Evaluation method of hazard level of occupational disease hazard factors under big data driving

Abstract

The invention discloses a big data driven occupational disease hazard factor hazard level evaluation method, and relates to the technical field of big data analysis, and the method comprises the steps: based on the occupational disease development stage big data of known post types, quantifying the occupational disease development stage demand vector of each post type; quantifying the difference exposure coefficient of the in-service personnel of the known post type; obtaining real-time hazard factor parameters of known post types, performing association mapping on the real-time hazard factor parameters and the occupational disease development stage demand vectors of the post types, and evaluating occupational disease development basic risk indexes of the known post types; according to the difference exposure coefficient of the in-service personnel of the known post type and the occupational disease development basic risk index of the known post type, establishing an occupational disease development accumulation model, and generating an occupational disease development hazard level grade of the known post type; the method has the beneficial effects that the subjectivity of traditional artificial experience is avoided, and objective and precise quantitative evaluation of hazard factors is realized.

Description

Technical Field

[0001] The present invention relates to the field of big data analysis technology, and in particular to a method for evaluating the hazard level of occupational hazard factors driven by big data. Background Art

[0002] Some occupational disease risk assessment methods utilize static models, failing to account for the impact of individual exposure factors (such as protective measures and working hours) or the cumulative effects of risk over time. They also rely on subjective weighting to integrate risk indicators, lacking objective data support. Threshold effect assessments distort assessments. Furthermore, they lack visualization and contribution quantification, making it difficult to accurately identify high-risk areas. Their hazard classification is overly simplistic, failing to capture the gradual evolution of risk. Consequently, these traditional methods struggle to meet the demand for precise and dynamic occupational disease risk quantification. Summary of the Invention

[0003] In order to solve the above technical problems, a method for evaluating the hazard level of occupational hazard factors driven by big data is provided. This technical solution solves the above problems.

[0004] In order to achieve the above objects, the technical solution adopted by the present invention is:

[0005] The evaluation method of occupational disease hazard level driven by big data includes:

[0006] Based on the big data of occupational disease development stages of known job types, we analyze the demand preferences for hazard factors in the occupational disease development stages of each job type and quantify the demand vectors of the occupational disease development stages of each job type;

[0007] Obtain the real-time protection status of employees in known job types during their work tasks to conduct work task-protection effectiveness matching analysis and quantify the differential exposure coefficients of employees in known job types.

[0008] Obtain real-time hazard factor parameters for known job types and correlate them with occupational disease development stage demand vectors for each job type to assess the basic risk indicators for occupational disease development for known job types;

[0009] Based on the differential exposure coefficients of employees in known job types and the basic risk indicators of occupational disease development in known job types, a cumulative model for occupational disease development is established to generate the occupational disease development hazard level grades for known job types.

[0010] Preferably, standardization is performed based on big data of occupational disease development stages of known job types;

[0011] Based on the big data of occupational disease development stages of known job types, the data is divided and labeled according to several occupational disease development stages of known job types to obtain characteristic data of occupational disease development stages of known job types;

[0012] The PCA principal component analysis method was used to reduce the dimension of the characteristic data of occupational disease development stages of known job types, and establish the characteristic vector matrix of occupational disease development stages of known job types.

[0013] Among them, a ij is the characteristic vector of the jth occupational disease development stage of the i-th job type, m is the total number of job types, and n is the total number of characteristic vectors of the occupational disease development stage;

[0014] Using the entropy weight method, based on the proportion of the number of occupational disease development stage eigenvectors of each known job type in the occupational disease development stage eigenvector matrix to the total number of occupational disease development stage eigenvectors of known job types, the occupational disease development stage weighted eigenvector matrix of known job types was established.

[0015] Among them, w ij is the characteristic vector weight of the j-th occupational disease development stage of the i-th job type;

[0016] Based on the occupational disease development stage eigenvector matrix of known job types and the weighted eigenvector matrix of occupational disease development stages of known job types, the weighted average risk index of each occupational disease development stage eigenvector of known job types is calculated.

[0017] Preferably, the weighted average risk index of the characteristic vector of each occupational disease development stage of a known job type is normalized;

[0018] Based on the K-means clustering algorithm, each occupational disease development stage type of a known job type is used as an initial cluster, and the Euclidean distance between the normalized value of the weighted average risk index of the characteristic vector of each occupational disease development stage of the known job type and the centroid of each initial cluster is calculated to divide the cluster, thereby generating a set of original feature samples of each occupational disease development stage type of the known job type; the development stage types include: low-risk occupational disease development stage type, medium-risk occupational disease development stage type, and high-risk occupational disease development stage type;

[0019] Based on the original feature sample set of each occupational disease development stage type of known job types, the SVR regression support vector machine is trained. The initial interval is preset according to each occupational disease development stage type. By minimizing the error function of the fitting data, the hyperplane boundary of each occupational disease development stage type of known job types is determined, and the occupational disease development stage demand vector of each job type is generated.

[0020] Preferably, a list of pending work tasks of known job types is obtained, and a sequence of pending work tasks of known job types is established;

[0021] Based on the big data of occupational disease development stages of known job types, the hazard factor parameters in the sequence of pending work tasks of known job types are marked, and a correlation matrix between pending work tasks and hazard factors of known job types is constructed;

[0022] Based on the association matrix of pending work tasks and hazard factors of known job types, the protection demand vector corresponding to the hazard factors of pending work tasks of known job types per unit time is analyzed according to the progress of the pending work tasks.

[0023] Preferably, based on edge vision sensors, real-time video data of employees of known job types performing work tasks is obtained;

[0024] According to the association matrix of pending work tasks and hazard factors of known job types, the corresponding protective equipment of pending work tasks and hazard factors are marked, and the association matrix of pending work tasks, hazard factors and protective equipment of known job types is established;

[0025] Based on each element in the association matrix of work tasks to be performed, hazards, and protective equipment for a given job type, a YOLOv7 object detection network is trained. Using real-time video data of employees performing work tasks for a given job type as input, the network generates a real-time wear compliance vector for employees performing work tasks for a given job type per unit time.

[0026] Using the cosine similarity formula, the protection matching degree between the real-time wear compliance vector of the employees of the known job type performing the work task per unit time and the protection demand vector corresponding to the hazard factors of the work task to be performed per unit time of the known job type is calculated. The time when the work task is performed under the non-positive protection matching degree is marked as the protection exposure time of the employees of the known job type performing the work task.

[0027] The exposure hazard factors of employees with known job types at the time of protective exposure during their work tasks were statistically analyzed, and a linear regression function was established with the exposure frequency and exposure duration to calculate the differential exposure coefficients of employees with known job types.

[0028] Preferably, a standardization process is performed based on a correlation matrix of work tasks to be performed and hazard factors of known job types to obtain a standardized correlation matrix of work tasks to be performed and hazard factors of known job types;

[0029] Using the PAC principal component analysis method, the dimension reduction process of the standardized correlation matrix of pending work tasks and hazard factors of known job types is carried out to obtain the standardized correlation vector matrix of pending work tasks and hazard factors of known job types.

[0030] Based on the cosine similarity, the occupational disease matching degree between the standardized association vector matrix of work tasks to be performed-hazard factors of known job types and the occupational disease development stage demand vectors of each job type is marked according to unit time, and a time series matrix of occupational disease development stages associated with hazard factors of known job types is established.

[0031] Preferably, the LSTM long short-term memory network is trained based on the time series matrix of occupational disease development stages associated with hazard factors of known job types, with the hazard factors per unit time as input and the basic risk indicators of occupational disease development of known job types as output.

[0032] Preferably, the initial cumulative risk value of occupational disease development of a known job type is calculated between the differential exposure coefficient of employees of a known job type and the basic risk index of occupational disease development of a known job type according to a weighted fusion formula;

[0033] Verify the changing trend of the initial cumulative risk value of occupational disease development of known job types under the increasing differential exposure coefficient of employees of known job types and the basic risk index of occupational disease development of known job types according to unit time, and establish the correlation time series between the differential exposure coefficient of employees of known job types and the basic risk index of occupational disease development;

[0034] Preferably, a thermal distribution diagram of differential exposure coefficients and basic risk indicators of occupational disease development is established based on a time series correlation between differential exposure coefficients and basic risk indicators of occupational disease development of employees of known job types;

[0035] Based on the contribution of cooperative game theory, the contribution value of the initial cumulative risk of occupational disease development of the differential exposure coefficient change in the differential exposure coefficient-basic risk distribution diagram of occupational disease development of known job types is traversed, and the differential exposure coefficient weight of the employees of known job types is assigned;

[0036] Based on the differential exposure coefficients of employees in known job types, the weights of the differential exposure coefficients of employees in known job types and the basic risk indicators of occupational disease development in known job types, the sliding window integration method is used to calculate the cumulative risk values ​​of occupational disease development in known job types and determine the hazard level of occupational disease development in known job types.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] The present invention proposes an evaluation scheme for the hazard level of occupational disease hazard factors driven by big data. Driven by big data, it automatically analyzes the occupational disease development stage demand vectors, real-time protection effectiveness matching and dynamic correlation of hazard factors of job types, avoiding the subjectivity of traditional manual experience and realizing the objective and precise quantitative evaluation of hazard factors; through dynamic modeling of differential exposure coefficients and basic risk indicators, it accurately identifies the cumulative risk level of occupational disease development in different positions, improves the real-time nature of hazard level evaluation, and provides data support for targeted protection decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 This is a flow chart of the evaluation method for the hazard level of occupational hazard factors driven by big data; DETAILED DESCRIPTION

[0040] The following description is intended to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are merely examples, and those skilled in the art may conceive of other obvious variations.

[0041] Reference Figure 1 As shown in the figure, the evaluation method of occupational hazard level driven by big data includes:

[0042] Step 1: Based on the big data of occupational disease development stages of known job types, analyze the demand preferences for occupational disease hazard factors in the development stages of occupational diseases of each job type, and quantify the demand vectors of occupational disease development stages of each job type;

[0043] The step 1 includes the following:

[0044] Step 101: Standardize the big data of occupational disease development stages based on known job types;

[0045] Based on the big data of occupational disease development stages of known job types, the data is divided and labeled according to several occupational disease development stages of known job types to obtain characteristic data of occupational disease development stages of known job types;

[0046] The PCA principal component analysis method is used to perform dimensionality reduction processing on the characteristic data of occupational disease development stages of known job types, and the characteristic vector matrix A of occupational disease development stages of known job types is established;

[0047]

[0048] Among them, a ij is the characteristic vector of the jth occupational disease development stage of the i-th job type, m is the total number of job types, and n is the total number of characteristic vectors of the occupational disease development stage;

[0049] Using the entropy weight method, based on the proportion of the number of occupational disease development stage eigenvectors of each known job type in the occupational disease development stage eigenvector matrix to the total number of occupational disease development stage eigenvectors of known job types, the occupational disease development stage weighted eigenvector matrix B of known job types was established.

[0050]

[0051] Among them, w ij is the characteristic vector weight of the j-th occupational disease development stage of the i-th job type;

[0052] Based on the occupational disease development stage eigenvector matrix of known job types and the weighted eigenvector matrix of occupational disease development stages of known job types, the weighted average risk index of each occupational disease development stage eigenvector of known job types is calculated as follows:

[0053]

[0054] Among them, P ij is the weighted average risk index of the characteristic vector of the j-th occupational disease development stage of the i-th job type;

[0055] Step 102: normalize the weighted average risk index of each occupational disease development stage characteristic vector of a known job type;

[0056] Based on the K-means clustering algorithm, each occupational disease development stage type of a known job type is used as an initial cluster, and the Euclidean distance between the normalized value of the weighted average risk index of the characteristic vector of each occupational disease development stage of the known job type and the centroid of each initial cluster is calculated to divide the cluster, thereby generating a set of original feature samples of each occupational disease development stage type of the known job type; the development stage types include: low-risk occupational disease development stage type, medium-risk occupational disease development stage type, and high-risk occupational disease development stage type;

[0057] Based on the original feature sample set of each occupational disease development stage type of known job types, the SVR regression support vector machine is trained. The initial interval is preset according to each occupational disease development stage type. By minimizing the error function of the fitting data, the hyperplane boundary of each occupational disease development stage type of known job types is determined, and the occupational disease development stage demand vector of each job type is generated as follows:

[0058]

[0059] Among them, V ij is the demand vector of the jth occupational disease development stage of the i-th job type, w j is the weight vector of the jth occupational disease development stage, fj (x) is the decision function value of the jth occupational disease development stage, j∈SV j The support vector set SV representing the j-th occupational disease development stage j , α i 、 are the Lagrange multiplier and its dual multiplier of the support vector, κ(x i ,x) is the measure of support vector x i The similarity between the input feature x, b j is the bias term for the jth occupational disease development stage, φ(x i ) is the support vector x i The feature map, m i is the feature mask of the i-th job type, and γ is the normalized scale factor.

[0060] When using, combine the contents in steps 101 to 102.

[0061] Furthermore, existing occupational disease development stage assessments for job types are often based on fixed thresholds or historical data, failing to reflect real-time trends in occupational hazards. Furthermore, the categorization of occupational disease development stages (e.g., low-risk, medium-risk, and high-risk) relies on expert experience and lacks data-driven dynamic classification. Statistical methods (e.g., mean and standard deviation) for calculating hazard factor weights struggle to capture complex nonlinear relationships and fail to account for differences in sensitivity to hazard factors across job types, resulting in insufficient generalization of assessment results. Consequently, traditional methods rely on expert scoring or empirical judgment, making them susceptible to human influence.

[0062] This scheme extracts the characteristics of key hazard factors through principal component analysis (PCA) to reduce noise interference and improve computational efficiency. It also uses the entropy weight method to objectively calculate the weight of each feature to avoid subjective bias and enhance the scientific nature of the assessment. It combines K-means clustering to automatically divide the developmental stages of low-, medium-, and high-risk occupational diseases, reducing manual classification errors. It fits the optimal hyperplane boundary based on SVR regression to improve the recognition accuracy of hazard factor demand preferences for different job types, update feature vectors in real time, realize dynamic monitoring and early warning of occupational hazards, and improve long-term prediction accuracy.

[0063] Step 2: Obtain the real-time protection status of employees in known job types during their work tasks, conduct work task-protection effectiveness matching analysis, and quantify the differential exposure coefficients of employees in known job types;

[0064] The second step includes the following:

[0065] Step 201: Obtain a list of pending work tasks of known job types and establish a sequence of pending work tasks of known job types;

[0066] Based on the big data of occupational disease development stages of known job types, the hazard factor parameters in the sequence of pending work tasks of known job types are marked, and a correlation matrix between pending work tasks and hazard factors of known job types is constructed;

[0067] Based on the association matrix of pending tasks and hazard factors of known job types, the protection demand vector corresponding to the hazard factors of pending tasks of known job types per unit time is analyzed according to the progress of the pending tasks. The method is as follows:

[0068]

[0069] Among them, D ik (t) is the protection demand vector corresponding to the kth hazard factor of the work task of the i-th known job type under the t-th unit time, R ik is the correlation strength of the kth hazard factor of the task to be performed of the i-th known position type in the task to be performed-hazard factor correlation matrix of the known position type, C ik (t) is the kth hazard factor value of the work task of the i-th known job type under the t-th unit time;

[0070] Step 202: Acquire real-time video data of employees of known job types performing work tasks based on edge vision sensors;

[0071] According to the association matrix of pending work tasks and hazard factors of known job types, the corresponding protective equipment of pending work tasks and hazard factors are marked, and the association matrix of pending work tasks, hazard factors and protective equipment of known job types is established;

[0072] Based on each element in the association matrix of work tasks to be performed, hazards, and protective equipment for a given job type, a YOLOv7 object detection network is trained. Using real-time video data of employees performing work tasks for a given job type as input, the network generates a real-time wear compliance vector for employees performing work tasks for a given job type per unit time.

[0073] Using the cosine similarity formula, the protection matching degree between the real-time wear compliance vector of the employees of the known job type performing the work task per unit time and the protection demand vector corresponding to the hazard factors of the work task to be performed per unit time of the known job type is calculated. The time when the work task is performed under the non-positive protection matching degree is marked as the protection exposure time of the employees of the known job type performing the work task.

[0074] The exposure hazard factors of employees with known job types at the time of protective exposure during their work tasks were statistically analyzed, and a linear regression function was established between the exposure frequency and exposure duration to calculate the differential exposure coefficients of employees with known job types.

[0075] When used, combine the contents in steps 201 to 202.

[0076] Furthermore, traditional methods rely primarily on regular manual inspections or spot checks, which are unable to monitor the wearing of protective equipment in real time, resulting in delayed detection of exposure risks. Furthermore, most protection standards are static and fail to incorporate the dynamic changes in hazard factors associated with specific work tasks (such as fluctuations in noise and dust concentrations in different processes). This makes it difficult to cover all work scenarios and can easily lead to missed detections of high-risk moments.

[0077] This solution dynamically links work tasks, hazards, and protective equipment to achieve precise protection demand mapping. Based on unit time analysis, it captures instantaneous protection gaps (such as brief exposure to high noise without earplugs). It uses real-time video streams to detect the wearing of protective equipment (masks, goggles, etc.) and generates a compliance vector, replacing manual inspections. By comparing the protection demand vector with the actual wearing compliance vector, it identifies the moment of protection exposure (such as the critical point of not wearing protective equipment). By integrating exposure frequency, duration, and hazard intensity, it quantifies the differential exposure coefficient through a regression function to more accurately reflect individual risk status.

[0078] Step 3: Obtain real-time hazard factor parameters of known job types and perform correlation mapping with occupational disease development stage demand vectors of each job type to evaluate the basic risk indicators of occupational disease development of the known job types;

[0079] The step three includes the following:

[0080] Step 301: Perform standardization based on the association matrix of pending work tasks and hazard factors of known job types to obtain a standardized association matrix of pending work tasks and hazard factors of known job types;

[0081] Using the PAC principal component analysis method, the dimension reduction process of the standardized correlation matrix of pending work tasks and hazard factors of known job types is carried out to obtain the standardized correlation vector matrix of pending work tasks and hazard factors of known job types.

[0082] Based on the cosine similarity, the occupational disease matching degree between the standardized correlation vector matrix of work tasks to be performed and hazard factors of known job types and the occupational disease development stage demand vector of each job type is marked according to unit time, and the time series matrix of occupational disease development stage associated with hazard factors of known job types is established;

[0083] Step 302: Based on the time series matrix of occupational disease development stages associated with hazard factors of known job types, an LSTM long short-term memory network is trained, with the hazard factors per unit time as input and the basic risk indicators of occupational disease development of known job types as output;

[0084] When used, combine the contents in steps 301 to 302.

[0085] As a further content, in the existing methods, the correlation analysis between the hazard factors of job types and the development stage of occupational diseases is relatively static, lacking dynamic time series modeling, and it is difficult to reflect the evolution of occupational disease risks under long-term exposure; and relying on static parameters (such as hazard factor concentration) does not comprehensively consider the dynamic matching relationship between the temporal characteristics of work tasks and the development stage of occupational diseases, resulting in insufficient accuracy of occupational disease development risk indicators of job types.

[0086] This solution constructs a time series matrix linking hazard factors to occupational disease developmental stages, dynamically linking static hazard factor parameters with occupational disease developmental requirements. This more accurately reflects the cumulative effects of long-term exposure. Combined with an LSTM network to capture temporal dependencies, it enhances the ability to predict the evolution of occupational disease risks. Furthermore, a standardized work task-hazard factor correlation matrix is ​​introduced, quantifying the matching degree between hazard factors and occupational disease developmental requirements using cosine similarity. This enables a fusion analysis of qualitative risks (such as developmental stages) and quantitative parameters (such as hazard factor values), providing quantitative parameters for risk changes per unit time.

[0087] Step 4: Based on the differential exposure coefficients of employees of known job types and the basic risk indicators of occupational disease development of known job types, a cumulative model for occupational disease development is established to generate occupational disease development hazard level grades for known job types;

[0088] The step 4 includes the following contents:

[0089] Step 401: Calculate the initial cumulative risk value of occupational disease development of a given job type based on the difference exposure coefficient of employees of the given job type and the basic risk index of occupational disease development of the given job type according to the weighted fusion formula, in the following manner:

[0090] C ij =β·E i +(1-β)·R ij

[0091] Among them, C ij is the initial cumulative risk value of occupational disease development of the jth job type, E i is the differential exposure coefficient of employees in the ith job type, R ij The jth occupational disease development basic risk index of the i-th job type, β is the weight coefficient of the heterogeneous exposure coefficient;

[0092] Verify the changing trend of the initial cumulative risk value of occupational disease development of known job types under the increasing differential exposure coefficient of employees of known job types and the basic risk index of occupational disease development of known job types according to unit time, and establish the correlation time series between the differential exposure coefficient of employees of known job types and the basic risk index of occupational disease development;

[0093] Step 402: establishing a thermal distribution diagram of differential exposure coefficients and basic risk indicators of occupational disease development based on the time series of differential exposure coefficients and basic risk indicators of occupational disease development of employees of known job types;

[0094] Based on the contribution of cooperative game theory, the contribution value of the initial cumulative risk of occupational disease development of the differential exposure coefficient change in the differential exposure coefficient-basic risk distribution diagram of occupational disease development of known job types is traversed, and the differential exposure coefficient weight of the employees of known job types is assigned;

[0095] Based on the differential exposure coefficients of employees of known job types, the differential exposure coefficient weights of employees of known job types, and the basic risk indicators of occupational disease development of known job types, the sliding window integration method is used to calculate the cumulative risk value of occupational disease development of known job types and determine the occupational disease development hazard level of known job types in the following manner:

[0096]

[0097] Among them, CumRisk(C ij ) is the occupational disease development hazard level of the known job type corresponding to the cumulative risk value of occupational disease development of the known job type, W β is the differential exposure coefficient weight of employees with known job types, is the differential exposure coefficient of employees in the ith position type in the tth unit time, is the basic risk indicator of occupational disease development of the jth job type in the ith unit of time, and T is the total length of the sliding window.

[0098] When used, combine the contents in steps 401 to 402.

[0099] As a further note, some occupational disease risk assessment methods utilize static models, failing to consider the impact of individual exposure factors (such as protective measures and working hours) or the cumulative effect of risk over time. They also rely on subjective weighting to integrate risk indicators, lacking objective data support. Threshold effect assessment methods lead to distorted assessments. Furthermore, they lack visualization analysis and contribution quantification, making it difficult to accurately identify high-risk areas. The hazard level classification is also overly simplistic, failing to capture the gradual evolution of risk. Consequently, these traditional methods struggle to meet the demand for precise and dynamic occupational disease risk quantification.

[0100] This solution realizes the dynamic cumulative calculation of risk value through the association time series of differential exposure coefficient and basic risk index of occupational disease development, combined with sliding window integration method. Combined with game theory contribution analysis, the weight of differential exposure coefficient is automatically calculated from the thermal distribution diagram to avoid subjective bias, so as to determine the critical exposure threshold and accelerated accumulation interval, and calculate the cumulative risk value using sliding window integration method. The hazard level is dynamically adjusted in combination with the time dimension to avoid misjudgment caused by fixed threshold.

[0101] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions merely illustrate the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. The evaluation method of occupational disease hazard level driven by big data is characterized by: include: S1. Based on the big data of occupational disease development stages of known job types, analyze the demand preferences for occupational hazard factors in the development stages of occupational diseases of each job type, and quantify the demand vectors of occupational disease development stages of each job type; S2. Obtain the real-time protection status of employees in known job types during their work tasks, conduct work task-protection effectiveness matching analysis, and quantify the differential exposure coefficients of employees in known job types; S3. Obtain real-time hazard factor parameters for known job types and correlate them with occupational disease development stage demand vectors for each job type to evaluate the basic risk indicators for occupational disease development for the known job types; S4. Based on the differential exposure coefficients of employees in known job types and the basic risk indicators of occupational disease development in known job types, a cumulative model for occupational disease development is established to generate occupational disease development hazard level grades for known job types.

2. The method for evaluating the hazard level of occupational hazard factors driven by big data according to claim 1 is characterized in that: Said S1 comprises: Standardized processing of big data on occupational disease development stages based on known job types; Based on the big data of occupational disease development stages of known job types, the data is divided and labeled according to several occupational disease development stages of known job types to obtain characteristic data of occupational disease development stages of known job types; The PCA principal component analysis method was used to reduce the dimension of the characteristic data of occupational disease development stages of known job types, and establish the characteristic vector matrix of occupational disease development stages of known job types. Among them, a ij is the characteristic vector of the jth occupational disease development stage of the i-th job type, m is the total number of job types, and n is the total number of characteristic vectors of the occupational disease development stage; Using the entropy weight method, based on the proportion of the number of occupational disease development stage eigenvectors of each known job type in the occupational disease development stage eigenvector matrix to the total number of occupational disease development stage eigenvectors of known job types, the occupational disease development stage weighted eigenvector matrix of known job types was established. Among them, w ij is the characteristic vector weight of the j-th occupational disease development stage of the i-th job type; Based on the occupational disease development stage eigenvector matrix of known job types and the weighted eigenvector matrix of occupational disease development stages of known job types, the weighted average risk index of each occupational disease development stage eigenvector of known job types is calculated.

3. The method for evaluating the hazard level of occupational hazard factors driven by big data according to claim 2 is characterized in that: Said S1 further comprises: Normalize the weighted average risk index of the characteristic vector of each occupational disease development stage of known job types; Based on the K-means clustering algorithm, each occupational disease development stage type of a known job type is used as an initial cluster, and the Euclidean distance between the normalized value of the weighted average risk index of the characteristic vector of each occupational disease development stage of the known job type and the centroid of each initial cluster is calculated to divide the cluster, thereby generating a set of original feature samples of each occupational disease development stage type of the known job type; the development stage types include: low-risk occupational disease development stage type, medium-risk occupational disease development stage type, and high-risk occupational disease development stage type; Based on the original feature sample set of each occupational disease development stage type of known job types, the SVR regression support vector machine is trained. The initial interval is preset according to each occupational disease development stage type. By minimizing the error function of the fitting data, the hyperplane boundary of each occupational disease development stage type of known job types is determined, and the occupational disease development stage demand vector of each job type is generated.

4. The method for evaluating the hazard level of occupational hazard factors driven by big data according to claim 1 is characterized in that: The S2 includes: Obtain a list of pending work tasks of known job types and establish a sequence of pending work tasks of known job types; Based on the big data of occupational disease development stages of known job types, the hazard factor parameters in the sequence of pending work tasks of known job types are marked, and a correlation matrix between pending work tasks and hazard factors of known job types is constructed; Based on the association matrix of pending work tasks and hazard factors of known job types, the protection demand vector corresponding to the hazard factors of pending work tasks of known job types per unit time is analyzed according to the progress of the pending work tasks.

5. The method for evaluating the hazard level of occupational hazard factors driven by big data according to claim 4 is characterized in that: Said S2 further comprises: Based on edge vision sensors, real-time video data of employees in known positions performing work tasks is obtained; According to the association matrix of pending work tasks and hazard factors of known job types, the corresponding protective equipment of pending work tasks and hazard factors are marked, and the association matrix of pending work tasks, hazard factors and protective equipment of known job types is established; Based on each element in the association matrix of work tasks to be performed, hazards, and protective equipment for a given job type, a YOLOv7 object detection network is trained. Using real-time video data of employees performing work tasks for a given job type as input, the network generates a real-time wear compliance vector for employees performing work tasks for a given job type per unit time. Using the cosine similarity formula, the protection matching degree between the real-time wear compliance vector of the employees of the known job type performing the work task per unit time and the protection demand vector corresponding to the hazard factors of the work task to be performed per unit time of the known job type is calculated. The time when the work task is performed under the non-positive protection matching degree is marked as the protection exposure time of the employees of the known job type performing the work task. The exposure hazard factors of employees with known job types at the time of protective exposure during their work tasks were statistically analyzed, and a linear regression function was established with the exposure frequency and exposure duration to calculate the differential exposure coefficients of employees with known job types.

6. The method for evaluating the hazard level of occupational hazard factors driven by big data according to claim 1 is characterized in that: The S3 includes: Based on the association matrix of pending work tasks and hazard factors of known job types, standardization processing is performed to obtain a standardized association matrix of pending work tasks and hazard factors of known job types; Using the PAC principal component analysis method, the dimension reduction process of the standardized correlation matrix of pending work tasks and hazard factors of known job types is carried out to obtain the standardized correlation vector matrix of pending work tasks and hazard factors of known job types. Based on the cosine similarity, the occupational disease matching degree between the standardized association vector matrix of work tasks to be performed-hazard factors of known job types and the occupational disease development stage demand vectors of each job type is marked according to unit time, and a time series matrix of occupational disease development stages associated with hazard factors of known job types is established.

7. The method for evaluating the hazard level of occupational hazard factors driven by big data according to claim 6, characterized in that: Said S3 further comprises: According to the time series matrix of occupational disease development stages associated with hazard factors of known job types, an LSTM long short-term memory network is trained, with the hazard factors per unit time as input and the basic risk indicators of occupational disease development of known job types as output.

8. The method for evaluating the hazard level of occupational hazard factors driven by big data according to claim 7 is characterized in that: The S4 includes: The initial cumulative risk value of occupational disease development of a known job type is calculated based on the difference exposure coefficient of employees in a known job type and the basic risk index of occupational disease development of a known job type according to the weighted fusion formula; By verifying the changing trend of the initial cumulative risk value of occupational disease development of known job types under the increasing differential exposure coefficient of employees in known job types and the basic risk index of occupational disease development of known job types per unit time, a time series of the correlation between the differential exposure coefficient of employees in known job types and the basic risk index of occupational disease development was established.

9. The method for evaluating the hazard level of occupational hazard factors driven by big data according to claim 8, characterized in that: Said S3 further comprises: Based on the time series of differential exposure coefficients and basic risk indicators of occupational disease development of employees with known job types, a thermal distribution diagram of differential exposure coefficients and basic risk indicators of occupational disease development was established; Based on the contribution of cooperative game theory, the contribution value of the initial cumulative risk of occupational disease development of the differential exposure coefficient change in the differential exposure coefficient-basic risk distribution diagram of occupational disease development of known job types is traversed, and the differential exposure coefficient weight of the employees of known job types is assigned; Based on the differential exposure coefficients of employees in known job types, the weights of the differential exposure coefficients of employees in known job types and the basic risk indicators of occupational disease development in known job types, the sliding window integration method is used to calculate the cumulative risk values ​​of occupational disease development in known job types and determine the hazard level of occupational disease development in known job types.

Images