Disease trend analysis and early warning system and method based on big data

Abstract

The invention relates to the technical field of medical health big data analysis, and discloses a disease trend analysis and early warning system and method based on big data, and the method comprises the steps: 1, obtaining multi-source medical data through a hospital information platform, carrying out the preprocessing of the data, building a correlation mapping relation, and forming a standardized data warehouse; 2, constructing a spatio-temporal data cube, and forming a three-dimensional relation graph of time, space and diseases; 3, establishing a disease analysis model according to the multiple influence factors, and performing dynamic analysis on the disease trend and the high-incidence diseases based on the disease analysis model; and 4, according to an analysis result, identifying a disease critical area, generating area early warning information and feeding back the area early warning information to a corresponding terminal. According to the method, the disease monitoring accuracy and real-time performance are effectively improved, the early warning capability is improved, resource configuration is fully integrated, and the disease response efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical and health big data analysis, and in particular to a disease trend analysis and early warning system and method based on big data. Background Art

[0002] Traditional disease surveillance systems rely primarily on aggregated reporting from medical institutions and static data analysis. These systems suffer from inherent flaws such as data lag, low early warning accuracy, and insufficient regional analysis. These systems struggle to accurately capture disease transmission patterns, often requiring only immediate action when a disease strikes, resulting in inefficient prevention and control. This is particularly true during large-scale infectious disease outbreaks, where traditional technologies are unable to quantify the specific impact of the disease across regions in real time. This leads to imbalanced resource allocation and delayed responses, severely hampering the emergency response capabilities of public health and regional medical institutions.

[0003] However, the existing technology currently has the following problems:

[0004] First, traditional models typically analyze and process existing data based on an idealized view of the subject. They often treat either the temporal or spatial dimension separately, neglecting the synergistic effects of both. Second, factor analysis often relies on medical personnel's experience and static weighting, making it difficult to adapt to the dynamic transmission characteristics of different diseases. This results in insufficient prediction accuracy, responsiveness, and regional adaptability of existing systems, making them inadequate for modern precision prevention and control. Summary of the Invention

[0005] The present invention aims to provide a disease trend analysis and early warning method based on big data to solve the problems of low detection accuracy, response efficiency and adaptability of existing disease monitoring systems.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] Solution 1: A disease trend analysis and early warning method based on big data. By analyzing patient consultation data and medical records, it identifies disease trends and high-incidence diseases within a certain period and provides regional disease early warnings, including:

[0008] Step 1: Obtain multi-source medical data through the hospital information platform, pre-process the data, establish association mapping relationships, and form a standardized data warehouse;

[0009] Step 2: Construct a spatiotemporal data cube to form a relationship map of the three dimensions of time, space, and disease;

[0010] Step 3: Establish a disease analysis model based on multiple influencing factors, and conduct dynamic analysis of disease trends and high-incidence diseases based on the disease analysis model;

[0011] Step 4: Based on the analysis results, identify the disease critical area and generate regional warning information to feed back to the corresponding terminal.

[0012] Solution 2: A disease trend analysis and early warning system based on big data, applied to the above-mentioned disease trend analysis and early warning method based on big data, includes a data acquisition module, a data preprocessing module, a trend analysis module, an early warning generation module and a visualization display module.

[0013] The principles and advantages of this solution are:

[0014] In existing technologies, traditional data analysis generally treats the analysis subject, such as a medical institution, as a fixed, unified, and unchanging idealized analysis object, ignoring the differences and connections between different analysis subjects. This solution creatively incorporates the impact of regional and environmental changes to conduct dynamic benchmarking analysis of analysis subjects. This fully considers the current disease impact criticality of the analysis subject at this time, location, and state, as well as the medical resources and allocation required at that time, thereby achieving dynamic and flexible resource adjustment.

[0015] In addition, the existing technology often conducts analysis based on the acquired data. This is because the existing technology requires a comprehensive and effective assessment of the macro environment and the overall situation to obtain the overall impact, and the data obtained is mostly static and fixed. Even if the acquired data is analyzed and processed in a timely manner, the data is generated after the event occurs. After processing and analysis, the state reflected by the data always has a lag. In addition, in the existing technology, the impact of the disease is often analyzed and processed as a post-event. Therefore, there is an inherent thinking logic that the analysis of the disease requires obtaining data for analysis and intervention only after the disease occurs. It is believed that such data will be more representative and feasible. This solution focuses on the concrete and precise analysis of a single analysis object in a single area. It is necessary to obtain the real-time impact on the calibrated analysis object at the current moment in order to make effective disease control measures or resource allocation plans in a timely manner.

[0016] Therefore, this solution achieves accurate multi-scale predictions of disease transmission trends by integrating multi-source data such as medical data, environmental parameters, and socioeconomic indicators within the current region in real time, utilizing a spatiotemporal weight matrix to quantify regional differences. This allows for precise predictions within a certain range and time period. This solution first discovered that the impact of environmental factors on disease transmission is not a simple linear superposition, but rather a threshold effect. Therefore, environmental parameters are divided by region, and a three-dimensional relationship map of time, space, and disease is established. Based on this, a disease analysis model is constructed that accounts for multiple factors. Combined with the disease analysis model, disease criticality values are analyzed and predicted for each region, with advance estimates and dynamic adjustments made within the acceptable error range. This improves disease monitoring accuracy while enhancing the model's adaptability. It can also analyze and analyze regional environmental differences, further improving the accuracy of early warnings and providing more tailored early warning solutions.

[0017] Implementing this solution has the following advantages:

[0018] 1. It breaks through the limitations of traditional methods of spatiotemporal dimension separation analysis and effectively improves the accuracy of regional early warning by constructing spatiotemporal data cubes.

[0019] 2. Through hybrid model analysis and simulation, nonlinear coupling of multiple influencing factors is achieved, solving the deviation problem caused by the need for empirical analysis and judgment in existing technologies.

[0020] 3. It effectively built a fully automatic analysis chain from data collection to closed-loop optimization, realized real-time data updates and dynamic trend analysis, effectively distinguished between individual and group predictions, improved monitoring efficiency, and enhanced early warning response efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 Schematic diagram of the process of the disease trend analysis and early warning method based on big data of the present invention;

[0022] Figure 2 This is a structural diagram of the disease trend analysis and early warning system based on big data of the present invention. DETAILED DESCRIPTION

[0023] The following is further described in detail through specific implementation methods:

[0024] Example 1

[0025] The disease trend analysis and early warning system and method based on big data in this embodiment integrates real-time collection of multi-source data, dynamic spatiotemporal weight matrix and multi-factor coupling modeling, breaking through the solidification of analysis objects in traditional disease monitoring. The analyzed data are mostly static post-event data, which makes it difficult for the analysis results to reflect the differences of the analysis objects, and there are problems such as data lag, insufficient early warning accuracy and lack of regional analysis. It realizes multi-scale feature extraction and dynamic propagation simulation of case data to solve the problems of improper resource allocation and slow response in the existing public health field. In this embodiment, the disease trend analysis and early warning method is as shown in the attached figure. Figure 1 As shown, including:

[0026] S1, obtains multi-source medical data through the hospital information platform, and establishes association mapping relationships after pre-processing the data to form a standardized data warehouse.

[0027] In this embodiment, multi-source data is collected through an interface layer accessing systems such as the hospital's HIS system, public health, meteorological bureau APIs, and mobile phone signaling. Distributed crawler technology and API interfaces are used to acquire data. This data includes, but is not limited to, real-time case data (including outpatient records, inpatient medical records, medical reports, test results, basic patient information, and location); environmental data (including temperature, humidity, air quality, and precipitation in the current area); and social data (including population mobility data and medical resource allocation within the current administrative region), ensuring data traceability and integrity.

[0028] The acquired multi-source data is preprocessed, starting with data cleansing to remove duplicate, erroneous, and incomplete data. Standardization is then performed to unify ICD-10 disease codes and standardize time formats. Then, correlation mapping relationships are established between the data, associating the time, location, and disease events with administrative regions to generate a structured spatiotemporal dataset.

[0029] S2, constructs a spatiotemporal data cube to form a relationship map of the three dimensions of time, space and disease.

[0030] In this embodiment, the extracted geographical distance, temporal proximity, regional disease characteristics, etc. are abstracted into calculable weight values to establish a spatiotemporal data cube with time as the X-axis, space as the Y-axis, and disease as the Z-axis. This is to eliminate analysis bias, correct statistical distortion caused by population density and uneven medical resources, and solve the problem of time and space separation analysis. The spatiotemporal data cube is established as follows:

[0031] Spatial dimension processing: Based on the acquired patient address information, the text address is converted into longitude and latitude coordinates through the geocoding API. The longitude and latitude coordinates are mapped to the divided geographic grid to form a spatial index matrix, resulting in a standardized spatial unit dataset for subsequent regional density calculation and hotspot detection.

[0032] In this embodiment, the city is divided into 500*500m grids to form a geographical grid with administrative areas as the scope, and the divided grids are numbered. At the same time, according to the population density in each area obtained from the mobile phone signaling, the size of the corresponding regional grid is adjusted. If the current area is a densely populated area (such as a residential area), it is automatically reduced to 200m; if the current area is a sparsely populated area (such as a suburb), it is enlarged to 1km, so as to flexibly adjust the division of the geographical grid, balance the amount of grid data and reduce computing power consumption, while ensuring the accuracy and real-time performance of the population area as much as possible, so as to avoid averaging processing and ignoring regional differences, and then accurately capture the changing status in each area. When an emergency occurs, only the grid data can be updated, thereby improving processing efficiency and real-time performance.

[0033] Time dimension processing: The original time data is uniformly formatted, multi-level time granularity is set, the case occurrence time is divided according to the sliding time window, a time series mark is formed, and a case event stream with a timestamp is obtained for trend analysis and periodic pattern recognition.

[0034] In this example, multiple time granularities are set, such as year, month, week, day, and hour. The raw time data is then standardized according to the set standards, such as the ISO8601 time standard, sliced at 15-minute granularity, and uniformly formatted. The case occurrence time is divided according to the set sliding time window to form a time series tag.

[0035] Disease dimension processing: Based on the obtained clinical diagnosis text, intelligent text processing is used to convert unstructured diagnostic descriptions into standardized disease classifications, and disease type dimensions are constructed to obtain a structured disease coding matrix for multi-disease association analysis.

[0036] In this embodiment, the unstructured diagnosis description is converted into a standardized disease classification based on the ICD-10 coding system, and the disease type dimension is constructed in combination with the disease ontology tree.

[0037] Spatiotemporal Coupling: Based on the established spatial index and timestamp tags, an initial cube is constructed by splicing three-dimensional tensors (i.e., time, space, and disease). The original case count is then modified using a set dual weight matrix to obtain a spatiotemporal coupled data cube. This outputs a weighted spatiotemporal disease distribution map for precise quantification of regional critical risk.

[0038] In existing technologies, simple statistics can lead to analytical bias due to the natural spatial clustering of medical data. However, using a dual-weight matrix to quantify spatial dependencies upgrades traditional single-space correlation analysis to a spatiotemporal coupling analysis model. This effectively eliminates false hotspots, improves data analysis accuracy, and effectively captures dynamic transmission characteristics. Furthermore, it can combine current environmental conditions for more concrete analysis and processing, facilitating timely adjustments to individual analysis entities and resource allocation.

[0039] The dual weight matrix includes a spatial weight matrix and a temporal weight matrix. Adaptive broadband kernel density estimation is used to determine the optimal domain range. Voronoi diagrams are used to ensure full coverage without overlap, thus constructing a true accessibility matrix.

[0040] The spatial weight matrix expression is:

[0041]

[0042] Where w ij is the spatial weight from region i to region j; d ij is the effective distance between regions i and j; d0 is the characteristic distance; α is the attenuation steepness coefficient; β ij is the cross-regional connectivity correction factor.

[0043] In this embodiment, an asymmetric time window is set based on the characteristics of disease transmission, and the front and back ranges of the analysis period are dynamically adjusted. For example, when analyzing the spread of influenza, due to the short incubation period (1-3 days) and strong infectivity, a long forward window and a short backward window (such as the past 7 days data + the next 3 days forecast) will be set to capture the characteristics of rapid transmission. If it is a chronic disease, because its incubation period is long but the spread is slow, a short front and long back window (past 30 days + next 90 days) is used. At the same time, wavelet analysis is used to extract multi-time scale features, establish dynamic adjustment of attenuation factors, and analyze the long-term laws and short-term mutations of disease trends to detect abnormal signals more promptly. The constructed time weight matrix expression can be

[0044] w t =e -λ·Δt ;

[0045] Where w t is the weight value at the current time point; λ is the decay rate coefficient; and Δt is the interval between the current time and the target time. This constructs a graph with dimensional relationships and can display a heat map of data density.

[0046] This embodiment also includes a dynamic update process, which updates the incremental data to the corresponding cube units to form a dynamically evolving cube sequence, achieving a dynamic update frequency of the weight matrix at the hourly level, and improving the efficiency of spatiotemporal queries.

[0047] S3, establish a disease analysis model based on multiple influencing factors, and conduct dynamic analysis of disease trends and high-incidence diseases based on the disease analysis model.

[0048] In this embodiment, the multiple influencing factors include meteorological factors, air quality, geographical characteristics, population mobility, distribution of medical resources, and socioeconomic levels. The disease analysis model adopts a three-layer modeling architecture. By nonlinearly combining the determined influencing factors according to the disease transmission mechanism, a multi-factor prediction model with weight distribution is formed to obtain an interpretable disease transmission dynamics equation for quantifying the contribution of each factor. In this embodiment, the three-layer architecture is the basic reproduction number estimation layer, the influencing factor coupling layer, and the spatiotemporal modulation layer. Among them, the basic reproduction number estimation layer is used to generate the basic reproduction number. Its basic parameters are estimated from the initial case data, and the basic reproduction number R0 is formed based on historical data, which can be expressed as

[0049]

[0050] Where β is the infection rate; is the recovery rate.

[0051] The influencing factor coupling layer is used to construct a neural network mapping relationship. Based on the acquired influencing factor data, the Spearman correlation coefficient between the factor and the disease rate is calculated, which is a non-parametric statistical indicator of the monotonic correlation between the two variables. 30-50 candidate factors are initially screened based on the coefficient value ρ. In this embodiment, variables with |ρ|>0.3 can be set to be retained. Then, LASSO regression processing is used to eliminate collinearity factors, and the top 10-15 core factors with coefficients ≠0 (weight values ≠0) in the current time period and space are determined for constructing the model input feature set.

[0052] Based on the determined core factors, the factor weights are calculated through the eigenvectors to quantify the contribution of meteorological, environmental, and social factors to the spread of the disease, form a network diagram of the association between factors and diseases, and construct a nonlinear coupling function to characterize the interaction between factors. In this embodiment, the number of nonlinear combinations based on the set disease transmission mechanism can be expressed as

[0053]

[0054] Where a and b are sigmoid parameters, i.e., curve adjustment values.

[0055] The spatiotemporal control layer introduces the spatiotemporal weight matrix w ij At the same time, the attention mechanism is used to dynamically adjust the contribution of factors to output an accurate disease analysis model.

[0056] In this embodiment, after the model is constructed, the model is verified by backtesting with historical data, such as historical data from the previous five years, real-time data verification, and forward-looking predictions, such as for the next three months, to ensure the accuracy and precision of the model.

[0057] According to the disease analysis model established, the short-term warning cycle is set to carry out statistical forecasting in hours, and the long-term warning cycle is carried out statistical forecasting in weeks. In the present embodiment, a 5-year baseline is established using the moving average method, and small changes are identified using the CUSUM control chart. According to the forecast cycle set, the disease trend is dynamically predicted and analyzed, the critical value and high-incidence area of the disease are judged, as short-term forecast (7 days), medium-term forecast (30 days) and long-term forecast (90 days), as well as the high-incidence aggregation area, are formed to analyze the disease trend, form a multi-period forecast curve, achieve 3-7 days early warning advance amount, and improve trend prediction accuracy. The early warning cycle can be dynamically set in conjunction with the current environmental state. When the early warning situation is analyzed and found within the early warning cycle, early warning can be carried out in time and corresponding intervention measures can be formulated, thereby flexibly adjusting the on-duty situation of the analysis subject, defense status, resource allocation and other preparatory measures to improve management and control effectiveness and timeliness. At the same time, this solution uses a dynamic combination of historical data and real-time data to conduct predictive analysis, achieving accurate and concrete predictions within a certain range and time period. During the predictive analysis, possible errors are controlled within a reasonable range to ensure the dual mechanisms of advance prediction and dynamic adjustment, thereby meeting the specific analysis of different analysis objects and improving the overall analysis accuracy and effectiveness.

[0058] S4, based on the analysis results, identifies the critical areas of the disease level and generates regional warning information and feeds it back to the corresponding terminal.

[0059] In this embodiment, a multi-scale prediction system is set up, a dynamic threshold warning is established, and the model output is graded according to the set dynamic threshold rules. For example, if the first-level warning is set to red: R t >1.5 and growth rate >50% / week; Level 2 warning is orange: 1.2 < R t ≤1.5; Level 3 warning is yellow: predicted value> baseline+2δ, where R t is the real-time reproduction number, i.e., the average number of people infected during the current period. In this embodiment, a transmission simulation is established based on the disease criticality risk. Based on current trends and simulation analysis, the critical state of the disease is promptly identified and displayed by region. This clarifies the prevalent diseases and the degree of crisis of each disease in each region, forming a disease hierarchy. This facilitates timely and appropriate resource allocation and effective intervention based on the regional environment, reducing the disease criticality and narrowing the scope of impact.

[0060] At the same time, based on the disease prediction structure, the warning area level is divided according to the geographical area. Combined with the geographical location, the weight of regional specific factors is adjusted. For example, for urban areas, the focus can be on population mobility factors, and its weight can be increased by 20% to adjust the simulation prediction of each disease; for rural areas, the focus can be on environmental media factors, considering the influence of factors such as mosquito density and vegetation density, and increasing the corresponding weight value, thereby forming a spatiotemporal risk level map and outputting a warning classification map, which makes it easier to obtain a trend forecast report with a confidence interval, and feedback to the corresponding terminal to guide the allocation of prevention and control resources.

[0061] In practical applications, this system can be used in hospitals to allocate resources based on outpatient volume forecasts, optimize physician scheduling, or dynamically manage drug inventory. It can also be applied to public health management to monitor the prevalence or spread of diseases in different regions, facilitate early detection and prevention, optimize regional allocation of public resources, and tailor intervention measures to local conditions. Furthermore, it can be used to monitor trends in common diseases based on regional environmental factors and develop effective improvement measures to effectively reduce disease incidence.

[0062] This solution creatively integrates real-time multi-source data collection, a dynamic spatiotemporal weight matrix, and multi-factor coupled modeling, overcoming technical barriers in traditional disease surveillance, such as data lag, inaccurate early warnings, and a lack of regionalized analysis. By constructing a spatiotemporal data cube and using asymmetric time window analysis, it achieves multi-scale feature extraction and dynamic propagation simulation of case data. Combined with wavelet analysis, it accurately captures both long-term patterns and short-term mutations in disease trends, increasing early warning accuracy to over 89% and shortening response time from 5.2 days to 16 hours, effectively addressing the pain points of resource mismatch and delayed response in the public health sector. Regional disparities are quantified through a spatiotemporal coupled weight matrix, and the innovative intelligent analysis chain of "factor screening-dynamic modeling-closed-loop optimization" is pioneered, overcoming the limitations of traditional methods, such as spatiotemporal fragmented analysis and static factor weighting.

[0063] Existing technologies, however, lack the ability of traditional statistical models to handle the nonlinear temporal and spatial correlations of medical data, and factor analysis relies on operational experience, resulting in delayed and crude predictions. However, this solution innovatively discovers that the coupling between environmental factors and disease transmission can be quantified using a dynamic neural network, rather than simple linear superposition. Furthermore, the spatiotemporal weight matrix breaks the traditional understanding of "distance decay," significantly improving early warning accuracy.

[0064] Example 2

[0065] In this embodiment, as shown in the attached Figure 2As shown, a disease trend analysis and early warning system based on big data is provided, which is applied to the above-mentioned disease trend analysis and early warning method based on big data, including a data acquisition module, a data preprocessing module, a trend analysis module, an early warning generation module and a visualization display module.

[0066] The data collection module is used to connect to multi-source data such as hospital HIS systems and public health platforms in real time, and simultaneously access external data such as meteorological, environmental, and population mobility data. It uses a distributed crawler and API dual-channel collection to ensure full data coverage. The data preprocessing module standardizes the collected multi-source data and constructs a spatiotemporal data cube to ensure data quality.

[0067] The trend analysis module is used to establish a disease analysis model based on multiple influencing factors, quantify the weights of multiple factors such as temperature, humidity, and population mobility, conduct dynamic analysis of disease trends, and detect abnormal aggregation based on the SaTScan algorithm.

[0068] The early warning generation module determines criticality based on disease analysis results and issues graded warnings according to set thresholds. It quickly locates warning areas based on the divided regional boundaries, improving warning accuracy while also adapting to disease criticality analysis in different environments. Furthermore, it integrates resources across regions to achieve more flexible resource allocation and deployment, increasing resource utilization while improving overall processing efficiency and reducing false alarm rates and resource allocation errors.

[0069] The visualization display module is used to output disease spatial distribution heat maps, time trend curves, multi-dimensional correlation analysis scatter plots, and graded push of early warning information, so that the disease critical values in each area can be displayed more intuitively, facilitating quick viewing and disease feature analysis, and providing more accurate intervention or prevention measures in a timely manner, reducing resource loss while ensuring the safety of the public environment and the stability of people's livelihood.

[0070] In this embodiment, by docking multiple data and performing preprocessing, the amount of data is reduced while ensuring the complete acquisition of valid data, providing comprehensive and accurate data support for the analysis of disease critical values. At the same time, a spatiotemporal data cube is used to construct a three-dimensional relationship map of time, space and disease, and the influence of the disease is divided and analyzed according to region and time. This can not only ensure the differential analysis of diseases in each region, improve the accuracy and fit of the analysis, but also ensure real-time updating of data and improve the timeliness of disease analysis. Combined with the environmental factors obtained locally, a more accurate simulation analysis is carried out to make disease monitoring more in line with the current environmental status, more conducive to the rational allocation and effective use of resources, reduce resource waste, and improve intervention efficiency, thereby improving the effectiveness and accuracy of disease monitoring as a whole.

[0071] Example 3

[0072] This embodiment also includes an update and feedback module for updating and optimizing the constructed model. In this embodiment, dynamic learning and closed-loop optimization are used to ensure that the model continuously adapts to the changing patterns of disease transmission.

[0073] First, the model performance indicators are monitored through real-time data streams. When it is detected that the prediction deviation exceeds the threshold or the data distribution has drifted significantly, the incremental learning process is automatically triggered. In this embodiment, a two-level optimization is designed based on the characteristics of the medical scenario. First, for short-term data fluctuations, an online random forest algorithm is used to quickly adjust the factor weights to maintain the sensitivity of the model; second, for long-term trend changes, a monthly full model reconstruction is initiated, and the optimal architecture combination is searched through AutoML technology. At the same time, a human-computer collaborative approach is introduced to convert the clinical experts' correction annotations on the warning results into reinforcement learning signals, and two-way interaction is carried out to enable the system to be continuously updated and optimized in actual applications.

[0074] Traditional update methods typically rely on periodic full updates, which are computationally expensive and fail to capture the characteristics of sudden outbreaks in a timely manner. This solution avoids catastrophic forgetting by preserving key spatiotemporal features. It also uses elastically triggered updates to dynamically adjust the learning rate based on data freshness and model confidence. This approach reduces model decay while maintaining hourly response efficiency, effectively improving system monitoring and identification accuracy and ensuring the system's adaptability and ubiquity.

[0075] The above is only an embodiment of the present invention, and the common knowledge such as the specific technical solutions and / or characteristics in the solution are not described in detail here. It should be pointed out that for those skilled in the art, without departing from the technical solution of the present invention, several variations and improvements can be made, which should also be regarded as the scope of protection of the present invention, and these will not affect the effect of the implementation of the present invention and the practicality of the patent. The scope of protection required by this application shall be based on the content of its claims, and the specific implementation methods and other records in the description can be used to interpret the content of the claims.

Claims

1. A disease trend analysis and early warning method based on big data, characterized by: include: Step 1: Obtain multi-source medical data through the hospital information platform, pre-process the data, establish association mapping relationships, and form a standardized data warehouse; Step 2: Construct a spatiotemporal data cube to form a relationship map of the three dimensions of time, space, and disease; Step 3: Establish a disease analysis model based on multiple influencing factors, and conduct dynamic analysis of disease trends and high-incidence diseases based on the disease analysis model; Step 4: Based on the analysis results, identify the disease critical area and generate regional warning information to feed back to the corresponding terminal.

2. The disease trend analysis and early warning method based on big data according to claim 1 is characterized in that: In step 2, the spatiotemporal data cube is created as follows: Spatial dimension processing: Based on the acquired patient address information, the text address is converted into longitude and latitude coordinate information, and the longitude and latitude coordinates are mapped to the divided geographic grid to form a spatial index matrix and obtain a standardized spatial unit data set; Time dimension processing: The original time data is uniformly formatted, multi-level time granularity is set, and the case occurrence time is divided into sliding time windows to form time series tags and obtain a case event stream with timestamps; Disease dimension processing: Based on the obtained clinical diagnosis text, the unstructured diagnosis description is converted into a standardized disease classification, and the disease type dimension is constructed to obtain a structured disease coding matrix; Spatiotemporal coupling: Based on the spatial index and timestamp mark, the initial cube is constructed, and the original number of cases is corrected by the set double weight matrix to obtain the spatiotemporal coupling data cube, and the spatiotemporal disease distribution map with weight labels is output.

3. The disease trend analysis and early warning method based on big data according to claim 2, characterized in that: The dual weight matrix includes a spatial weight matrix and a temporal weight matrix; the spatial weight matrix expression is: Where w ij is the spatial weight from region i to region j; d ij is the effective distance between regions i and j; d0 is the characteristic distance; α is the attenuation steepness coefficient; β ij is the cross-regional connectivity correction factor; The time weight matrix expression is: w t =e -λ·Δt ; Where w t is the weight value at the current time point; λ is the decay rate coefficient; Δt is the interval between the current time and the target time.

4. The disease trend analysis and early warning method based on big data according to claim 2, characterized in that: It also includes a dynamic update process to update the incremental data into the corresponding cube units to form a dynamically evolving cube sequence.

5. The disease trend analysis and early warning method based on big data according to claim 1 is characterized by: The multiple influencing factors include meteorological factors, air quality, geographical characteristics, population mobility, distribution of medical resources and socioeconomic levels.

6. The disease trend analysis and early warning method based on big data according to claim 1, characterized in that: The disease analysis model adopts a three-layer modeling architecture, and forms a multi-factor prediction model with weight distribution by performing nonlinear combination of determined influencing factors according to the disease transmission mechanism.

7. The disease trend analysis and early warning method based on big data according to claim 6, characterized in that: For the multiple influencing factors obtained, the Spearman correlation coefficient between the factors and the disease rate was calculated, and 30-50 candidate factors were preliminarily screened out based on the coefficient values; then regression processing was used to eliminate collinear factors, and the top 10-15 core factors with coefficients ≠ 0 in the current time period and space were determined.

8. A disease trend analysis and early warning system based on big data, characterized by: The method applied to any one of claims 1-7 comprises a data acquisition module, a data preprocessing module, a trend analysis module, an early warning generation module and a visualization display module.

9. The disease trend analysis and early warning system based on big data according to claim 8, characterized in that: The visualization display module is used to output disease spatial distribution heat maps, time trend curves, multi-dimensional correlation analysis scatter plots, and graded push notifications of early warning information.

10. The disease trend analysis and early warning system based on big data according to claim 8, characterized in that: It also includes an update and feedback module for updating and feedback optimization of the constructed model.

Images