Method and device for training classification model of key areas of hazardous chemical substances

By obtaining and pretreating characteristic indicators of hazardous chemical areas, training classification models, and using Wasserstein center of gravity and model fusion technology, the problem of non-standardization of key areas of hazardous chemicals was solved, and accurate prediction of key areas and levels was achieved, and improvement of the improvement of the The accuracy and reliability of classification are achieved.

CN118940098APending Publication Date: 2024-11-12CHINA PETROLEUM & CHEMICAL CORP +1
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202310532858.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-05-11
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The classification of key areas for hazardous chemicals in the prior art is not standardized and rationalized enough, and there is a lack of quantitative evaluation methods, resulting in unclear division standards.

Method used

By obtaining the characteristic indicators of multiple hazardous chemical areas, forming an index data set, marking and preprocessing, training classification models, using Wasserstein center of gravity and model fusion technology, establishing a classification model for key areas of hazardous chemicals, and realizing the key areas and levels of key areas and levels accurate prediction.

Benefits of technology

The quantification, standardization and rationalization classification of hazardous chemicals regions have been achieved, the accuracy and reliability of the division of key areas have been improved, and the stability and generalization capabilities of the model have been ensured.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118940098A_ABST
    Figure CN118940098A_ABST
Patent Text Reader

Abstract

The embodiment of the invention provides a training method and device of a hazardous chemical substance key area classification model and a hazardous chemical substance key area classification method and device, and belongs to the field of industrial site safety level classification. The training method comprises the following steps: acquiring a plurality of different types of feature indexes of a plurality of hazardous chemical substance areas to form an index data set; marking whether the index data set is a key area of the hazardous chemical substance or not and marking the grade of the key area to obtain a training set; and inputting the training set into a to-be-trained classification model, and performing model training to obtain a hazardous chemical substance key area classification model. Therefore, the hazardous chemical substance key area classification model obtained after training can accurately predict and classify whether the to-be-predicted area is the key area and the key area grade of the to-be-predicted area according to the feature index of the to-be-predicted area, and quantification, standardization and rationalization of hazardous chemical substance area classification are achieved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of industrial site safety level classification, and in particular to a training method and device for a hazardous chemicals key area classification model, and a hazardous chemicals key area classification method and device. Background Art

[0002] Key areas for hazardous chemicals refer to areas that have a significant impact on the environment and society. They are usually divided based on whether or not they are hazardous chemical production and operation enterprises, the number of enterprises operating them, and the layout of chemical parks. They are divided into national, provincial, and municipal levels according to their importance.

[0003] At present, the consideration dimensions and data for the classification of key areas for hazardous chemicals and the important classification of key areas are limited and the standards are unclear, and there is no corresponding quantitative evaluation method. Therefore, a data analysis method is needed to quantitatively evaluate and classify the division of key areas for hazardous chemicals, so as to achieve the standardization and rationalization of the division of key areas for hazardous chemicals. Summary of the invention

[0004] The purpose of the embodiments of the present invention is to overcome the problem that the division of key areas of hazardous chemicals in the prior art is not standardized and rational enough, and to provide a training method and device for a classification model of key areas of hazardous chemicals, as well as a method and device for classifying key areas of hazardous chemicals.

[0005] The first aspect of the present application provides a method for training a classification model for key areas of hazardous chemicals, the method comprising:

[0006] Obtain multiple different types of characteristic indicators of multiple hazardous chemical areas to form an indicator data set;

[0007] The indicator data set is labeled to determine whether it is a key area for hazardous chemicals, and the key area level is labeled to obtain a training set;

[0008] The training set is input into the classification model to be trained and the model is trained to obtain a classification model for key areas of hazardous chemicals.

[0009] In one embodiment of the present application, before labeling the indicator data set as to whether it is a key area of ​​hazardous chemicals and labeling the key area level to obtain a training set, the method further includes:

[0010] The characteristic indicators whose values ​​in the screening indicator data set exceed the preset range are regarded as abnormal values;

[0011] Determine the characteristic indicator type to which the outlier belongs;

[0012] Determine the mean and standard deviation of all feature indicators included in the feature indicator type;

[0013] Determine the expected guarantee range based on the mean and standard deviation;

[0014] Replace outliers with random numbers in the desired guaranteed range.

[0015] In one embodiment of the present application, the classification model to be trained includes a first model to be trained for classifying whether it is a key area, and a second model to be trained for classifying the level of the key area, the second model to be trained includes multiple sub-classification models of different types, and the training set is input into the classification model to be trained, and the model training is performed to obtain a classification model for key areas of hazardous chemicals, including:

[0016] The training set is input into a first model to be trained, and the first model to be trained is trained using the Wasserstein center of gravity to obtain a first model;

[0017] Input the training set into each sub-classification model for model training respectively, and then perform model fusion to obtain a second model;

[0018] The first model and the second model are combined into a classification model for key areas of hazardous chemicals.

[0019] In one embodiment of the present application, a training set is input into a first model to be trained, and the first model to be trained is trained using the Wasserstein center of gravity to obtain a first model, including:

[0020] During the training process of each layer of the first model to be trained, a local gradient is determined according to the samples provided by the training set and the loss function of the first model to be trained, and the local gradient is scaled and normalized;

[0021] Determine the Wasserstein center of gravity of all scaled and normalized local gradients and use the Wasserstein center of gravity as the global gradient;

[0022] Normalize the global gradient;

[0023] The model parameters of the first model to be trained are updated using the normalized global gradient.

[0024] In one embodiment of the present application, the first model is a lightweight logistic regression model.

[0025] In one embodiment of the present application, the training set is input into each sub-classification model for model training respectively, and then the model is fused to obtain a second model, including:

[0026] The five-fold cross-validation method is used to train each sub-classification model to obtain five prediction results for each sub-classification model;

[0027] The five prediction results of each sub-classification model are stacked vertically to form a prediction result matrix of each sub-classification model;

[0028] All prediction result matrices are horizontally spliced ​​to form a training set matrix;

[0029] Determine the average of the five prediction results of each sub-classification model to obtain a matrix of the average prediction results of each sub-classification model;

[0030] All the prediction result average matrices are horizontally transformed into a test set matrix;

[0031] The training set matrix is ​​used to train the fusion model, and the test set matrix is ​​used to test the trained fusion model to be trained;

[0032] When the test results meet the preset requirements, all trained sub-classification models and the trained fusion model are merged into a second model according to the hierarchical relationship that the output of the sub-classification model is the input of the fusion model.

[0033] In one embodiment of the present application, the sub-classification models include an XGBoost model, a LightGBM model, and a CATBoost model, and the fusion model is a Bayesian regression model.

[0034] In one embodiment of the present application, the method further includes:

[0035] According to the indicator data set, a heat map representing the correlation between multiple indicator feature types included in the indicator data set and whether it is a key area for hazardous chemicals is determined;

[0036] Determine the correlation coefficient between all characteristic indicator types and whether it is a key area for hazardous chemicals based on the heat map;

[0037] Multiple important indicator feature types are determined based on all correlation coefficients.

[0038] The second aspect of the present application provides a classification method for key areas of hazardous chemicals, the method comprising:

[0039] Obtain characteristic indicators of the area to be predicted and form a data set to be predicted;

[0040] The data set to be predicted is input into the hazardous chemicals key area classification model, and the classification result of the area to be predicted output by the hazardous chemicals key area classification model is obtained, wherein the classification result is whether the area to be predicted is a key area, or the key area level of the area to be predicted, wherein the hazardous chemicals key area classification model is obtained according to the training method of the hazardous chemicals key area classification model provided in the first aspect of the present application.

[0041] In one embodiment of the present application, before inputting the data set to be predicted into the hazardous chemicals key area classification model and obtaining the classification result of the area to be predicted output by the hazardous chemicals key area classification model, the method further includes:

[0042] Filter the characteristic indicators whose values ​​in the data set to be predicted exceed the preset range as outliers;

[0043] Determine the characteristic indicator type to which the outlier belongs;

[0044] Determine the mean and standard deviation of all feature indicators included in the feature indicator type;

[0045] Determine the expected guarantee range based on the mean and standard deviation;

[0046] Replace outliers with random numbers in the desired guaranteed range.

[0047] In one embodiment of the present application, the hazardous chemicals key area classification model includes a first model for classifying whether it is a key area, and a second model for classifying the level of the key area. The preprocessed prediction data set is input into the hazardous chemicals key area classification model to obtain the classification result of the area to be predicted output by the hazardous chemicals key area classification model, including:

[0048] Inputting the prediction data set into the first model to determine whether the area to be predicted is a key area;

[0049] When the area to be predicted is not a key area, the area to be predicted is not a key area as a classification result output;

[0050] In the case where the area to be predicted is a key area, the prediction data set is input into the second model;

[0051] The key area level of the area to be predicted of the second model data is output as a classification result.

[0052] The third aspect of the present application provides a training device for a hazardous chemicals key area classification model, the device comprising:

[0053] An indicator data set building module is used to obtain multiple different types of characteristic indicators of multiple hazardous chemicals areas to form an indicator data set;

[0054] The labeling module is used to label the preprocessed indicator data set to determine whether it is a key area for hazardous chemicals, and to label the level of the key area, so as to obtain a training set;

[0055] The model training module is used to input the training set into the classification model to be trained, perform model training, and obtain a classification model for key areas of hazardous chemicals.

[0056] A fourth aspect of the present application provides a classification device for key areas of hazardous chemicals, the device comprising:

[0057] The module for forming the data set to be predicted is used to obtain characteristic indicators of the area to be predicted and form the data set to be predicted;

[0058] A classification result acquisition module is used to input a prediction data set into a hazardous chemicals key area classification model, and obtain a classification result of the area to be predicted output by the hazardous chemicals key area classification model, wherein the classification result is whether the area to be predicted is a key area, or the key area level of the area to be predicted, wherein the hazardous chemicals key area classification model is obtained according to the training method of the hazardous chemicals key area classification model provided in the first aspect of the present application.

[0059] The fifth aspect of the present application provides an electronic device, including a processor and a memory, the memory storing machine executable instructions that can be executed by the processor, and the processor executing the machine executable instructions to implement the training method of the hazardous chemicals key area classification model provided in the first aspect of the present application, or the hazardous chemicals key area classification method provided in the second aspect of the present application.

[0060] The sixth aspect of the present application provides a machine-readable storage medium, which stores instructions. When the instructions are executed by a processor, the processor implements the training method of the hazardous chemicals key area classification model provided in the first aspect of the present application, or the classification method of hazardous chemicals key areas provided in the second aspect of the present application.

[0061] Through the above technical scheme, a training set for training a classification model for key hazardous chemical areas can be established based on the quantified characteristic indicators of multiple hazardous chemicals areas. The classification model for key hazardous chemical areas obtained after training can accurately predict and classify whether the area to be predicted is a key area and the key area level of the area to be predicted based on the characteristic indicators of the area to be predicted, thereby realizing the quantification, standardization and rationalization of hazardous chemical area classification.

[0062] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings:

[0064] Figure 1 A flowchart of a method for training a classification model for key areas of hazardous chemicals according to an embodiment of the present application is schematically shown;

[0065] Figure 2A flowchart schematically illustrates a method for classifying key areas of hazardous chemicals according to an embodiment of the present application;

[0066] Figure 3 A workflow logic block diagram of a hazardous chemicals key area classification model according to an embodiment of the present application is schematically shown;

[0067] Figure 4 A structural block diagram of a training device for a hazardous chemicals key area classification model according to an embodiment of the present application is schematically shown;

[0068] Figure 5 The structural block diagram of a classification device for key areas of hazardous chemicals according to an embodiment of the present application is schematically shown. DETAILED DESCRIPTION

[0069] The specific implementation of the present application is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the present application, and is not used to limit the present application.

[0070] It should be noted that if the embodiments of the present application involve directional indications (such as up, down, left, right, front, back, etc.), the directional indications are only used to explain the relative position relationship, reversal status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0071] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of the present application, the descriptions of "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or suggesting their relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by this application.

[0072] Figure 1 The flowchart of a method for training a classification model for key areas of hazardous chemicals according to an embodiment of the present application is schematically shown. Figure 1 As shown, in one embodiment of the present application, a training method for a classification model for key areas of hazardous chemicals is provided, comprising steps S100-S300:

[0073] Step S100: Acquire multiple different types of characteristic indicators of multiple hazardous chemical areas to form an indicator data set;

[0074] Hazardous chemical areas refer to areas related to the production, operation, and storage of hazardous chemicals. Hazardous chemical key areas refer to areas that have a significant impact on the environment and society. They are usually divided based on whether they are hazardous chemical production and operation enterprises, the number of enterprises, and the layout of chemical parks. They are divided into national, provincial, and municipal levels according to their importance. Therefore, the training data can be derived from the National Hazardous Chemicals Safety Production Risk Monitoring and Early Warning System Database. The characteristic indicators of multiple hazardous chemical areas across the country can be queried through multi-table queries. The characteristic indicators are quantitative values ​​that characterize various types of information in hazardous chemical areas.

[0075] Characteristic indicators include many different types. For example, characteristic indicator types may include: characteristic indicator types that characterize basic information of hazardous chemicals areas, such as the number of hazardous chemical production, operation, and use enterprises, the number of chemical parks, the enterprise entry rate, the proportion of A / B-level chemical parks, the number of storage tanks, the number of monitoring units, the number of enterprises involved in 18 key regulated hazardous chemical processes, and the number of enterprises in high-risk sub-sectors; characteristic indicator types that characterize the safety production level of hazardous chemicals areas, such as the number of major hazard sources, the number of low-alarm alarms for liquid level indicators, the number of pressure indicator alarms, the number of flammable and toxic gas indicator alarms, the number of temperature indicator alarms, the total number of major hazard source alarms, the safety commitment announcement rate, and the online performance rates of the main, technical, and operational responsible persons for safety protection of hazardous chemical enterprises, which are counted on a semi-annual, quarterly, and weekly basis ... Types of characteristic indicators of emergency command capabilities in hazardous chemicals areas: average score of "fire-land cooperation" supervision and inspection, rate of enterprises owning emergency command platforms, rate of emergency drills organized and implemented in accordance with relevant regulations; types of characteristic indicators that characterize the level of employees in hazardous chemicals areas: proportion of safety engineering professionals with a certain chemical professional background, proportion of personnel with professional qualification certificates or professional skill registration certificates, proportion of qualified training for the three types of responsible persons for major hazardous sources; types of characteristic indicators that characterize the digital supervision capabilities of hazardous chemicals areas: proportion of enterprise data access to the hazardous chemicals production safety risk monitoring and early warning system, rate of compliance with inspections and random checks, rate of completion of major hazardous source enterprises, rate of excellent operating results, rate of completion of government-side construction, rate of completion of enterprise platform construction for the construction and application of digital intelligent management and control platforms, and rate of completion of chemical park platform construction.

[0076] After obtaining multiple different types of characteristic indicators of the above-mentioned multiple hazardous chemical areas, the processor can form an indicator data set.

[0077] In one embodiment of the present application, before labeling the indicator data set as to whether it is a key area of ​​hazardous chemicals and labeling the key area level to obtain a training set, the method further includes:

[0078] Step S110: screening the characteristic indicators whose values ​​in the indicator data set are not within a preset range as abnormal values;

[0079] Step S120: Determine the characteristic indicator type to which the abnormal value belongs;

[0080] Step S130: determining the mean and standard deviation of all characteristic indicators included in the characteristic indicator type;

[0081] Step S140: determining the expected guarantee range according to the mean and standard deviation;

[0082] Step S150: Replace the abnormal value with a random number in the expected guaranteed range.

[0083] In order to ensure the accuracy of the indicator data set, after the indicator data set is formed, the indicator data set can also be preprocessed to correct the outliers in the indicator data set. Specifically, it is first necessary to filter out outliers whose values ​​in the indicator data set are not within the preset range, such as the characteristic indicator types of compliance rate, completion rate, and completion rate, whose preset range is [0, 1], that is, the characteristic indicators of the compliance rate, completion rate, and completion rate that are not within this range are all outliers. The preset ranges of other characteristic indicator types can be adaptively set according to the actual situation of the hazardous chemicals area.

[0084] After filtering out the outliers in the indicator data set, determine which characteristic indicator type the outlier belongs to, and then calculate the mean and standard deviation of all characteristic indicators contained in the characteristic indicator type. If the characteristic indicator type to which the outlier belongs is the inspection and spot check compliance rate, then calculate the mean and standard deviation of the characteristic indicators of all inspection and spot check compliance rates.

[0085] After determining the above mean and standard deviation, the processor can determine an expected guarantee range based on the mean and standard deviation. For example, if the mean is 0.5 and the standard deviation is 0.1, the expected guarantee range determined by the mean ± 3 times the standard deviation can ensure that 99.73% of the data meets expectations, so the above expected guarantee range can be determined based on the mean ± 3 times the standard deviation, that is, the expected guarantee range is [0.2, 0.8].

[0086] The processor generates a random number within a certain expected guarantee range to replace the abnormal value, thereby ensuring the rationality and accuracy of the characteristic indicators in the indicator data set.

[0087] Step S200: labeling the indicator data set to determine whether it is a key area for hazardous chemicals and labeling the level of the key area to obtain a training set.

[0088] The training method of the hazardous chemicals key area classification model provided in this embodiment is intended to train a hazardous chemicals key area classification model that can classify hazardous chemicals areas as key areas and which key area level they are. Therefore, the indicator data set is labeled for whether it is a hazardous chemicals key area, and for the key area level, to obtain a training set for model training. Exemplarily, whether it is a hazardous chemicals key area is labeled with 0 or 1, and the key area level (national, provincial, and municipal) is labeled using one-hot encoding.

[0089] Step S300: input the training set into the classification model to be trained, and perform model training to obtain a classification model for key areas of hazardous chemicals.

[0090] The training set is input into the classification model to be trained to perform model training. After the model training is completed, the obtained classification model for key areas of hazardous chemicals can predict and classify whether the area to be predicted is a key area and the specific level of the key area.

[0091] Through the training method of the hazardous chemicals key area classification model in the above embodiment, a training set for training the hazardous chemicals key area classification model can be established according to the quantified characteristic indicators of multiple hazardous chemicals areas. The hazardous chemicals key area classification model obtained after the training can accurately predict and classify whether the area to be predicted is a key area and the key area level of the area to be predicted according to the characteristic indicators of the area to be predicted, thereby realizing the quantification, standardization and rationalization of hazardous chemicals area classification.

[0092] In one embodiment of the present application, the classification model to be trained includes a first model to be trained for classifying whether it is a key area, and a second model to be trained for classifying the level of the key area. The second model to be trained includes multiple sub-classification models of different types. The training set is input into the classification model to be trained, and model training is performed to obtain a key area classification model for hazardous chemicals, including steps S310-S330.

[0093] Step S310: inputting the training set into the first model to be trained, and performing model training on the first model to be trained using the Wasserstein centroid to obtain a first model;

[0094] The training method of the hazardous chemicals key area classification model provided in this embodiment is intended to train a hazardous chemicals key area classification model that can classify whether a hazardous chemicals area is a key area, and which key area level it is, that is, the hazardous chemicals key area classification model needs to first perform a binary classification task - whether it is a key area, and then perform a multi-classification task - which key area level it is. The embodiment of the present application uses two different models to complete the above two classification tasks, namely, the first model that completes the binary classification task and the second model that completes the multi-classification task. Among them, the first model is obtained by training the first model to be trained using the Wasserstein centroid.

[0095] In one embodiment of the present application, a training set is input into a first model to be trained, and the first model to be trained is trained using the Wasserstein center of gravity to obtain a first model, including:

[0096] Step S311: during the training process of each layer of the first model to be trained, the local gradient is determined according to the samples provided by the training set and the loss function of the first model to be trained, and the local gradient is scaled and normalized;

[0097] Step S312: determining the Wasserstein centroid of all the local gradients that have been scaled and normalized, and using the Wasserstein centroid as the global gradient;

[0098] Step S313: normalizing the global gradient;

[0099] Step S314: using the normalized global gradient to update the model parameters of the first model to be trained.

[0100] Wasserstein distance (also known as Earth Mover's Distance) is a metric used to measure the difference between two probability distributions. It computes the minimum transportation cost required to transform one distribution into another, where the cost of each point is its distance to the corresponding point in the other distribution multiplied by its weight. Wasserstein distance is a more efficient way to measure the difference between distributions because it takes into account the geometric structure and probability mass between the distributions. Wasserstein centroid is a method of merging multiple distributions into one. The idea is to find a new distribution that minimizes the sum of its Wasserstein distances to the original distributions. The centroid can be thought of as an average distribution that preserves the structure and probability mass of each original distribution.

[0101] The embodiment of the present application uses the Wasserstein center of gravity obtained by Sinkhorn's algorithm to update the global gradient of model training, which maximizes the retention of the characteristics of each gradient selected in the gradient descent and fuses them to find the most effective gradient descent direction in a shorter number of iterations, and achieves the best model training effect with fewer training times.

[0102] Specifically, first input the number of iterations niter, batch size batch size, learning rate i, initial model parameter θ0, training number epoch, and distribution dimension N and regularization coefficient ∈ required by Sinkhorn's algorithm, and then use the training set obtained in step S200 to perform model training:

[0103] Within the epoch number, each batch is trained;

[0104] In each batch, each network layer of the first model to be trained is trained;

[0105] In each layer, the local gradient is calculated based on the samples provided by the training set and the loss function of the first model to be trained, and then scaled and normalized.

[0106] Use these local gradients to calculate the Wasserstein center of gravity and get the global gradient;

[0107] The global gradient is normalized and used together with the learning rate i to update the model parameters.

[0108] The calculation of Wasserstein centroid uses Sinkhorn's algorithm, which is an iterative algorithm used to solve the inverse of a positive definite matrix with a sparse structure. Assume that there are R distributions (local gradients), each distribution can be represented by a probability matrix a k , the goal of computing R distributions is to find a probability matrix b that is the sum of all a k The weighted average of b and a k The sum of the Wasserstein distances between them is the smallest. Therefore, the Wasserstein center of gravity b can be expressed as:

[0109]

[0110] In formula (1), λ k is a positive weight, and ∑ k λ k =1.

[0111] First, using Sinkhorn's algorithm, we can transform the R distributions into a matrix K whose diagonal elements are all 1, so that ak The Wasserstein distance between a and b can be expressed as the Frobenius norm of K, so the matrix K can be expressed as:

[0112]

[0113] In the above formulas (1) and (2), ∈ is a regularization coefficient that controls the accuracy and speed of the solution.

[0114] According to Sinkhorn's algorithm, the optimal optimal coupling matrix for entropy regularization problems is calculated It can be expressed as:

[0115] P k =diag(u k )Kdiag(v k ) (3)

[0116] Among them, u k 、v k It can be understood as a variable in Sinkhorn's algorithm, u k It can be calculated by the following formula:

[0117]

[0118] v k It can be calculated by the following formula:

[0119]

[0120] According to Kantorovich relaxation theory, the Wasserstein center of gravity b can be iteratively updated by the mass conservation constant of the optimal transmission until convergence is reached, that is, the difference between the current Wasserstein center of gravity b and the Wasserstein center of gravity b obtained by the previous iterative update satisfies the preset convergence condition. The mass conservation constant of the optimal transmission can be expressed as the following formula:

[0121] P T II=v⊙(K T u)=b (6)

[0122] Where ⊙ represents the Hadamard product, which is the product of the corresponding elements between matrices. T Ⅱ Table 2 The transposed matrix of matrix P is the matrix obtained after each iteration.

[0123] By using the method in the above embodiment to perform model training based on the Wasserstein center of gravity on the first model to be trained, the features of each gradient selected in the gradient descent can be retained to the maximum extent and integrated to find the most effective gradient descent direction within a shorter number of iterations, thereby improving the stability and convergence speed of the model, making the acquisition process of the first model more high-quality and efficient.

[0124] In one embodiment of the present application, the first model is a lightweight logistic regression model.

[0125] The first model can be obtained by selecting a lightweight logistic regression model for training. The lightweight logistic regression model has the advantages of fast convergence speed and suitability for binary classification problems. It is more suitable for the binary classification task of whether it is a key area for hazardous chemicals that needs to be completed in the embodiment of the present application.

[0126] Step S320: input the training set into each sub-classification model to perform model training respectively, and then perform model fusion to obtain a second model.

[0127] The second model is used to complete the multi-classification task of "which specific key area level". The models to be trained corresponding to the second model include multiple sub-classification models of different types. First, it is necessary to train the multiple sub-classification models separately, and then perform model fusion operation on the trained sub-classification models to combine the advantages and disadvantages of multiple models and improve the generalization ability of the model.

[0128] In one embodiment of the present application, the training set is input into each sub-classification model for model training respectively, and then the model is fused to obtain a second model, including:

[0129] Step S321: Perform model training on each sub-classification model using a five-fold cross-validation method to obtain five prediction results for each sub-classification model;

[0130] Step S322: stacking the five prediction results of each sub-classification model vertically to form a prediction result matrix of each sub-classification model;

[0131] Step S323: horizontally splicing all prediction result matrices to form a training set matrix;

[0132] Step S324: determining the average value of the five prediction results of each sub-classification model to obtain a prediction result average value matrix of each sub-classification model;

[0133] Step S325: Transversely transform all prediction result average matrices to form a test set matrix;

[0134] Step S326: using the training set matrix to perform model training on the fusion model, and using the test set matrix to test the trained fusion model to be trained;

[0135] Step S327: When the test result meets the preset requirements, all trained sub-classification models and the trained fusion model are merged into a second model according to the hierarchical relationship that the output of the sub-classification model is the input of the fusion model.

[0136] The embodiment of the present application adopts the Stacking method to perform model fusion. The Stacking fusion model generally has two layers of models. The first layer of models is the above-mentioned multiple sub-classification models, which train the original training set and output new features for the next layer of models to use; the learning "learning material" of the second layer of models is the output of the previous layer, and the output of the second layer is the final classification result. Therefore, in order to fuse the above-mentioned multiple sub-classification models, a fusion model is required to use the output of the above-mentioned multiple sub-classification models for learning to output the final classification result.

[0137] In one embodiment of the present application, the sub-classification models include an XGBoost model, a LightGBM model, and a CATBoost model, and the fusion model is a Bayesian regression model.

[0138] The sub-classification models selected in the embodiments of the present application include the following three: XGBoost model, LightGBM model and CATBoost model, and the fusion model selects the Bayesian regression model.

[0139] The XGBoost model uses Classification and Regression Trees (CART) as the basic classifier and adopts the gradient boosting decision tree method in ensemble learning for training to improve the training speed and accuracy of the model. It can solve problems such as classification and regression. In addition, the XGBoost model uses a Boosting algorithm based on gradient decision trees, which has the advantages of fast, efficient, and strong generalization ability, and can achieve good results in the prediction field.

[0140] Similarly, LightGBM is also a decision tree based on a learning algorithm. It uses a histogram-based decision tree algorithm. The basic idea is to first discretize the continuous floating-point eigenvalues ​​into k integers, and construct a histogram with a width of k. When traversing the data, the statistics are accumulated in the histogram based on the discretized values ​​as indexes. After traversing the data once, the histogram accumulates the required statistics, and then traverses to find the optimal segmentation point based on the discrete values ​​of the histogram.

[0141] CATBoost uses a symmetric full binary tree. Compared with XGBoost and LightGBM, the tree used by CatBoost is a simpler structure, avoiding the danger of overfitting. In addition, due to the simple structure of CATBoost's basic model, it can predict faster.

[0142] First, the training set is used to train the three sub-classification models using the five-fold cross-validation method. For example, the training set data is first divided into five parts, and then training and testing are performed as follows:

[0143] In the first round, the data from the first to fourth parts are used to train the model, and the data from the fifth part is used to test the model;

[0144] In the second round, the data from the second to fifth parts are used to train the model, and the data from the first part is used to test the model;

[0145] In the third round, the data from the third to the first part is used to train the model, and the data from the second part is used to test the model.

[0146] In the fourth round, the data from the fourth to the second part is used to train the model, and the data from the third part is used to test the model.

[0147] In the fifth round, the data from the fifth to third parts are used to train the model, and the data from the fourth part is used to test the model.

[0148] That is, in each round, different parts of the training set data are used to train the sub-classification model, and the remaining part of the data is used to test the sub-classification model to obtain the prediction results of the sub-classification model. In this way, five prediction results of each sub-classification model can be obtained, and the five prediction results of each sub-classification model are stacked vertically to obtain the prediction result matrix of each sub-classification model. Then the three prediction result matrices are spliced ​​horizontally to obtain the training set matrix for training the fusion model.

[0149] The fusion model also needs to be tested after training. Therefore, the average or weighted average of the five prediction results of each sub-classification model is calculated to obtain the prediction result average matrix of each sub-classification model. The three prediction result average matrices are then horizontally spliced ​​to obtain the test set matrix for fusion model testing.

[0150] After the training set matrix and the test set matrix are determined, the training set matrix is ​​used to train the fusion model. After the training is completed, the test set matrix is ​​used to test the fusion model.

[0151] When the test result of the fusion model meets the preset requirements, that is, the difference between the classification result output by the fusion model and the true classification meets the preset requirements, the training process of the second model is completed. As long as the output of the sub-classification model is the input of the fusion model, the trained sub-classification model is merged with the fusion model to obtain the second model.

[0152] By using the method in the above embodiment to train each sub-classification model included in the second model to be trained, and then using the Stacking method to fuse the models, the advantages and disadvantages of multiple models can be combined to improve the generalization ability of the model and make the classification of the model more accurate.

[0153] Step S330: merging the first model and the second model into a hazardous chemicals key area classification model.

[0154] By merging the trained first model with the second model, we can obtain a hazardous chemicals key area classification model that can perform two classification tasks: whether it is a hazardous chemicals key area and the key area level, which can be used for the classification of hazardous chemicals key areas in subsequent applications.

[0155] In one embodiment of the present application, the method further includes:

[0156] According to the indicator data set, a heat map representing the correlation between multiple indicator feature types included in the indicator data set and whether it is a key area for hazardous chemicals is determined;

[0157] Determine the correlation coefficient between all characteristic indicator types and whether it is a key area for hazardous chemicals based on the heat map;

[0158] Multiple important indicator feature types are determined based on all correlation coefficients.

[0159] In order to propose targeted assistance plans for each key hazardous chemical area in the future, it is necessary to provide a reference for the formulation of the assistance plan. Therefore, in the method provided in the embodiment of the present application, after obtaining the indicator data set, a heat map representing the correlation between multiple indicator feature types included in the indicator data set and whether it is a key hazardous chemical area can be drawn according to the indicator data set. The correlation coefficient between all characteristic indicator types and whether it is a key hazardous chemical area can be determined according to the heat map. After the correlation coefficient is determined, the correlation coefficients of all characteristic indicator types can be sorted, and the characteristic indicator types with the top ten correlation coefficients can be selected as important indicators as a reference for the formulation of subsequent assistance plans.

[0160] Through the above technical scheme, according to the quantified characteristic indicators of multiple hazardous chemicals areas, a training set for training the classification model of key hazardous chemicals areas can be established. The classification model of key hazardous chemicals areas obtained after training can accurately predict and classify whether the predicted area is a key area and the key area level of the predicted area according to the characteristic indicators of the area to be predicted, thereby realizing the quantification, standardization and rationalization of the classification of hazardous chemicals areas. At the same time, the first model included in the classification model of key hazardous chemicals areas adopts a model optimizer based on the Wasserstein center of gravity during the training process, which improves the stability and convergence speed of the model, making the acquisition process of the first model more high-quality and efficient; the second model included in the classification model of key hazardous chemicals areas adopts a multi-classification model Stacking fusion method during the training process, combining the advantages and disadvantages of multiple models and improving the generalization ability of the model. In addition, the indicator data set containing a large number of quantitative indicators can also be used to determine the important characteristic indicator types related to whether it is a key hazardous chemical area, providing a reference for the subsequent designated assistance plan.

[0161] Figure 2 A flowchart of a method for classifying key areas of hazardous chemicals according to an embodiment of the present application is schematically shown. Figure 2 As shown, in one embodiment of the present application, a classification method for key areas of hazardous chemicals is provided, the method comprising:

[0162] Step S400: Acquire characteristic indicators of the area to be predicted to form a data set to be predicted;

[0163] Step S500: Input the data set to be predicted into the hazardous chemicals key area classification model, and obtain the classification result of the area to be predicted output by the hazardous chemicals key area classification model, wherein the classification result is whether the area to be predicted is a key area, or the key area level of the area to be predicted, wherein the hazardous chemicals key area classification model is obtained according to the training method of the hazardous chemicals key area classification model provided in the above embodiment.

[0164] The area to be predicted, that is, whether it is a key area or not, and the hazardous chemicals area to be classified into the key area grade classification, firstly, various characteristic indicators of the area to be predicted are obtained to form a data set to be predicted, and then the data set to be predicted is input into the hazardous chemicals key area classification model obtained by the training method of the hazardous chemicals key area classification model mentioned above, so as to obtain the classification result output by the hazardous chemicals key area classification model. If the area to be predicted is a hazardous chemicals key area, the key area grade of the area to be predicted is output; if the area to be predicted is not a hazardous chemicals key area, the classification result that the area to be predicted is not a hazardous chemicals key area is output, thereby realizing the quantification, standardization and rationalization of hazardous chemicals area classification.

[0165] In one embodiment of the present application, before inputting the data set to be predicted into the hazardous chemicals key area classification model and obtaining the classification result of the area to be predicted output by the hazardous chemicals key area classification model, the method further includes:

[0166] Filter the characteristic indicators whose values ​​in the data set to be predicted exceed the preset range as outliers;

[0167] Determine the characteristic indicator type to which the outlier belongs;

[0168] Determine the mean and standard deviation of all feature indicators included in the feature indicator type;

[0169] Determine the expected guarantee range based on the mean and standard deviation;

[0170] Replace outliers with random numbers in the desired guaranteed range.

[0171] After the data set to be predicted is determined, before the data set to be predicted is input into the classification model for key areas of hazardous chemicals, the data set to be predicted can also be preprocessed in the same way as steps S110-S150 above to correct abnormal values ​​in the data set to be predicted and ensure the rationality and accuracy of the characteristic indicators in the data set to be predicted.

[0172] In one embodiment of the present application, the hazardous chemicals key area classification model includes a first model for classifying whether it is a key area, and a second model for classifying the key area level. The data set to be predicted is input into the hazardous chemicals key area classification model, and the classification result of the area to be predicted output by the hazardous chemicals key area classification model is obtained, including:

[0173] Step S510: inputting the data set to be predicted into the first model to determine whether the area to be predicted is a key area;

[0174] Step S520: if the area to be predicted is not a key area, outputting the classification result that the area to be predicted is not a key area;

[0175] Step S530: when the area to be predicted is a key area, inputting the prediction data set into the second model;

[0176] Step S540: outputting the key area level of the to-be-predicted area of ​​the second model data as a classification result.

[0177] Figure 3 Schematically shows a workflow logic block diagram of a hazardous chemicals key area classification model according to an embodiment of the present application, please refer to Figure 2 and Figure 3, the data set to be predicted first enters the first model for a binary classification task to determine whether it is a key area for hazardous chemicals. If it is not a key area for hazardous chemicals, the classification result that the area to be predicted is not a key area for hazardous chemicals is input. If the result of the binary classification task is that the predicted area is a key area for hazardous chemicals, the data set to be predicted is input into the second model to perform a multi-classification task and output the specific key area level of the area to be predicted. Specifically, after the data set to be predicted enters the second model, it first enters multiple sub-classification models to obtain the classification results of multiple sub-classification models, and the classification results of multiple sub-classification models are then used as the input of the fusion model to obtain the final classification result.

[0178] Figure 4 The structural block diagram of a training device for a hazardous chemicals key area classification model according to an embodiment of the present application is schematically shown, such as Figure 4 As shown, the device 1000 includes:

[0179] The indicator data set forming module 1001 is used to obtain multiple different types of characteristic indicators of multiple hazardous chemical areas to form an indicator data set;

[0180] The labeling module 1002 is used to label the preprocessed indicator data set as to whether it is a key area for hazardous chemicals, and label the level of the key area, so as to obtain a training set;

[0181] The model training module 1003 is used to input the training set into the classification model to be trained, perform model training, and obtain a classification model for key areas of hazardous chemicals.

[0182] The training device 1000 for the hazardous chemicals key area classification model provided in the embodiment of the present application can implement each process of step S100 to step S300 in the method embodiment and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0183] Figure 5 A structural block diagram of a classification device for key areas of hazardous chemicals according to an embodiment of the present application is schematically shown, such as Figure 5 As shown, the device 2000 includes:

[0184] The module 2001 for forming a data set to be predicted is used to obtain characteristic indicators of the area to be predicted and form a data set to be predicted;

[0185] The classification result acquisition module 2002 is used to input the data set to be predicted into the hazardous chemicals key area classification model, and obtain the classification result of the area to be predicted output by the hazardous chemicals key area classification model, wherein the classification result is whether the area to be predicted is a key area, or the key area level of the area to be predicted, wherein the hazardous chemicals key area classification model is obtained according to the training method of the hazardous chemicals key area classification model as claimed in claims 1-8.

[0186] The classification device 2000 for key areas of hazardous chemicals provided in the embodiment of the present application can implement each process of step S400 to step S500 in the method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0187] In one embodiment of the present application, an electronic device is provided, including a processor and a memory, the memory storing machine executable instructions that can be executed by the processor, and the processor can execute the machine executable instructions to implement the training method of the hazardous chemicals key area classification model in the above embodiment, or the hazardous chemicals key area classification method in the above embodiment.

[0188] In one embodiment of the present application, a machine-readable storage medium is provided, on which instructions are stored. When the instructions are executed by a processor, the processor implements the training method of the hazardous chemicals key area classification model in the above embodiment, or the classification method of the hazardous chemicals key areas in the above embodiment.

[0189] In one embodiment of the present application, a computer program product is also provided, including a computer program, which, when executed by a processor, implements the training method of the hazardous chemicals key area classification model in the above embodiment, or the classification method of the hazardous chemicals key area in the above embodiment.

[0190] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD ROM, optical storage, etc.) that include computer-usable program code.

[0191] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including an instruction device, which implements the functions specified in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0192] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0193] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0194] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0195] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0196] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A training method for a classification model for key areas of hazardous chemicals, characterized in that: The method comprises: Obtain multiple different types of characteristic indicators of multiple hazardous chemical areas to form an indicator data set; The indicator data set is labeled as to whether it is a key area for hazardous chemicals, and the key area level is labeled to obtain a training set; The training set is input into the classification model to be trained, and model training is performed to obtain the classification model for key areas of hazardous chemicals.

2. The method according to claim 1, characterized in that Before labeling the indicator data set as to whether it is a key area for hazardous chemicals and labeling the key area level to obtain a training set, the method further includes: Screening the characteristic indicators whose values ​​in the indicator data set are not within a preset range as the abnormal values; Determine the characteristic indicator type to which the abnormal value belongs; Determine the mean and standard deviation of all characteristic indicators included in the characteristic indicator type; Determine an expected guarantee range based on the mean and the standard deviation; The abnormal value is replaced by a random number in the expected guaranteed range.

3. The method according to claim 1, characterized in that The classification model to be trained includes a first model to be trained for classifying whether it is a key area, and a second model to be trained for classifying the key area level, the second model to be trained includes a plurality of sub-classification models of different types, and the training set is input into the classification model to be trained to perform model training to obtain the classification model for the key area of ​​hazardous chemicals, including: Inputting the training set into the first model to be trained, and performing model training on the first model to be trained using the Wasserstein center of gravity to obtain a first model; Inputting the training set into each of the sub-classification models to perform model training respectively, and then performing model fusion to obtain a second model; The first model and the second model are combined into the hazardous chemicals key area classification model.

4. The method according to claim 3, characterized in that The step of inputting the training set into the first model to be trained, and performing model training on the first model to be trained using the Wasserstein center of gravity to obtain a first model includes: During the training process of each network layer of the first model to be trained, a local gradient is determined according to the samples provided by the training set and the loss function of the first model to be trained, and the local gradient is scaled and normalized; Determine the Wasserstein centroid of all the local gradients that have undergone the scaling and normalization processing, and use the Wasserstein centroid as the global gradient; Normalizing the global gradient; The model parameters of the first model to be trained are updated using the global gradient that has undergone the normalization processing.

5. The method according to claim 3, characterized in that: The first model is a lightweight logistic regression model.

6. The method according to claim 3, characterized in that: The step of inputting the training set into each of the sub-classification models to perform model training respectively, and then performing model fusion to obtain a second model, comprises: Performing model training on each of the sub-classification models using a five-fold cross-validation method to obtain five prediction results for each of the sub-classification models; Stacking the five prediction results of each sub-classification model vertically to form a prediction result matrix of each sub-classification model; All the prediction result matrices are horizontally spliced ​​to form a training set matrix; Determine the average of the five prediction results of each of the sub-classification models to obtain a prediction result average matrix of each of the sub-classification models; All the prediction result average matrices are horizontally transformed to form a test set matrix; The training set matrix is ​​used to perform model training on the fusion model, and the test set matrix is ​​used to test the trained fusion model to be trained; When the result of the test meets the preset requirements, all the trained sub-classification models and the trained fusion model are merged into the second model according to the hierarchical relationship that the output of the sub-classification model is the input of the fusion model.

7. The method according to claim 6, characterized in that The sub-classification models include an XGBoost model, a LightGBM model and a CATBoost model, and the fusion model is a Bayesian regression model.

8. The method according to claim 1, characterized in that The method further comprises: Determine, according to the indicator data set, a heat map representing the correlation between a plurality of indicator feature types included in the indicator data set and whether the area is a key area for hazardous chemicals; Determine the correlation coefficient between all the characteristic indicator types and whether it is a key area for hazardous chemicals according to the heat map; A plurality of important indicator feature types are determined according to all the correlation coefficients.

9. A classification method for key areas of hazardous chemicals, characterized in that: The method comprises: Obtain characteristic indicators of the area to be predicted and form a data set to be predicted; The data set to be predicted is input into a hazardous chemicals key area classification model, and a classification result of the area to be predicted output by the hazardous chemicals key area classification model is obtained, wherein the classification result is whether the area to be predicted is a key area, or the key area level of the area to be predicted, wherein the hazardous chemicals key area classification model is obtained according to the training method of the hazardous chemicals key area classification model as described in claims 1-8.

10. The method according to claim 9, characterized in that Before inputting the to-be-predicted data set into the hazardous chemicals key area classification model and obtaining the classification result of the to-be-predicted area output by the hazardous chemicals key area classification model, the method further includes: Screening the characteristic indicators whose values ​​in the to-be-predicted data set exceed a preset range as the abnormal values; Determine the characteristic indicator type to which the abnormal value belongs; Determine the mean and standard deviation of all characteristic indicators included in the characteristic indicator type; Determine an expected guarantee range based on the mean and the standard deviation; The abnormal value is replaced by a random number in the expected guaranteed range.

11. The method according to claim 9, characterized in that The hazardous chemicals key area classification model includes a first model for classifying whether it is a key area, and a second model for classifying the level of the key area. The inputting of the prediction data set into the hazardous chemicals key area classification model to obtain the classification result of the area to be predicted output by the hazardous chemicals key area classification model includes: Inputting the to-be-predicted data set into the first model to determine whether the to-be-predicted area is a key area; In a case where the area to be predicted is not a key area, outputting that the area to be predicted is not a key area as the classification result; In a case where the area to be predicted is a key area, inputting the prediction data set into the second model; The key area level of the to-be-predicted area of ​​the second model data is output as the classification result.

12. A training device for a classification model of key areas of hazardous chemicals, characterized in that: The device comprises: An indicator data set building module is used to obtain multiple different types of characteristic indicators of multiple hazardous chemicals areas to form an indicator data set; A labeling module, used to label the indicator data set as to whether it is a key area for hazardous chemicals, and label the level of the key area, so as to obtain a training set; The model training module is used to input the training set into the classification model to be trained to perform model training to obtain the classification model for key areas of hazardous chemicals.

13. A classification device for key areas of hazardous chemicals, characterized in that: The device comprises: The module for forming the data set to be predicted is used to obtain characteristic indicators of the area to be predicted and form the data set to be predicted; A classification result acquisition module is used to input the data set to be predicted into the hazardous chemicals key area classification model, and obtain the classification result of the area to be predicted output by the hazardous chemicals key area classification model, wherein the classification result is whether the area to be predicted is a key area, or the key area level of the area to be predicted, wherein the hazardous chemicals key area classification model is obtained according to the training method of the hazardous chemicals key area classification model as described in claims 1-8.

14. An electronic device, characterized in that: It includes a processor and a memory, the memory stores machine executable instructions that can be executed by the processor, and the processor can execute the machine executable instructions to implement the training method of the hazardous chemicals key area classification model described in any one of claims 1 to 8, or the hazardous chemicals key area classification method described in claims 9 to 11.

15. A machine-readable storage medium, characterized in that: The machine-readable storage medium stores instructions, which, when executed by a processor, enable the processor to implement the training method for the hazardous chemicals key area classification model according to any one of claims 1 to 8, or the hazardous chemicals key area classification method according to claims 9 to 11.

Citation Information

Patent Citations

  • Fault diagnosis logic algorithm and system

    CN112001440A

  • Method for establishing risk prediction model, regional risk prediction method and corresponding device

    CN112508300A

  • Regional risk level assessment method and device, computer equipment and storage medium

    CN113298438A

  • Model training method and device, waterlogging prediction method and device and electronic equipment

    CN115879007A

  • Machine learning based ranking of hydrocarbon prospects for field exploration

    US20220414522A1