Hospital infectious disease identification early warning method and system based on big data and artificial intelligence

Abstract

The invention discloses a hospital infectious disease recognition and early warning method and system based on big data and artificial intelligence, and relates to the technical field of infectious disease recognition and early warning, and the method comprises a data collection module which is used for collecting multi-source data, and the multi-source data comprises information data of a hospital, environment sensor data and migrant bird migration track data; the data processing module distributes data source weights for multi-source data, and the intelligent early warning module is used for generating a self-adaptive early warning threshold value according to standardized data and historical data and triggering graded early warning according to the generated self-adaptive early warning threshold value; a migrant bird migration track and environmental sensor data are fused, for example, a migrant bird migration abnormal track and environmental virus survival rate parameters are linked, a hospital information system and non-medical data are in butt joint through a standardized interface, and a dynamic time warping algorithm is adopted to calibrate time axis offset of multi-source data, so that the migrant bird migration accuracy is improved. The problem that sensor data and medical record time sequences are inconsistent is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of infectious disease identification and early warning, and more specifically, to a hospital infectious disease identification and early warning method and system based on big data and artificial intelligence. Background Art

[0002] The infectious disease identification and early warning system automatically captures patient symptoms, test results, imaging features, and other data by connecting to hospital information systems (HIS, LIS, and PACS), replacing the traditional manual reporting model and addressing the issue of lag. The National Infectious Disease Intelligent Monitoring and Early Warning Pre-Software can extract keywords (such as "fever" and "lymphocytopenia") from electronic medical records in real time and, combined with laboratory pathogen test results, trigger early warnings. For unknown pathogens (such as new Bunyaviruses and drug-resistant strains), dynamic models analyze syndromic data (such as clustered fever cases) to issue risk warnings even without a clear diagnosis. Intelligent infectious disease monitoring and early warning uses electronic medical records (symptom descriptions), testing systems (pathogen detection results), imaging systems (CT image features), and environmental sensors (temperature, humidity / air particulate matter) to connect with hospital information systems through standardized interfaces, achieving "one data source, one collection location" to avoid duplicate data entry. When the number of fever cases in a ward exceeds the historical level for the same period (such as the moving percentile), an early warning is triggered. The geographical distribution and temporal clustering of cases are analyzed to predict the spread of the epidemic. However, in actual use, the existing system mainly relies on structured data from hospital HIS, LIS, PACS and other systems (such as confirmed case reports and test results), and lacks the integration of non-medical data, such as environmental sensors (temperature, humidity, air particulate matter), migratory bird trajectories, etc. This makes it difficult to capture early warning signals (such as syndrome clusters and environmental risks), causing the warning starting point to lag behind the actual transmission cycle. Summary of the Invention

[0003] To solve the above problems, the present invention provides a hospital infectious disease identification and early warning method and system based on big data and artificial intelligence.

[0004] The present invention provides a hospital infectious disease identification and early warning system based on big data and artificial intelligence, comprising a data acquisition module for collecting multi-source data, wherein the multi-source data includes hospital information data, environmental sensor data, and migratory bird migration trajectory data; A data processing module, which is used to calibrate the time axis offset of multi-source data, assign data source weights to the calibrated multi-source data, calculate and transmit standardized data based on the weights of the multi-source data sources; An intelligent early warning module, the intelligent early warning module is used to receive the standardized data and obtain historical data, the historical data being the standardized data before the current time period; An adaptive warning threshold is generated based on the standardized data and historical data, and a graded warning is triggered based on the generated adaptive warning threshold. Specifically, the generated adaptive warning threshold is compared with the standardized data to obtain a graded warning result.

[0005] Preferably, the specific steps of the data processing module calibrating the time axis offset of multi-source data are as follows: Dynamically adjust the hospital's information data and environmental sensor data to compensate for time offset, and calculate the compensation time ΔT based on the hospital's information data and environmental sensor data; The timestamps of the hospital information data and migratory bird migration trajectory data are uniformly added with ΔT to match them with the real-time data of the environmental sensors.

[0006] Preferably, the specific working steps of the data acquisition module are as follows: For hospital information data, it is connected to the hospital information system and the test results and imaging features in the patient's electronic medical record are captured in real time through a standardized interface; Environmental sensor data is collected through the IoT sensor network to collect environmental data inside and outside the hospital, including temperature, humidity, air particulate matter concentration, and ultraviolet intensity. For the migratory trajectory data of migratory birds, first use the formula: , The deviation between the migratory bird migration trajectory and the historical baseline path is calculated as DTW(S,T), where S is the time series of the current migratory bird migration trajectory, and T is the longitude and latitude coordinate sequence recorded by GPS. Similarly, T is the time series of the current migratory bird trajectory, and is the longitude and latitude coordinate sequence recorded by GPS. ; in is the optimal alignment path, consisting of a series of point pairs Composition, which means that the first point and T's Point alignment; A deviation threshold is set in advance. If the deviation DTW(S,T) between the migratory bird migration trajectory and the historical baseline path exceeds the corresponding threshold, the formula , calculate and obtain the virus survival probability P, and use the virus survival probability P as the migratory trajectory data of migratory birds; Where T is the real-time ambient temperature obtained from the environmental sensor data, and H is the real-time ambient humidity obtained from the environmental sensor data; The timestamps of the hospital information data and migratory bird migration trajectory data are uniformly added with ΔT to match them with the real-time data of the environmental sensors.

[0007] Preferably, the specific working steps of the data processing module for assigning data source weights to multi-source data are as follows: Set a data acquisition cycle W, and then calculate the information entropy of each data source based on the entropy weight method, and then according to the formula: , calculate and obtain the weight of each data source ,in For the The entropy value of the day sample, For the The entropy value of the data source reflects the discrete degree of the data under this indicator. Since the number of data sources is 3, The value of is 1 to 3. The value range is 1 to W.

[0008] Preferably, the steps of calculating the information entropy of each data source based on the entropy weight method include the following: First, normalize the data according to the formula: , calculate the normalized value of the jth data source of the i-th day sample ; Then according to the formula , calculated .

[0009] Preferably, the data processing module calculates the standardized data according to the weights of the multiple data sources, and the specific steps include the following: According to the formula: , Calculate the standardized data ,in is the first after Z-Score standardization data sources, from The data contained in each data source is obtained after Z-Score standardization.

[0010] Preferably, the specific working steps of the intelligent early warning module are as follows: Use historical data to calculate static thresholds; According to the formula , get the static threshold ,in For standardized data The historical average of To standardize data The standard deviation of Then the adaptive warning threshold is calculated based on the static threshold To trigger a graded warning.

[0011] Preferably, the specific steps of obtaining the adjustment coefficient k include the following: Set a data acquisition cycle W; For each time point t in the time window W, Assign time decay weights to obtain the adjusted mean and the adjusted standard deviation ; Then according to the formula , calculate and obtain the adjusted adaptive warning threshold , where 2 is the adjustment coefficient; Then, based on the generated adaptive warning threshold To trigger a graded warning.

[0012] Preferably, the adaptive warning threshold generated according to The specific steps to trigger a graded warning include the following: if , it indicates normal fluctuation and is marked as low risk; if , it is marked as medium risk; if Not present If it is within the range, it is marked as high risk.

[0013] The present invention also proposes a hospital infectious disease identification and early warning method based on big data and artificial intelligence, comprising the following steps: Step 1: Integrate hospital information systems, environmental sensor networks, and migratory bird trajectories, and then use a dynamic time warping algorithm to adjust the timing offset of multi-source data; Step 2: Based on the calibrated multi-source data, assign data source weights to the multi-source data, and then calculate the standardized data based on the weights of the multi-source data sources; Step 3: Receive the standardized data and obtain historical data. The historical data is the standardized data before the current time period, and generate an adaptive warning threshold based on the standardized data and the historical data. Step 4: Compare the generated adaptive warning threshold with the standardized data to obtain the graded warning results.

[0014] Beneficial effects: By fusing migratory bird trajectories with environmental sensor data, for example, linking abnormal migratory bird trajectories with environmental virus survival rate parameters, connecting hospital information systems and non-medical data through standardized interfaces, and using dynamic time warping algorithms to calibrate the time axis offset of multi-source data, the problem of inconsistency between sensor data and medical records is solved, and then using the entropy weight method + random forest model to dynamically assign data source weights, the probability of false alarms can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flow chart of the system of the present invention. DETAILED DESCRIPTION

[0016] Application Scenario: Existing systems primarily rely on structured data from hospital HIS / LIS / PACS (such as confirmed case reports and test results) and lack the integration of non-medical data (environmental sensors and migratory bird trajectories). This makes it difficult to capture early signals such as symptom clusters and environmental risks, and the initial warning period lags behind the actual transmission cycle. This technical solution uses multi-source data to deeply integrate migratory bird trajectories (animal host movement paths) with environmental sensor data (temperature, humidity, and air particulate matter concentration). For example, abnormal migratory bird trajectories can be linked to environmental virus survival rate parameters. This solution connects to hospital information systems and non-medical data through standardized interfaces. A dynamic time warping algorithm is used to calibrate the time axis offset of multi-source data to solve the problem of inconsistent timing between sensor data and medical records. Then, the entropy weight method + random forest model is used to dynamically assign data source weights to reduce the probability of false alarms. like Figure 1 The figure shows a hospital infectious disease identification and early warning system based on big data and artificial intelligence, including a data acquisition module, a data processing module, and an intelligent early warning module. The data acquisition module is used to collect multi-source data, where the multi-source data includes hospital information data, environmental sensor data, and migratory bird migration trajectory data. It should be noted that the data processing module provides raw data and shares real-time data streams with the transmission modeling module through an API interface. The data processing module is used to calibrate the time axis offset of the multi-source data, assign data source weights to the calibrated multi-source data, calculate and transmit the standardized data based on the weights of the multi-source data sources; it should be noted that it is used to receive the original multi-source data from the data acquisition module, output the standardized data set to the propagation modeling module, and feedback the data quality assessment results to the early warning module; The intelligent early warning module is used to receive the standardized data and obtain historical data, where the historical data is the standardized data before the current time period. It should be noted that, for example, if the current time period is T, the historical data is the data before T, and both are standardized data. Generate adaptive warning thresholds based on standardized data and historical data, and trigger graded warnings based on the generated adaptive warning thresholds. Specifically, the generated adaptive warning thresholds are compared with the standardized data to obtain graded warning results. It should be noted that the prediction results of the transmission modeling module are received, and key monitoring parameter optimization suggestions are fed back to the data acquisition module. The function of infectious disease identification and warning is realized based on the results of the graded warning. It should also be noted that the traditional solution uses standardized historical data to calculate a static threshold, providing a stable and reliable benchmark for the early warning system. The solution of this technology is based on this static threshold and further calculates an adaptive early warning threshold using standardized multi-source data collected in the current cycle; It can dynamically adjust according to changes in real-time data, meeting the needs of early warning of epidemic infectious diseases within a short period of time. It not only improves the accuracy of early warning, reduces false alarms and missed alarms, but also enhances the system's adaptability and timeliness to different monitoring scenarios; Existing systems primarily rely on structured data from hospital HIS / LIS / PACS (such as confirmed case reports and test results), lacking integration of non-medical data (environmental sensors and migratory bird trajectories). This makes it difficult to capture early signals such as symptom clusters and environmental risks, and the starting point of early warning lags behind the actual transmission cycle. This technical solution uses multi-source data to deeply integrate migratory bird trajectories (animal host movement paths) with environmental sensor data (temperature, humidity, and air particulate matter concentration). For example, abnormal migratory bird trajectories can be linked to environmental virus survival rate parameters. This solution connects to hospital information systems and non-medical data through standardized interfaces. A dynamic time warping algorithm is used to calibrate the time axis offset of multi-source data to solve the problem of inconsistent timing between sensor data and medical records. Then, the entropy weight method + random forest model is used to dynamically assign data source weights to reduce the probability of false alarms. Moreover, the existing system relies on fixed thresholds and cannot adapt to dynamic scenarios such as pathogen mutations, environmental changes, temperature, and humidity, resulting in an increase in false alarm rates. The above-mentioned technical solution adjusts the threshold parameters in real time, which can further reduce the probability of false alarms.

[0017] As an optional embodiment, the specific steps of the data processing module calibrating the time axis offset of multi-source data are as follows: Dynamically adjust the hospital information data and environmental sensor data to compensate for time offset according to the formula: , Where E(t) is the real-time data sequence of the environmental sensor, specifically including temperature, humidity, air particulate matter concentration, and ultraviolet intensity. It should be noted that E(t) is a multidimensional time series, and each time point t corresponds to a vector of environmental parameters. For example: E(t) = [temperature (t), humidity (t), PM2.5 (t), ultraviolet intensity (t)]; is a real-time data sequence of test results and imaging features in the patient's electronic medical record, where is the average time delay of the sample transportation and testing process; it should be noted that the test results in the electronic medical record include the concentration of the original body, white blood cell technology, the image characteristics are CT image characteristics, and the average time delay In this embodiment, the initial value is 3 hours, and the unit is hours; It is also a multidimensional sequence, including laboratory test results or imaging features; L(t+ )=[pathogen concentration (t+ ), white blood cell count (t+ ), CT imaging characteristics (t+ )]; argmin means finding the solution that minimizes the DTW distance. value, To traverse the range, set according to the real-time delay possibility, in this embodiment, it is 0 to 14 hours; The specific calculation steps are: Traversal Candidate values: For example =0,1,...,23 (hours), for each , shift the L sequence right units, accumulate the local alignment distances of all time slices, and select the optimal , the τ with the smallest total distance is Δ; like = 3, the total DTW distance is the smallest, indicating that the laboratory data needs to be shifted right by 3 hours to match the environmental data trend (for example, the pathogen concentration peak occurs 3 hours after the temperature and humidity increase). This is suitable for aligning multi-dimensional sequences of unequal length and asynchronous lengths (such as environmental monitoring and lagged laboratory data). Through flexible time warping, the impact of equipment delays or sampling frequency differences can be eliminated. It should also be noted that DTW is a dynamic time warping algorithm that calculates the minimum cumulative alignment distance between two sequences. DTW constructs a distance matrix between the two sequences and dynamically plans to find the optimal alignment path. Before calculating the DTW distance, multidimensional data is usually normalized (such as Z-score standardization or Min-Max scaling) to eliminate the impact of numerical magnitude caused by dimensional differences. Moreover, the DTW algorithm itself only focuses on the similarity of sequence morphology, rather than the absolute numerical size. Even if the units of sensor data and laboratory data are different, their morphological change trends (such as peak values ​​and fluctuation periods) can still be aligned through methods such as derivative dynamic time warping (DTW) or weighted Euclidean distance. Allowing the sequence to be locally stretched or compressed on the time axis, if the laboratory data lags behind due to detection delay, DTW can flexibly align the morphologically similar parts of the two sequences by traversing different time offsets. , finds the ΔT that minimizes the cumulative DTW distance. For example, when an environmental sensor detects a sudden temperature rise in a ward (optimizing virus survival conditions), and the laboratory reports a positive pathogen result three hours later due to a process delay, the system uses ΔT to align the two data sets, ensuring temporal consistency in subsequent transmission modeling. When the time axes are not aligned, the correlation between laboratory data and sensor data may be obscured by noise (for example, laboratory data with a 3-hour lag is misaligned with the real-time temperature peak). After ΔT calibration, the temporal matching of the two data is improved, the information entropy calculation is more accurate, and the accuracy of the subsequent allocation of data source weights for multi-source data is increased. The timestamps of the hospital information data and migratory bird migration trajectory data are uniformly added with ΔT to match them with the real-time data of the environmental sensors; It should be noted that if the timestamps of the hospital's information data and migratory bird migration trajectory data do not match the real-time data of the environmental sensors, the multi-source data obtained will affect the subsequent calculation results. For example, the adaptive warning threshold is calculated based on the multi-source data of the previous cycle. If the timestamps do not match, it may actually be calculated based on the hospital's information data and migratory bird migration trajectory data of the previous cycle and the real-time data of the environmental sensors of the previous cycle, which will cause calculation deviations. This technical problem can be solved by compensating for time offsets through this technical solution to ensure the accuracy of the multi-source data used for subsequent calculations.

[0018] As an optional embodiment: the specific working steps of the data acquisition module are as follows: For hospital information data, it is connected to the hospital information system and the test results and imaging features in the patient's electronic medical record are captured in real time through a standardized interface; Environmental sensor data is collected through the IoT sensor network to collect environmental data inside and outside the hospital, including temperature, humidity, air particulate matter concentration, and ultraviolet intensity. For the migratory trajectory data of migratory birds, first use the formula: , The deviation between the migratory bird migration trajectory and the historical baseline path is calculated as DTW(S,T), where S is the time series of the current migratory bird migration trajectory, and T is the longitude and latitude coordinate sequence recorded by GPS. Similarly, T is the time series of the current migratory bird trajectory, and is the longitude and latitude coordinate sequence recorded by GPS. ; in is the optimal alignment path, consisting of a series of point pairs Composition, which means that the first point and T's Points are aligned; it should be noted that the path must meet the monotonicity (no backtracking in time sequence) and continuity constraints; is the square of the Euclidean distance, which is used to quantify the spatial deviation between individual point pairs; Dynamic programming is used to find the path with the minimum cumulative distance, allowing the trajectory to be stretched or compressed on the time axis. The cumulative distance reflects the overall similarity of the two trajectories rather than strict point-by-point alignment. A deviation threshold is set in advance. If the deviation DTW(S,T) between the migratory bird migration trajectory and the historical baseline path exceeds the corresponding threshold, the formula , calculate and obtain the virus survival probability P, and use the virus survival probability P as the migratory trajectory data of migratory birds; Where T is the real-time ambient temperature obtained from the environmental sensor data, and H is the real-time ambient humidity obtained from the environmental sensor data. Mapped to the interval from 0 to 1, it represents the probability of virus survival; The timestamps of the hospital information data and migratory bird migration trajectory data are uniformly added with ΔT to match them with the real-time data of the environmental sensors.

[0019] As an optional embodiment, the specific working steps of the data processing module for assigning data source weights to multi-source data are as follows: Set a data acquisition cycle W, and then calculate the information entropy of each data source based on the entropy weight method, and then according to the formula: , calculate and obtain the weight of each data source ,in For the The entropy value of the day sample, For the The entropy value of the data source reflects the discrete degree of the data under this indicator. Since the number of data sources is 3, The value of is 1 to 3. The value range of is 1 to W. It should be noted that Indicates that the model importance is directly superimposed on the entropy weight method weight, and its influence strength is adjusted by 0.5, where 0.5 is the adjustment coefficient. In this embodiment, 0.5 is the basic value and can be adjusted in the actual process; It should also be noted that the molecules is the difference coefficient, which represents the information redundancy of indicator i. The sum of the difference coefficients of all indicators is used for normalization to ensure that the sum of the weights of the entropy weight method is 1; For example, hospital data (j=1) has a small variance after standardization (stable symptom reporting rate) and low information redundancy. The information entropy calculated by the formula is 0.22; Environmental sensors (j=2) collect data at a high frequency (minute-level), resulting in a high entropy value, large fluctuations in temperature and humidity, and sudden peaks in PM2.5. The calculated information entropy result is 0.68; Migratory bird trajectories (j=3) show seasonal periodicity in their migration path deviations and have medium redundancy. The information entropy calculation result is 0.45. A random forest was trained using a disease spread prediction model (the target variable was the case growth rate over the next seven days) to output the feature importance of each data source: hospital data (j=1), with the nucleic acid test positivity rate contributing the most, with a feature importance of 0.52; Environmental sensor (j=2), PM2.5 is strongly correlated with virus survival rate, with a feature importance of 0.28; Migratory bird trajectories (j=3), the deviation of migratory bird paths warns of avian influenza entry, with a feature importance of 0.20; Substituting the above data into the formula, we get the value of j equal to 1. is 0.626; when j is equal to 2 is 0.278, when j is equal to 3 is 0.393; In this embodiment, a sliding window W (window size is 7 days) is used to recalculate .

[0020] As an optional embodiment, the steps of calculating the information entropy of each data source based on the entropy weight method include the following: First, normalize the data according to the formula; , calculate the normalized value of the jth data source of the i-th day sample ; Then according to the formula , calculated Among them, it should be noted that The calculation method and same.

[0021] As an optional embodiment, the data processing module calculates the standardized data according to the weights of the multiple data sources, and the specific steps include the following: According to the formula: , calculate the standardized data ,in is the first after Z-Score standardization data sources, from The data contained in each data source is obtained through Z-Score normalization. It should be noted that Z-Score normalization is achieved by converting the data into a distribution with a mean of 0 and a standard deviation of 1, eliminating different dimensions. The Z-Score is obtained by subtracting the historical mean from the original data value and dividing it by the historical standard deviation.

[0022] As an optional embodiment: the specific working steps of the intelligent early warning module are as follows: Use historical data to calculate static thresholds; According to the formula , get the static threshold ,in For standardized data The historical average of To standardize data The standard deviation of Then the adaptive warning threshold is calculated based on the static threshold To trigger a graded warning.

[0023] It should be noted that static thresholds calculated using standardized historical data provide a stable and reliable benchmark for the early warning system. Based on this static threshold, the adaptive early warning thresholds further calculated can be dynamically adjusted according to changes in real-time data, thereby achieving precise monitoring and early warning of system status. This design not only improves the accuracy of early warnings and reduces false alarms and missed alarms, but also enhances the system's adaptability to different monitoring scenarios. Through a graded early warning mechanism, the intelligent early warning module can provide different levels of alerts based on the severity of the anomaly, helping users take appropriate measures in a timely manner, effectively preventing potential risks, and ensuring stable system operation. In addition, the module's adaptive nature reduces reliance on manual intervention, reduces maintenance costs, and improves the reliability and efficiency of the overall system.

[0024] As an optional embodiment, the specific steps of obtaining the adjustment coefficient k include the following: Set a data acquisition period, W. It should be noted that, based on the business cycle settings (e.g., the default for hospital infection alerts is 7 days), the data within the statistical window is counted. If the data variance within the window exceeds twice the historical mean (sudden fluctuation), the window is shortened to a smaller value (e.g., 3 days) to capture the change. For each time point t in the time window W, Assign time decay weights to obtain the adjusted mean and the adjusted standard deviation ; Then according to the formula , calculate and obtain the adjusted adaptive warning threshold , where 2 is the adjustment coefficient; Then, based on the generated adaptive warning threshold To trigger a graded warning. It should be noted that by dynamically adjusting the time window W, it is possible to effectively cope with sudden fluctuations in data, improve the response speed and adaptability of the warning system, and assign a time-decay weight to the data at each time point t within the time window W, so that recent data has a greater impact on the warning threshold, thereby more accurately reflecting the current trend. By calculating the adjusted mean and standard deviation, combined with the adjustment coefficient k, the calculated adaptive warning threshold can be dynamically adjusted according to changes in real-time data, thereby improving the accuracy of the warning. This design not only reduces false alarms and missed alarms, but also enhances the system's adaptability to different business scenarios. The graded warning mechanism can provide different levels of alerts according to the severity of the anomaly, helping users take corresponding measures in a timely manner, effectively preventing potential risks, and ensuring the stable operation of the system. In addition, the adaptive characteristics of this module reduce dependence on manual intervention, reduce maintenance costs, and improve the reliability and efficiency of the overall system.

[0025] As an optional embodiment: the adaptive warning threshold generated according to The specific steps to trigger a graded warning include the following: if , it indicates normal fluctuation and is marked as low risk; it should be noted that the local mean Reflects the recent data concentration trend. In this example, 1.5 times the standard deviation is used. , 1.5 times the standard deviation It can cover about 87% of the normally distributed data, which is within a reasonable fluctuation range; if , it is marked as medium risk; it should be noted that the data is between the dynamic threshold and the static basic threshold, indicating that it exceeds the local fluctuation but does not reach the global abnormal level; if Not present If it is within the range, it is marked as high risk.

[0026] It should be noted that the data completely exceeds the dynamic threshold range and requires immediate intervention; it should also be noted that the specific steps of triggering graded warnings can evaluate and classify the system status in real time according to the dynamic changes of the adaptive warning threshold. This graded warning mechanism not only improves the accuracy of the warning, but also enhances the system's response capability, allowing users to deal with potential risks in a timely and effective manner and ensure the stable operation of the system. Graded warnings help reduce false alarms and missed alarms, reduce unnecessary waste of resources and manual intervention costs, and improve the reliability and efficiency of the overall system.

[0027] The present invention also proposes a hospital infectious disease identification and early warning method based on big data and artificial intelligence, comprising the following steps: Step 1: Integrate hospital information systems, environmental sensor networks, and migratory bird trajectories, and then use a dynamic time warping algorithm to adjust the timing offset of multi-source data; Step 2: Based on the calibrated multi-source data, assign data source weights to the multi-source data, and then calculate the standardized data based on the weights of the multi-source data sources; Step 3: Receive the standardized data and obtain historical data. The historical data is the standardized data before the current time period, and generate an adaptive warning threshold based on the standardized data and the historical data. Step 4: Compare the generated adaptive warning threshold with the standardized data to obtain the graded warning results.

[0028] The above are only preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions that fall within the scope of protection of the present invention are within the scope of protection of the present invention. It should be pointed out that for ordinary technical personnel in this technical field, certain improvements and modifications that do not depart from the principles of the present invention should also be considered as the scope of protection of this template.

Claims

1. A hospital infectious disease identification and early warning system based on big data and artificial intelligence, characterized by: It includes a data acquisition module, which is used to collect multi-source data, wherein the multi-source data includes hospital information data, environmental sensor data and migratory bird migration trajectory data; A data processing module, which is used to calibrate the time axis offset of multi-source data, assign data source weights to the calibrated multi-source data, calculate and transmit standardized data based on the weights of the multi-source data sources; An intelligent early warning module, the intelligent early warning module is used to receive the standardized data and obtain historical data, the historical data being the standardized data before the current time period; An adaptive warning threshold is generated based on the standardized data and historical data, and a graded warning is triggered based on the generated adaptive warning threshold. Specifically, the generated adaptive warning threshold is compared with the standardized data to obtain a graded warning result.

2. A hospital infectious disease identification and early warning system based on big data and artificial intelligence according to claim 1, characterized in that: The specific steps of the data processing module to calibrate the time axis offset of multi-source data are as follows: Dynamically adjust the hospital's information data and environmental sensor data to compensate for time offset, and calculate the compensation time ΔT based on the hospital's information data and environmental sensor data; The timestamps of the hospital information data and migratory bird migration trajectory data are uniformly added with ΔT to match them with the real-time data of the environmental sensors.

3. The hospital infectious disease identification and early warning system based on big data and artificial intelligence according to claim 1 is characterized in that: The specific working steps of the data acquisition module are as follows: For hospital information data, it is connected to the hospital information system and the test results and imaging features in the patient's electronic medical record are captured in real time through a standardized interface; Environmental sensor data is collected through the IoT sensor network to collect environmental data inside and outside the hospital, including temperature, humidity, air particulate matter concentration, and ultraviolet intensity. For the migratory trajectory data of migratory birds, first use the formula: , calculate the deviation DTW(S,T) between the migratory bird migration trajectory and the historical baseline path, where S is the time series of the current migratory bird migration trajectory, and T is the longitude and latitude coordinate sequence recorded by GPS. Similarly, T is the time series of the current migratory bird trajectory, and is the longitude and latitude coordinate sequence recorded by GPS. ; in is the optimal alignment path, consisting of a series of point pairs Composition, which means that the first point and T's Point alignment; A deviation threshold is set in advance. If the deviation DTW(S,T) between the migratory bird migration trajectory and the historical baseline path exceeds the corresponding threshold, the formula , calculate and obtain the virus survival probability P, and use the virus survival probability P as the migratory trajectory data of migratory birds; Where T is the real-time ambient temperature obtained from the environmental sensor data, and H is the real-time ambient humidity obtained from the environmental sensor data; The timestamps of the hospital information data and migratory bird migration trajectory data are uniformly added with ΔT to match them with the real-time data of the environmental sensors.

4. The hospital infectious disease identification and early warning system based on big data and artificial intelligence according to claim 3 is characterized in that: The specific working steps of the data processing module for assigning data source weights to multi-source data are as follows: Set a data acquisition cycle W, and then calculate the information entropy of each data source based on the entropy weight method, and then according to the formula: , calculate and obtain the weight of each data source ,in For the The entropy value of the day sample, For the The entropy value of the data source reflects the discrete degree of the data under this indicator. Since the number of data sources is 3, The value of is 1 to 3. The value range is 1 to W.

5. The hospital infectious disease identification and early warning system based on big data and artificial intelligence according to claim 4 is characterized in that: The steps for calculating the information entropy of each data source based on the entropy weight method include the following: First, normalize the data according to the formula; , calculate the normalized value of the jth data source of the i-th day sample ; Then according to the formula , calculated .

6. The hospital infectious disease identification and early warning system based on big data and artificial intelligence according to claim 4 is characterized in that: The specific steps of the data processing module calculating the standardized data according to the weights of the multiple data sources include the following: According to the formula: , calculate the standardized data ,in is the first after Z-Score standardization data sources, from The data contained in each data source is obtained after Z-Score standardization.

7. The hospital infectious disease identification and early warning system based on big data and artificial intelligence according to claim 6 is characterized in that: The specific working steps of the intelligent early warning module are as follows: Use historical data to calculate static thresholds; According to the formula , get the static threshold ,in For standardized data The historical average of To standardize data The standard deviation of Then the adaptive warning threshold is calculated based on the static threshold To trigger a graded warning.

8. The hospital infectious disease identification and early warning system based on big data and artificial intelligence according to claim 7 is characterized in that: The specific steps of obtaining the adjustment coefficient k include the following: Set a data acquisition cycle W; For each time point t in the time window W, Assign time decay weights to obtain the adjusted mean and the adjusted standard deviation ; Then according to the formula , calculate and obtain the adjusted adaptive warning threshold , where 2 is the adjustment coefficient; Then, based on the generated adaptive warning threshold To trigger a graded warning.

9. The hospital infectious disease identification and early warning system based on big data and artificial intelligence according to claim 8 is characterized in that: The adaptive warning threshold generated according to The specific steps to trigger a graded warning include the following: if , it indicates normal fluctuation and is marked as low risk; if , it is marked as medium risk; if Not present If it is within the range, it is marked as high risk.

10. A hospital infectious disease identification and early warning method based on big data and artificial intelligence, applicable to a hospital infectious disease identification and early warning system based on big data and artificial intelligence according to any one of claims 1 to 9, characterized in that: The following steps are involved: Step 1: Integrate hospital information systems, environmental sensor networks, and migratory bird trajectories, and then use a dynamic time warping algorithm to adjust the timing offset of multi-source data; Step 2: Based on the calibrated multi-source data, assign data source weights to the multi-source data, and then calculate the standardized data based on the weights of the multi-source data sources; Step 3: Receive the standardized data and obtain historical data. The historical data is the standardized data before the current time period, and generate an adaptive warning threshold based on the standardized data and the historical data. Step 4: Compare the generated adaptive warning threshold with the standardized data to obtain the graded warning results.

Images