Occupational health risk assessment system based on big data analysis
By integrating multi-source data and constructing a multimodal risk assessment model through a big data-based occupational health risk assessment system, the system addresses the lag problem of traditional assessment methods, enabling dynamic assessment and real-time early warning of occupational health risks, and improving the accuracy of risk management and the effectiveness of protection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HUANENG POWER INT INC DALIAN POWER PLANT
- Filing Date
- 2026-04-13
- Publication Date
- 2026-06-12
Smart Images
Figure CN122198658A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of big data and artificial intelligence technology, and specifically to an occupational health risk assessment system based on big data analysis. Background Technology
[0002] Occupational health risk assessment is an important technical means to prevent occupational diseases and protect workers' health. Traditional occupational health risk assessment methods mainly rely on regular inspections, fixed-point environmental monitoring, and expert judgment, which have gradually revealed various limitations when facing complex and ever-changing work environments.
[0003] Traditional occupational health risk assessment practices commonly suffer from a bias of prioritizing inherent risks over management effectiveness. For example, a steel company's dust concentration monitoring data consistently met standards, but due to insufficient employee protective equipment usage and a regular training coverage rate of only 60%, the incidence of pneumoconiosis remained higher than the industry average. The root cause of this problem lies in the fact that existing assessment models often employ single-dimensional, static threshold judgments (such as analyzing only workplace monitoring data). Furthermore, multi-source, heterogeneous data (such as occupational health monitoring records, real-time sensor data, and company training records) are scattered across different systems, lacking effective data fusion and dynamic analysis capabilities. This results in delayed risk assessment results, making it difficult to support real-time risk warnings and precise intervention decisions for enterprises. To address these issues, designing an occupational health risk assessment system based on big data analytics is essential. Summary of the Invention
[0004] This invention provides an occupational health risk assessment system based on big data analysis to solve the above-mentioned problems in the prior art.
[0005] This invention is implemented as follows: an occupational health risk assessment system based on big data analysis, comprising the following modules:
[0006] Data acquisition module: used to integrate multi-source occupational health data, including occupational hazard factor detection data, worker health monitoring data, real-time monitoring data of the work environment, enterprise occupational health management data, and external public data;
[0007] Big Data Processing Module: Employing a distributed storage and computing framework, this module cleans, denoises, normalizes, and extracts features from multi-source heterogeneous data to generate standardized occupational health datasets.
[0008] Multimodal risk assessment module: Based on machine learning algorithms, a dynamic risk assessment model is built, which outputs a comprehensive risk score by integrating essential risk indicators and management risk indicators;
[0009] Dynamic early warning module: Based on risk scores, risk levels are divided into low, medium and high, and risk trend prediction and abnormal operating condition alarms are realized based on real-time data stream.
[0010] As a preferred embodiment, in the multimodal risk assessment module, the inherent risk indicator includes exposure frequency, and the managed risk indicator includes training pass rate;
[0011] The dynamic risk assessment model adopts a weighted fusion strategy, in which the inherent risk has a weight of 30% and the management risk has a weight of 70%.
[0012] As a preferred option, the machine learning algorithm used in the multimodal risk assessment module is an ensemble learning model, which obtains a comprehensive risk score by fusing the outputs of two sub-models, Gradient Boosting Decision Tree (GBDT) and Multilayer Perceptron (MLP), and performing a weighted average.
[0013] As a preferred embodiment, the multimodal risk assessment module adopts an integrated learning and neural network fusion architecture, and achieves dynamic risk assessment through the following steps:
[0014] Feature organization: Standardized data is divided into two modal feature groups, including the essential risk feature group and the management risk feature group; numerical features are normalized and categorical features are encoded.
[0015] Base model construction and training:
[0016] The essential risk prediction sub-model (weight 30%): adopts the gradient boosting decision tree algorithm, specifically XGBoost, and iteratively constructs multiple CART decision trees to capture the nonlinear relationship between exposure factors and health consequences through residual fitting.
[0017] Risk prediction sub-model (70% weight): It adopts a multilayer perceptron neural network, sets multiple hidden layers and applies the ReLU activation function to learn the deep mapping relationship between management indicators and risk suppression effect;
[0018] Weighted fusion strategy: The normalized risk scores output by the two sub-models are linearly weighted and fused.
[0019] As a preferred embodiment, the multimodal risk assessment module further includes:
[0020] Critical Control Point Identification Submodule: Based on real-time monitoring data from the Internet of Things, it uses spatial clustering and anomaly detection algorithms to locate high-exposure areas or operational processes and dynamically adjusts occupational hazard critical control points.
[0021] As a preferred option, the dynamic early warning module pushes tiered early warning information to enterprises, regulatory authorities, and workers through a multi-level linkage mechanism.
[0022] As a preferred embodiment, a security management module is also included, which is used to monitor data access behavior, mark abnormal modification operations with risk indicators, and trigger permission audits.
[0023] As a preferred embodiment, it includes a user terminal and a base station terminal, which are used for data input, result visualization, and interactive query.
[0024] As a preferred embodiment, the system also includes a non-readable storage medium storing a computer program, which, when executed, implements the operational process of the occupational health risk assessment model, including data collection, risk calculation, risk classification, early warning generation, and report output.
[0025] As a preferred embodiment, the model further includes a display module electrically connected to the dynamic early warning module, the display module being used to visually display the risk level, early warning information, and risk trend.
[0026] The present invention has the following advantages:
[0027] The multimodal risk assessment module employs a "GBDT+MLP" integrated learning architecture, using a weighted fusion strategy to achieve synergistic quantification of dual risk dimensions. The intrinsic risk sub-model (XGBoost) iteratively constructs a CART decision tree to capture the nonlinear relationship between structured data such as exposure frequency, intensity, and duration and health consequences, accurately quantifying the objective risks of physical / chemical / biological hazards. The management risk sub-model (MLP) utilizes the nonlinear mapping capabilities of a multilayer perceptron to deeply explore the complex correlation between management indicators such as training pass rates and protective equipment compliance rates and risk mitigation effects, revealing the actual control effectiveness of management measures. The weighted fusion mechanism generates a comprehensive risk score through linear weighted fusion, reflecting both objective exposure risks and highlighting the subjective regulatory value of management measures.
[0028] The dynamic early warning module uses real-time data streams to predict risk trends and alarm for abnormal operating conditions, supporting dynamic classification and real-time updates of risk levels. The model has built-in online learning or periodic fine-tuning functions, and automatically updates parameters through the accumulation of new data or error thresholds to ensure that the model continuously adapts to changes in the operating environment and management measures, achieving closed-loop management of "assessment-early warning-optimization".
[0029] The model outputs a quantitative comprehensive risk score, providing a scientific basis for occupational health management decisions. By allocating the weights of inherent risk and management risk (3:7), it guides enterprises to emphasize the core role of management measures in risk control, promotes the transformation from "passive protection" to "proactive management," and ultimately achieves a systematic reduction in occupational health risks and a substantial decrease in the incidence of occupational diseases. Attached Figure Description
[0030] Figure 1 This is a block diagram of the overall system architecture of the present invention;
[0031] Figure 2 This is a schematic diagram of the structure of the weighted multimodal ensemble model of the present invention;
[0032] Figure 3 This is a flowchart illustrating the workflow of the essential risk prediction sub-model of the present invention.
[0033] Figure 4 This is a flowchart of the management risk prediction sub-model of the present invention. Detailed Implementation
[0034] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0035] Example 1
[0036] like Figures 1 to 4 As shown, the occupational health risk assessment system based on big data analysis includes the following modules:
[0037] Data acquisition module: used to integrate multi-source occupational health data, including occupational hazard factor detection data, worker health monitoring data, real-time monitoring data of the working environment, enterprise occupational health management data, and external public data, including meteorological and regional economic data;
[0038] Big Data Processing Module: Employing a distributed storage and computing framework, this module cleans, denoises, normalizes, and extracts features from multi-source heterogeneous data to generate standardized occupational health datasets.
[0039] Multimodal risk assessment module: Based on machine learning algorithms, a dynamic risk assessment model is built, which outputs a comprehensive risk score by integrating essential risk indicators and management risk indicators;
[0040] Dynamic early warning module: Based on risk scores, risk levels are divided into low, medium and high, and risk trend prediction and abnormal operating condition alarms are realized based on real-time data stream.
[0041] In the multimodal risk assessment module, the essential risk indicator includes exposure frequency, and the management risk indicator includes training pass rate.
[0042] The dynamic risk assessment model adopts a weighted fusion strategy, in which the inherent risk has a weight of 30% and the management risk has a weight of 70%.
[0043] The machine learning algorithm used in the multimodal risk assessment module is an ensemble learning model, which combines the outputs of two sub-models, Gradient Boosting Decision Tree (GBDT) and Multilayer Perceptron (MLP), and performs a weighted average to obtain a comprehensive risk score.
[0044] The multimodal risk assessment module adopts an ensemble learning and neural network fusion architecture to achieve dynamic risk assessment through the following steps:
[0045] Feature organization: Standardized data is divided into two modal feature groups, including the essential risk feature group and the management risk feature group; numerical features are normalized and categorical features are encoded.
[0046] Base model construction and training:
[0047] The essential risk prediction sub-model (weight 30%) adopts the gradient boosting decision tree algorithm, specifically XGBoost, and iteratively constructs multiple CART decision trees to capture the nonlinear relationship between exposure factors and health consequences through residual fitting.
[0048] Risk prediction sub-model (70% weight): It adopts a multilayer perceptron neural network, sets multiple hidden layers and applies the ReLU activation function to learn the deep mapping relationship between management indicators and risk suppression effect;
[0049] Weighted fusion strategy: The normalized risk scores output by the two sub-models are linearly weighted and fused.
[0050] In this embodiment, the dynamic risk assessment model adopts a multimodal architecture that combines ensemble learning and neural networks. Its core objective is to process multi-source heterogeneous data and fuse it according to set weights (30% for intrinsic risk and 70% for management risk).
[0051] The model first organizes the features of the standardized data generated by the big data processing module. The input features are explicitly divided into two modalities:
[0052] The inherent risk characteristics group mainly includes the frequency, intensity (such as dust concentration and noise decibel level) and duration of exposure to chemical, physical and biological hazards.
[0053] The risk management characteristics group mainly includes enterprise management indicators such as training pass rate, compliance rate of protective equipment wearing, occupational health monitoring coverage, and completeness of emergency plans.
[0054] Numerical features are normalized, and categorical features are encoded to ensure the stability of model training.
[0055] The core of the model is a weighted multimodal ensemble model, the structure of which is as follows: Figure 2 As shown.
[0056] Base model selection and training:
[0057] The essential risk prediction sub-model (weight 30%) employs Gradient Boosting Decision Tree (GBDT), such as XGBoost. This algorithm effectively processes structured data by integrating multiple weak learners (decision trees) to capture the complex nonlinear relationship between exposure factors and health consequences, such as the association between exposure frequency and potential health damage.
[0058] The core idea of XGBoost is to iteratively build multiple decision trees (CART) through additive training and a greedy algorithm. The learning objective of each new tree is the residual between the predictions of all previous trees and the true value.
[0059] Management Risk Prediction Sub-model (70% weight): Employs a Multilayer Perceptron (MLP), a basic feedforward neural network. MLPs contain multiple hidden layers and non-linear activation functions (such as ReLU), and are adept at learning and mapping deep, complex patterns between management metrics and risk mitigation effects, such as how training pass rates reduce risk by influencing employee behavior.
[0060] Each sub-model is trained independently using historical data, and its output ( and The normalized risk score is (e.g., between 0 and 1).
[0061] Weighted fusion strategy:
[0062] The integration layer uses a weighted average method. (Essential risk score) Management risk score The scores were assigned weights of 30% and 70% respectively to calculate the overall risk score. :
[0063]
[0064] This weighting process can be a simple linear weighting, or it can introduce an attention mechanism to dynamically fine-tune the weights, enabling the model to adjust its attention to different risk sources according to the specific context.
[0065] The model's dynamism is reflected in its ability to process real-time data streams and make real-time predictions. Once new operational environment monitoring data or management data is input, the model can quickly calculate and update the risk score.
[0066] To achieve consistent accuracy, the model supports online learning or periodic fine-tuning. The system can be set with trigger conditions (such as the accumulation of a certain amount of new data or the prediction error exceeding a threshold) to automatically update the model parameters with the latest data, adapting it to changes in the operating environment and management conditions.
[0067] The workflow of the Essential Risk Prediction Sub-model (GBDT) in the occupational health risk assessment model can be found by referring to [reference needed]. Figure 3 As shown, the specific workflow is as follows.
[0068] Input and output:
[0069] Input: Intrinsic risk characteristics group, mainly including structured data such as exposure frequency, exposure intensity (e.g., dust concentration, noise decibel value), and exposure duration of occupational hazard factors.
[0070] Output: Essential Risk Score This score is usually normalized to a specific range (such as between 0 and 1) so that it can be weighted and integrated with the management risk score later.
[0071] Model training and prediction:
[0072] Training Phase: This XGBoost sub-model is trained using historical, labeled worker health data or expert-assessed risk levels as supervisory signals. It learns the complex mapping relationship between intrinsic risk indicators and historical health outcomes or known risks.
[0073] Prediction Phase: When new work environment monitoring data is received, the trained XGBoost model will automatically calculate an intrinsic risk score based on features such as the current exposure frequency. .
[0074] Weighted fusion:
[0075] Essential Risk Score It is then fed into a weighted fusion layer and compared with the score output by the Management Risk Prediction Submodel (MLP). The scores are then integrated according to predetermined weights to arrive at a comprehensive risk score. .
[0076] MLP (Multi-Level Processing) is a type of feedforward artificial neural network capable of learning and approximating any complex functional relationship through multi-layered structures and nonlinear transformations. It is well-suited for assessing how management measures (such as training and policy implementation) affect ultimate occupational health risks. (Reference) Figure 4 The flowchart illustrates the workflow of MLP in the model.
[0077] The basic structure of an MLP is a network of interconnected layers, consisting of three or more layers:
[0078] Input layer: Each neuron represents a management risk characteristic. The inputs are multiple management indicators such as training pass rate, compliance rate of protective equipment wearing, and occupational health monitoring coverage rate.
[0079] Hidden layers: This is the "brain" of the MLP, and there can be one or more layers. Each neuron is connected to all neurons in the previous layer (this type of layer is also called a fully connected layer or dense layer). The number of hidden layers and the number of neurons they contain are key hyperparameters.
[0080] Output layer: The output layer contains only one neuron, used to output the management risk score. .
[0081] The connections between these layers all have corresponding weights and biases, which are parameters that the model needs to continuously adjust through learning.
[0082] When a set of management metrics (such as [training pass rate = 0.95, wearing compliance rate = 0.88, ...]) is input into the network, the information flows from the input layer through the hidden layers to the output layer. This process mainly occurs in two steps within each neuron:
[0083] Linear weighted summation: Each neuron receives input from all neurons in the previous layer, performs a weighted summation, and adds a bias term, i.e., z = (w1*x1) + (w2*x2) + ... + (wn*xn) + b.
[0084] Nonlinear activation: The linear result z is input into an activation function, such as ReLU, to obtain the final output a = f(z) of the neuron. Activation functions enable neural networks to solve nonlinear problems. Without them, no matter how deep the network, it can only represent linear relationships.
[0085] The model is initially "unknowing," with its weights and biases randomly initialized. It learns through training. The goal of training is to find an optimal set of weights and biases that makes the model's predicted output (managed risk score) more accurate. To be as close to the real situation as possible.
[0086] This process relies on the backpropagation algorithm:
[0087] Calculate the loss: Quantify the difference between the model's current prediction and the true value using a loss function (such as mean squared error).
[0088] Error backpropagation: Using the chain rule from calculus, the loss is propagated backward from the output layer to the input layer, calculating the gradient of the loss function with respect to each weight and bias. The gradient indicates how the parameters should be fine-tuned to reduce the loss.
[0089] Update parameters: Using an optimizer (such as Adam), the weights and biases in the network are iteratively updated based on the calculated gradients. This cycle of "forward propagation to calculate loss - backpropagation to calculate gradient - parameter update" will be repeated continuously until the model converges.
[0090] The impact of management measures on risk is often not a simple linear relationship. MLP can automatically capture and learn complex patterns such as "the risk reduction effect of increasing the training pass rate from 90% to 95% may be much more significant than that from 70% to 75%." On the other hand, MLP can perform advanced nonlinear combinations of different management indicators (such as training pass rate and the completeness of emergency plans) in the hidden layer, thereby assessing the comprehensive governance level's overall risk mitigation effect.
[0091] In summary, the core algorithm of the multimodal risk assessment module is a machine learning model that integrates GBDT and MLP, employing a weighted fusion strategy. It processes intrinsic and managerial risks separately, then fuses them according to preset weights to ultimately output a comprehensive risk score. This model can adapt to real-time data streams, enabling dynamic risk assessment and providing direct, quantitative evidence for subsequent risk level classification and early warning.
[0092] Example 2
[0093] The occupational health risk assessment system based on big data analysis includes all the contents of Example 1. In addition, the multimodal risk assessment module further includes:
[0094] Critical Control Point Identification Submodule: Based on real-time monitoring data from the Internet of Things, it uses spatial clustering and anomaly detection algorithms to locate high-exposure areas or operational processes and dynamically adjusts occupational hazard critical control points.
[0095] The critical control point identification submodule works closely with the risk assessment model. This submodule uses spatial clustering and anomaly detection algorithms (such as DBSCAN) to analyze real-time IoT monitoring data and accurately locate high-exposure areas or abnormal operation links.
[0096] The spatial clustering and anomaly detection algorithms relied upon by the critical control point identification submodule rely on a data-driven approach to automatically identify spatially clustered areas or operational processes with abnormally high risk levels from real-time IoT monitoring data. In occupational health scenarios, the distribution of hazard factors at the work site is often uneven, forming specific "hotspot" areas. Therefore, density-based spatial clustering algorithms are ideal because they can discover dense areas of arbitrary shapes and directly identify points in low-density areas as anomalies (or noise points). Its main advantage is that it does not require pre-specifying the number of clusters and can effectively identify noise points, which can themselves be considered anomalies. This is crucial for identifying high-exposure points that deviate from normal patterns.
[0097] How the algorithm works: DBSCAN defines "density" through two key parameters:
[0098] Neighborhood radius (eps): Using a data point (such as a sensor monitoring point) as the center, search for other points within its surrounding radius.
[0099] Minimum number of points (min_samples): A point is considered a core point only if its neighborhood contains at least this number of other points.
[0100] Based on core points, the algorithm connects high-density regions into clusters using relationships such as density reachability and density accessibility. Points that cannot be reached by any core point are marked as noise points or outliers.
[0101] The application process of this algorithm can be broken down into the following steps:
[0102] Data Preparation and Feature Extraction: After the big data processing module cleans and standardizes the real-time IoT monitoring data, the key control point identification submodule needs to extract features for each monitoring point. These features typically include geospatial coordinates (such as location number within the workshop, GPS coordinates) and multi-dimensional risk-related indicators (such as instantaneous concentration, time-weighted average concentration, exposure duration, peak frequency, etc. of specific occupational hazard factors). These features together constitute the input data for algorithm analysis.
[0103] Perform spatial clustering to identify high-density regions: Input the processed data into the DBSCAN algorithm. The algorithm will output two types of results:
[0104] Clustering: These clusters represent spatially continuous and horizontally similar areas of hazard concentration. A cluster may correspond to a specific workstation, production line, or functional area of a workshop. By analyzing these clusters, the spatial distribution patterns of hazard factors can be visualized.
[0105] Noise points: These are isolated points that do not conform to any high-density area. In the context of occupational health, an isolated, abnormally high reading may indicate a sudden leak or violation of regulations, and also requires close attention.
[0106] Anomaly and Critical Control Point Identification: In this step, clustering results are closely integrated with anomaly detection.
[0107] Intra-cluster analysis: Even within a cluster, the sub-regions with the highest relative risk can be further identified by comparing risk indicators (such as average concentration) at various points within the cluster.
[0108] Noise points are considered anomalies: All points marked as noise by DBSCAN are directly identified as spatial anomalies. These may be "critical control points" that require immediate intervention.
[0109] Dynamic Adjustment: Because IoT data flows in in real time, the algorithm can periodically (e.g., every minute or hour) re-execute cluster analysis. When new monitoring data forms new high-density clusters, or new anomalies appear, the system can dynamically update the list of critical control points, enabling real-time monitoring and early warning of risks.
[0110] By analyzing the spatial location and risk indicators of monitoring points, high-exposure areas can be identified (clustering) and abnormal operating conditions can be alerted (noise point detection) with a single click. This method can automatically and efficiently pinpoint risk sources from massive amounts of real-time data, greatly improving the initiative and accuracy of occupational health risk management.
[0111] Example 3
[0112] The occupational health risk assessment system based on big data analysis includes all the contents of Example 2. In addition, the dynamic early warning module pushes graded early warning information to enterprises, regulatory authorities and workers through a multi-level linkage mechanism.
[0113] It also includes a security management module, which monitors data access behavior, marks abnormal modification operations with risk indicators, and triggers permission audits.
[0114] It includes user terminals and base station terminals, which are used for data input, result visualization, and interactive querying.
[0115] It also includes non-readable storage media containing computer programs, which, when executed, implement the operational process of the occupational health risk assessment model, including data collection, risk calculation, level classification, early warning generation, and report output.
[0116] The model further includes a display module electrically connected to the dynamic early warning module, which is used to visualize the risk level, early warning information and risk trends.
Claims
1. An occupational health risk assessment system based on big data analysis, characterized in that, Includes the following modules: Data acquisition module: used to integrate multi-source occupational health data, including occupational hazard factor detection data, worker health monitoring data, real-time monitoring data of the work environment, enterprise occupational health management data, and external public data; Big Data Processing Module: Employing a distributed storage and computing framework, this module cleans, denoises, normalizes, and extracts features from multi-source heterogeneous data to generate standardized occupational health datasets. Multimodal risk assessment module: Based on machine learning algorithms, a dynamic risk assessment model is built, which outputs a comprehensive risk score by integrating essential risk indicators and management risk indicators; Dynamic early warning module: Based on risk scores, risk levels are divided into low, medium and high, and risk trend prediction and abnormal operating condition alarms are realized based on real-time data stream.
2. The occupational health risk assessment system based on big data analysis as described in claim 1, characterized in that, In the multimodal risk assessment module, the inherent risk indicator includes exposure frequency, and the managed risk indicator includes training pass rate; The dynamic risk assessment model adopts a weighted fusion strategy, in which the inherent risk has a weight of 30% and the management risk has a weight of 70%.
3. The occupational health risk assessment system based on big data analysis as described in claim 2, characterized in that, The machine learning algorithm used in the multimodal risk assessment module is an ensemble learning model, which integrates the outputs of two sub-models, Gradient Boosting Decision Tree (GBDT) and Multilayer Perceptron (MLP), and performs a weighted average to obtain a comprehensive risk score.
4. The occupational health risk assessment system based on big data analysis as described in claim 3, characterized in that, The multimodal risk assessment module adopts an integrated learning and neural network fusion architecture, and achieves dynamic risk assessment through the following steps: Feature organization: Standardized data is divided into two modal feature groups, including the essential risk feature group and the management risk feature group; numerical features are normalized and categorical features are encoded. Base model construction and training: The essential risk prediction sub-model (weight 30%) adopts the gradient boosting decision tree algorithm, specifically XGBoost, and iteratively constructs multiple CART decision trees to capture the nonlinear relationship between exposure factors and health consequences through residual fitting. Risk prediction sub-model (70% weight): It adopts a multilayer perceptron neural network, sets multiple hidden layers and applies the ReLU activation function to learn the deep mapping relationship between management indicators and risk suppression effect; Weighted fusion strategy: The normalized risk scores output by the two sub-models are linearly weighted and fused.
5. The occupational health risk assessment system based on big data analysis as described in claim 4, characterized in that, The multimodal risk assessment module further includes: Critical Control Point Identification Submodule: Based on real-time monitoring data from the Internet of Things, it uses spatial clustering and anomaly detection algorithms to locate high-exposure areas or operational processes and dynamically adjusts occupational hazard critical control points.
6. The occupational health risk assessment system based on big data analysis as described in claim 1, characterized in that, The dynamic early warning module pushes tiered early warning information to enterprises, regulatory authorities, and workers through a multi-level linkage mechanism.
7. The occupational health risk assessment system based on big data analysis as described in claim 1, characterized in that, It also includes a security management module, which is used to monitor data access behavior, mark abnormal modification operations with risk indicators and trigger permission audits.
8. The occupational health risk assessment system based on big data analysis as described in claim 1, characterized in that, It includes a user terminal and a base station terminal, which are used for data input, result visualization, and interactive query.
9. The occupational health risk assessment system based on big data analysis as described in claim 1, characterized in that, It also includes a non-readable storage medium containing a computer program that, when executed, implements the operational process of the occupational health risk assessment model, including data collection, risk calculation, level classification, early warning generation, and report output.
10. The occupational health risk assessment system based on big data analysis as described in claim 6, characterized in that, The model further includes a display module electrically connected to the dynamic early warning module, which is used to visually display the risk level, early warning information and risk trend.