovirus based on multi-source data fusion, characterized in that, Includes the following steps: S1: Obtain environmental time-series data and multi-pathogen monitoring time-series data for the same chicken flock; S2: Extract the pre-outbreak directional time window from the multi-pathogen surveillance time series data, extract the detection time series relationship between each other pathogen and the remaining pathogens, and screen out pathogens with time series mutual exclusion characteristics as a candidate pathogenic introductory pathogen set; S3: Extract environmental time-series data for each pathogen in the candidate pathogenic causative pathogen set during the pathogen detection period and the outbreak period, compare the difference in the outbreak probability of FAdV-4 when the environmental data after the detection of the same pathogen is in different intervals, and identify conditional causative pathogens and independent causative pathogens. S4: Obtain feed intake time-series data and daily weight gain time-series data of chicken flocks within the same time period after the detection of each conditional and independent ground pathogen. Calculate the coupling phase desynchronization index between feed intake time-series data and daily weight gain time-series data. Identify pathogens with coupling phase desynchronization index greater than the preset disorder threshold as immunosuppressive pathogens. S5: Based on the classification results of immunosuppressive pathogens and their identification as conditional or independent caching pathogens, determine the differential inhibition weight of immunosuppressive pathogen detection signals in subsequent fusion warning. S6: Based on the differentiated suppression weight, the multi-pathogen monitoring data is suppressed. The environmental time series data and the suppressed multi-pathogen monitoring data are input into the time series prediction model, and the FAdV-4 outbreak warning result is output.

2. The method for time-series early warning of avian adenovirus based on multi-source data fusion according to claim 1, characterized in that, S1 includes: Environmental time-series data are obtained from the farm's sensor equipment at a fixed collection cycle, including time-series data of temperature and humidity and ammonia concentration. Nucleic acid detection time series results of multiple pathogens obtained from samples sent from breeding farms at a fixed sampling period are used as time series data for multi-pathogen monitoring.

3. The method for time-series early warning of avian adenovirus based on multi-source data fusion according to claim 2, characterized in that, in, The multi-pathogen surveillance time series data records the positive or negative detection status of each pathogen at each sampling time point.

4. The avian adenovirus time-series early warning method based on multi-source data fusion according to claim 1, characterized in that, S2 include: Using the occurrence time of the FAdV-4 outbreak as the anchor point, a preset time frame is extracted to serve as the pre-outbreak directional time window; Within the targeted time window before the outbreak, for each pathogen, the ratio of the intersection size to the union size of the positive detection time points with each other pathogen is calculated as the co-detection overlap. Pathogen pairs with a co-detection overlap lower than a preset mutual exclusion threshold are identified as having temporally mutually exclusive characteristics. Pathogens that exhibit at least one set of mutually exclusive temporal characteristics will be selected for inclusion in the candidate pathogenic precursor set.

5. The avian adenovirus time-series early warning method based on multi-source data fusion according to claim 1, characterized in that, S3 include: The pathogen detection time period is formed by extending a preset time forward and backward from the positive detection time point of each pathogen. Extract time-series data of temperature and humidity and time-series data of ammonia concentration during the period of pathogen detection; The environmental time series data within the pathogen detection period were divided into normal environmental intervals and environmental stress intervals. The probability of FAdV-4 outbreak when environmental time-series data after the detection of the same pathogen are within the normal range is used as the first outbreak probability. The probability of FADV-4 outbreak when environmental time-series data after the detection of the same pathogen falls within the environmental stress range is used as the second outbreak probability. By comparing the first outbreak probability with the second outbreak probability, pathogens whose difference between the second outbreak probability and the first outbreak probability exceeds a preset difference threshold are identified as conditional precipitating pathogens, while pathogens whose difference does not exceed the preset difference threshold are identified as independent precipitating pathogens.

6. The method for time-series early warning of avian adenovirus based on multi-source data fusion according to claim 1, characterized in that, S4 includes: Extract time-series data of feed intake and daily weight gain within the same period after the detection of each conditional and independent bed pathogen from the farm's production records. Phase extraction was performed on the time series data of feed intake and the time series data of daily weight gain to obtain the phase sequence of feed intake and the phase sequence of daily weight gain. Calculate the phase difference sequence between the feed intake phase sequence and the daily weight gain phase sequence; Determine the standard deviation of the phase difference sequence as the coupling phase desynchronization index; Pathogens with a coupling phase desynchronization index greater than a preset disorder threshold are identified as immunosuppressive pathogens.

7. The avian adenovirus time-series early warning method based on multi-source data fusion according to claim 1, characterized in that, S5 include: For immunosuppressive pathogens that are identified as conditional cushioning pathogens, the suppression weight of their detection signal is set as the product of the basic suppression coefficient and the real-time environmental stress correction coefficient. For immunosuppressive pathogens that are identified as independent caking pathogens, the inhibition weight of their detection signals is set to a fixed inhibition coefficient.

8. The method for time-series early warning of avian adenovirus based on multi-source data fusion according to claim 7, characterized in that, The basic suppression coefficient is determined based on the magnitude of the coupling phase desynchronization index exceeding the preset disorder threshold, while the real-time environmental stress correction coefficient is determined based on the degree to which the current environmental time series data deviates from the normal environmental range.

9. A time-series early warning method for avian adenovirus based on multi-source data fusion according to claim 1, characterized in that, S6 include: Extract the detection signals corresponding to immunosuppressive pathogens from the time-series data of multi-pathogen surveillance; The extracted immunosuppressive pathogen detection signals were weighted and suppressed based on the differential suppression weights to obtain the suppressed immunosuppressive pathogen detection signals. The immunosuppressed pathogen detection signal was combined with the detection signals of other pathogens in the multi-pathogen monitoring time series data to obtain the multi-pathogen monitoring time series data after suppression treatment. Environmental time-series data and multi-pathogen monitoring time-series data after suppression treatment are input into the time-series prediction model to output the FAdV-4 outbreak early warning result.

10. A time-series early warning system for avian adenovirus based on multi-source data fusion, used to implement the time-series early warning method for avian adenovirus based on multi-source data fusion as described in any one of claims 1-9, characterized in that, Includes the following modules: The data acquisition module is used to acquire environmental time-series data and multi-pathogen monitoring time-series data for the same flock of chickens. The pathogen screening module is used to extract the pre-outbreak time window from the multi-pathogen monitoring time series data, extract the detection time series relationship between each other pathogen and the remaining pathogens, and screen out pathogens with time series mutual exclusion characteristics as a candidate pathogenic introductory pathogen set. The pathogen classification module is used to extract environmental time-series data of each pathogen in the candidate pathogenic incubation pathogen set during the pathogen detection period and the outbreak period, compare the difference in the outbreak probability of FAdV-4 when the environmental data after the detection of the same pathogen is in different intervals, and identify conditional incubation pathogens and independent incubation pathogens. The desynchronization identification module is used to acquire feed intake time-series data and daily weight gain time-series data of chicken flocks within the same time period after the detection of each conditional and independent bed pathogen, calculate the coupling phase desynchronization index between feed intake time-series and daily weight gain time-series, and identify pathogens with coupling phase desynchronization index greater than the preset disorder threshold as immunosuppressive pathogens. The weight determination module is used to determine the differential suppression weight of the immunosuppressive pathogen detection signal in the subsequent fusion warning based on the classification results of the immunosuppressive pathogen and its classification as a conditional or independent caching pathogen. The early warning output module is used to suppress multi-pathogen monitoring data based on differentiated suppression weights. It inputs environmental time series data and suppressed multi-pathogen monitoring data into the time series prediction model and outputs FAdV-4 outbreak early warning results.

Images