Object classification processing method and device, electronic equipment and storage medium

Abstract

The invention provides an object classification processing method and device, electronic equipment and a storage medium, relates to the technical field of artificial intelligence, and is applied to the field of financial science and technology. The method comprises the following steps: acquiring carbon emission reference data of a carbon emission reference object; performing sample expansion according to the historical claim settlement data of the carbon emission reference object and the carbon emission reference data to obtain a carbon emission sample; performing model training on a preset initial decision tree model according to the carbon emission sample to obtain a target decision tree model; acquiring insurance basic data and current carbon emission data of the carbon emission prediction object; wherein the carbon emission prediction object belongs to or does not belong to one of the carbon emission reference objects; performing risk prediction on the insurance basic data and the current carbon emission data through a target decision tree model to obtain a target risk prediction degree; and classifying the carbon emission prediction objects according to the target risk prediction degree. According to the invention, the classification accuracy of the carbon emission objects in the insurance scene can be improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology and is applied to the field of fintech, and particularly relates to an object classification processing method and device, an electronic device, and a storage medium. Background Art

[0002] Object classification processing technology can be used to predict the risks of objects and then classify the objects. In the insurance scenario in the field of fintech, the object classification processing technology can specifically be used to predict whether there is a claim risk for a carbon emission object (a certain enterprise), and then classify the carbon emission object. For example, an insurance company will provide insurance services related to carbon accounts for carbon emission objects. In order to predict whether there is a claim risk for the future insurance service, the historical insurance claim data of the carbon emission object can be used for risk prediction and estimation to classify whether the carbon emission object is suitable or not suitable for continued insurance. The insurance company can also adjust the insurance pricing according to the classification results.

[0003] In the related art, when performing classification processing in an insurance scenario, risk prediction is often only based on data related to insurance applications (such as enterprise scale, historical insurance pricing, historical claim amount, etc.). However, for carbon emission objects, it is difficult to comprehensively reflect the insurance claim risks of carbon emission objects only relying on these data. Therefore, the accuracy of the classification processing of carbon emission objects in the insurance scenario is relatively poor. Summary of the Invention

[0004] The main purpose of the embodiments of the present application is to propose an object classification processing method and device, an electronic device, and a storage medium, which can improve the classification accuracy of carbon emission objects in the insurance scenario.

[0005] To achieve the above object, a first aspect of the embodiments of the present application proposes an object classification processing method, and the method includes:

[0006] Obtain the carbon emission reference data of the carbon emission reference object;

[0007] Perform sample expansion according to the historical claim data and the carbon emission reference data of the carbon emission reference object to obtain carbon emission samples;

[0008] Train a preset initial decision tree model according to the carbon emission samples to obtain a target decision tree model;

[0009] Obtain the insurance basic data and the current carbon emission data of the carbon emission prediction object; wherein, the carbon emission prediction object belongs to or does not belong to one of the carbon emission reference objects;

[0010] Performing risk prediction on the insured basic data and the current carbon emission data through the target decision tree model to obtain a target risk prediction level; wherein, the target risk prediction level indicates the risk level of insurance claims for the carbon emission prediction object;

[0011] Classifying the carbon emission prediction object according to the target risk prediction level.

[0012] In some embodiments, the carbon emission sample has a risk label value; the method of training a preset initial decision tree model according to the carbon emission sample to obtain a target decision tree model includes:

[0013] Performing risk prediction on the carbon emission sample through the initial decision tree model to obtain a sample risk prediction level; wherein, the sample risk prediction level indicates the risk level of insurance claims for the carbon emission reference object;

[0014] Adjusting the parameters of the initial decision tree model according to the sample risk prediction level and the risk label value to obtain the target decision tree model;

[0015] Wherein, the risk label value is a first value or a second value, the first value indicates that the carbon emission reference object has a claim risk, the second value indicates that the carbon emission reference object has no claim risk, the first value is determined based on a first claim label value and an influence degree label value, the first claim label value indicates that a claim has been made for the carbon emission reference object, the influence degree label value indicates the influence degree of the carbon emission reference data on the carbon emission reference object, the second value is determined based on a second claim label value and the influence degree label value, and the second claim label value indicates that no claim has been made for the carbon emission reference object.

[0016] In some embodiments, before training a preset initial decision tree model according to the carbon emission sample to obtain a target decision tree model, the method further includes:

[0017] Obtaining a target carbon emission reduction task to be performed by the carbon emission reference object;

[0018] Generating the influence degree label value according to the influence degree of the carbon emission reference data on the execution of the target carbon emission reduction task;

[0019] Updating the first claim label value of the carbon emission sample according to the influence degree label value to obtain the first value, or updating the second claim label value of the carbon emission sample according to the influence degree label value to obtain the second value.

[0020] In some embodiments, before training the preset initial decision tree model according to the carbon emission samples to obtain the target decision tree model, the method further includes:

[0021] Constructing a carbon emission sample set based on at least two carbon emission samples;

[0022] Calculating the gradient of each sample in the carbon emission sample set to obtain the sample gradient;

[0023] Sorting the samples in the carbon emission sample set from largest to smallest according to the absolute value of the sample gradient to obtain sample sorting data;

[0024] Starting from the first sample in the sample sorting data, extracting a first preset number ratio of samples to obtain the first retained samples;

[0025] Randomly sampling a second preset number ratio of samples from the sample sorting data except for the first retained samples to obtain the second retained samples; wherein, the second preset number ratio is greater than the first preset number ratio;

[0026] Replacing the samples in the carbon emission sample set according to the first retained samples and the second retained samples to update the carbon emission sample set.

[0027] In some embodiments, the adjusting the parameters of the initial decision tree model according to the sample risk prediction degree and the risk label value to obtain the target decision tree model includes:

[0028] Taking the sample risk prediction degree corresponding to the first retained samples as the first sample risk prediction degree, and taking the risk label value corresponding to the first retained samples as the first risk label value;

[0029] Performing sampling bias learning on the second retained samples through a preset residual neural network to obtain sampling compensation weights;

[0030] Multiplying the sample risk prediction degree corresponding to the second retained samples by the sampling compensation weights to obtain the second sample risk prediction degree, and taking the risk label value corresponding to the second retained samples as the second risk label value;

[0031] Adjusting the parameters of the initial decision tree model according to the first sample risk prediction degree and the first risk label value, and the second sample risk prediction degree and the second risk label value to obtain the target decision tree model.

[0032] In some embodiments, before predicting the risk of the carbon emission sample through the initial decision tree model to obtain the sample risk prediction degree, the method further includes:

[0033] Construct a first object graph network, where the first object graph network indicates the carbon emission reference data of at least two candidate carbon emission objects and the carbon emission association relationship between any two of the candidate carbon emission objects; wherein, the carbon emission reference object belongs to one of the candidate carbon emission reference objects;

[0034] Search for the candidate carbon emission objects in the first object graph network according to the carbon emission reference object to extract the relevant carbon emission reference data and the carbon emission association relationship, and obtain the first carbon emission graph embedding feature;

[0035] Add the first carbon emission graph embedding feature to the carbon emission sample to update the carbon emission sample.

[0036] In some embodiments, the predicting the risk of the insured basic data and the current carbon emission data through the target decision tree model to obtain the target risk prediction degree includes:

[0037] Extract the carbon emission indicators and the carbon emission association relationship related to the carbon emission prediction object from a preset second object graph network to obtain a second carbon emission graph embedding feature; wherein, the second object graph network indicates the carbon emission indicators of at least two candidate carbon emission objects and the carbon emission association relationship between any two of the candidate carbon emission objects; wherein, the carbon emission prediction object belongs to one of the candidate carbon emission reference objects;

[0038] Extract features from the current carbon emission data to obtain the current carbon emission feature;

[0039] Extract features from the insured basic data to obtain the insured basic feature;

[0040] Generate a risk prediction basic feature according to the second carbon emission graph embedding feature, the current carbon emission feature and the insured basic feature;

[0041] Predict the risk of the risk prediction basic feature through the target decision tree model to obtain the target risk prediction degree.

[0042] To achieve the above object, a second aspect of the embodiments of the present application proposes an object classification processing device, and the device includes:

[0043] A first data acquisition module, configured to acquire the carbon emission reference data of the carbon emission reference object;

[0044] A sample augmentation module, configured to perform sample augmentation based on the historical claim settlement data of the carbon emission reference object and the carbon emission reference data to obtain carbon emission samples;

[0045] A model training module, configured to perform model training on a preset initial decision tree model according to the carbon emission samples to obtain a target decision tree model;

[0046] A second data acquisition module, configured to acquire the insured basic data and the current carbon emission data of the carbon emission prediction object; wherein, the carbon emission prediction object belongs to or does not belong to one of the carbon emission reference objects;

[0047] A risk prediction module, configured to perform risk prediction on the insured basic data and the current carbon emission data through the target decision tree model to obtain a target risk prediction degree; wherein, the target risk prediction degree indicates the risk degree of insurance claim settlement for the carbon emission prediction object;

[0048] An object classification module, configured to classify the carbon emission prediction object according to the target risk prediction degree.

[0049] To achieve the above object, a third aspect of the embodiments of the present application proposes an electronic device, which includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, it implements the object classification processing method described in the first aspect above.

[0050] To achieve the above object, a fourth aspect of the embodiments of the present application proposes a storage medium, which is a computer-readable storage medium, the storage medium stores a computer program, and when the computer program is executed by a processor, it implements the object classification processing method described in the first aspect above.

[0051] In the object classification processing method, object classification processing device, electronic device and computer-readable storage medium proposed by the present application, in the model training stage, a decision tree model is selected to provide the risk prediction function, and the historical claim settlement data and carbon emission reference data belonging to the same carbon emission reference object are used for sample augmentation to obtain carbon emission samples, and then the initial decision tree model is trained according to the carbon emission samples to obtain a target decision tree model. In this way, the target decision tree model can perform risk prediction based on insurance-related data and carbon emission-related data. In the model usage stage, risk prediction is performed on the insured basic data and the current carbon emission data through the target decision tree model, and then the carbon emission prediction object is classified according to the target risk prediction degree. In this way, by considering both insurance-related data and carbon emission-related data, the accuracy of risk prediction is improved, and thus the accuracy of object classification is improved. In summary, the present application can improve the classification accuracy of carbon emission objects in the insurance scenario.

[0052] Additional aspects and advantages of the present application will be given in part in the following description, become apparent in part from the following description, or be learned through the practice of the present application. Brief Description of the Drawings

[0053] Figure 1 is a flowchart of an object classification processing method provided by an embodiment of the present application;

[0054] Figure 2 is Figure 1 a flowchart of step 103 in

[0055] Figure 3 is a flowchart of an object classification processing method provided by another embodiment of the present application;

[0056] Figure 4 is a flowchart of an object classification processing method provided by another embodiment of the present application;

[0057] Figure 5 is Figure 2 a flowchart of step 202 in

[0058] Figure 6 is a block diagram of the module structure of an object classification processing device provided by an embodiment of the present application;

[0059] Figure 7 is a schematic diagram of the hardware structure of an electronic device provided by an embodiment of the present application. Detailed Description of the Embodiments

[0060] In order to make the objectives, technical solutions and advantages of the present application more clear and understandable, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0061] It should be noted that although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order from the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the description, claims and the above-mentioned drawings are used to distinguish similar objects and do not necessarily need to describe a specific order or sequence.

[0062] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present application belongs. The terms used herein are only for the purpose of describing the embodiments of the present application and are not intended to limit the present application.

[0063] First, several nouns involved in the present application are analyzed:

[0064] Artificial Intelligence (AI): It is a new technical science that studies and develops theories, methods, technologies, and application systems for simulating, extending, and expanding human intelligence. Artificial intelligence is a branch of computer science. It attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a way similar to human intelligence. The research in this field includes robots, speech recognition, image recognition, natural language processing, and expert systems, etc. Artificial intelligence can simulate the information process of human consciousness and thinking. It also uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to obtain the best results in terms of theories, methods, technologies, and application systems.

[0065] Natural Language Processing (NLP): NLP uses computers to process, understand, and apply human languages (such as Chinese, English, etc.). NLP is a branch of artificial intelligence and an interdisciplinary field of computer science and linguistics, and is often referred to as computational linguistics. Natural language processing includes syntactic analysis, semantic analysis, discourse understanding, etc. Natural language processing is commonly used in technical fields such as machine translation, recognition of handwritten and printed characters, speech recognition, text-to-speech conversion, information image processing, information extraction and filtering, text classification and clustering, public opinion analysis, and opinion mining. It involves data mining, machine learning, knowledge acquisition, knowledge engineering, artificial intelligence research related to language processing, and linguistic research related to language computing, etc.

[0066] A carbon account refers to a record account that defines the carbon footprint, carbon emission rights boundary, and carbon reduction contributions of social entities such as individuals and enterprises. The carbon account provides a dynamic monitoring tool for implementing the carbon reduction responsibilities of each participating entity (also known as the carbon emission object), and also provides the corresponding micro foundation for the effective operation of the carbon market. However, the construction of the carbon account is still in its infancy, and the construction of the incentive mechanism and scenario application is also in its initial stage. Especially in the insurance scenario in the field of fintech, the carbon account has a certain potential impact on insurance claim risks and insurance pricing, but currently, this potential impact cannot be effectively explored, resulting in poor accuracy in classifying the carbon emission objects.

[0067] Based on this, the embodiments of this application propose an object classification processing method, an object classification processing device, an electronic device, and a computer-readable storage medium. An augmented sample generated based on historical insurance claim data and carbon emission data during the training stage is used to train a target decision tree model, so as to perform risk prediction through the target decision tree model during the application stage, improve the accuracy of risk prediction, and further improve the accuracy of object classification.

[0068] The object classification processing method provided by the embodiments of the present application can be applied to terminals and server sides, and can also be software running on the server side. The server side can be configured as an independent physical server, or as a server cluster or distributed system composed of multiple physical servers, and can also be configured as a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements the object classification processing method, etc., but is not limited to the above forms.

[0069] The present application can be used in many general or special computer system environments or configurations. For example: server computers, multi-processor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, distributed computing environments including any of the above systems or devices, and so on. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0070] The embodiments of the present application provide an object classification processing method, an object classification processing device, an electronic device, and a storage medium, which will be specifically described through the following embodiments. First, the object classification processing method in the embodiments of the present application will be described.

[0071] It should be noted that in each specific implementation manner of the present application, when it comes to relevant processing that needs to be based on user behavior data and other data related to user identity or characteristics, user permission or consent will be obtained first, and moreover, the collection, use, and processing of these data will comply with relevant laws, regulations, and standards.

[0072] Refer to Figure 1 , Figure 1 is an optional flowchart of the object classification processing method provided by the embodiments of the present application, which may include but is not limited to steps 101 to 106.

[0073] Step 101, obtain the carbon emission reference data of the carbon emission reference object;

[0074] Step 102, perform sample expansion according to the historical claim data and carbon emission reference data of the carbon emission reference object to obtain carbon emission samples;

[0075] Step 103: Train a preset initial decision tree model according to the carbon emission samples to obtain a target decision tree model;

[0076] Step 104: Obtain the insured basic data and current carbon emission data of the carbon emission prediction object; wherein, the carbon emission prediction object belongs to or does not belong to one of the carbon emission reference objects;

[0077] Step 105: Perform risk prediction on the insured basic data and current carbon emission data through the target decision tree model to obtain a target risk prediction level; wherein, the target risk prediction level indicates the risk level of insurance claims for the carbon emission prediction object;

[0078] Step 106: Classify the carbon emission prediction objects according to the target risk prediction level.

[0079] In steps 101 to 106 illustrated in the embodiments of the present application, in the model training stage, a decision tree model is selected to provide a risk prediction function, and historical claim data and carbon emission reference data belonging to the same carbon emission reference object are used for sample expansion to obtain carbon emission samples. Then, the initial decision tree model is trained according to the carbon emission samples to obtain a target decision tree model. In this way, the target decision tree model can perform risk prediction based on insurance-related data and carbon emission-related data. In the model usage stage, the target decision tree model is used to perform risk prediction on the insured basic data and current carbon emission data, and then the carbon emission prediction objects are classified according to the target risk prediction level. In this way, by considering both insurance-related data and carbon emission-related data, the accuracy of risk prediction is improved, and thus the accuracy of object classification is improved. In summary, the present application can improve the classification accuracy of carbon emission objects in the insurance scenario.

[0080] In step 101 of some embodiments, obtain the carbon emission reference data of the carbon emission reference object. The carbon emission reference object is a carbon emission object used as training data. The carbon emission object refers to an object that consumes carbon emission indicators, generally referring to energy-consuming enterprises. The carbon emission indicator is, for example, carbon dioxide. The carbon emission reference data refers to the relevant data generated by the carbon emission reference object for emitting carbon emission indicators. For example, the carbon emission reference data includes carbon emissions, achievement of emission reduction targets, carbon trading activity, etc.

[0081] In one embodiment, the carbon emission reference data is recorded in the carbon account of the carbon reference object, and the carbon account is maintained in the target database. Therefore, step 101 may include: determining the carbon account of the carbon emission reference object to obtain a reference carbon account identifier; querying the target database according to the reference carbon account identifier to obtain the carbon emission reference data. Wherein, the target database can be maintained by the carbon emission reference object itself or by the blockchain system.

[0082] The blockchain system involved in the embodiments of the present application can be a distributed system formed by connecting terminal devices and multiple nodes in the form of network communication. In this way, the security of the carbon account reference data is higher.

[0083] In step 102 of some embodiments, sample augmentation is performed based on the historical claim data and carbon emission reference data of the carbon emission reference object to obtain carbon emission samples. Historical claim data refers to various information collected and recorded during the process of processing carbon emission insurance claims, which covers the entire process from reporting a claim to final payment. Specifically, historical claim data usually includes the following content: policy information (such as: insured's name, policy number, insurance type, insurance period, insurance amount, etc.), claim reporting information (such as: claim reporting time, claimant, claim reporting method, accident occurrence time, location, process, etc.), accident information (such as: accident type (e.g., disease, accident, theft, etc.), accident cause, accident liability determination, loss situation (e.g., medical expenses, property losses, etc.)), claim application information (such as: claim application time, applicant, claim amount, submitted claim materials, etc.), investigation information (such as: investigation report of the claim investigator, on-site photos, medical reports, etc.), review information (such as: review opinions of the claim reviewer, approval records, reasons for rejection / payment, etc.), and payment information (such as: payment amount, payment time, payment method, payee information, etc.).

[0084] In an example, an insurance company provides carbon emission insurance for carbon emission reference objects, and the insurance coverage of the carbon emission insurance includes, but is not limited to, the risk of carbon emission exceeding the standard caused by natural disasters, accidents, etc., and the carbon trading losses caused by policy changes, market risks, etc.

[0085] It can be seen that algorithms or models that only rely on insurance claim data for risk prediction in the current related technologies are difficult to comprehensively evaluate the claim risks of carbon emission objects, resulting in relatively low object classification accuracy. However, in the embodiments of the present application, specifically, sample augmentation is performed based on the historical claim data and carbon emission reference data of the carbon emission reference object to obtain carbon emission samples. In this way, the carbon emission samples can not only indicate the characteristic information of the historical claim data but also indicate the characteristic information of the carbon emission reference data. Therefore, based on this carbon emission sample for model training, the model can jointly consider claim data and carbon emission data, thereby improving the risk prediction accuracy.

[0086] In one embodiment, step 102 may include:

[0087] Performing feature extraction on the historical claim data to obtain at least one historical claim feature;

[0088] Performing feature extraction on the carbon emission reference data to obtain at least one carbon emission reference feature;

[0089] Concatenate at least one historical claim settlement feature and at least one carbon emission reference feature to obtain an augmented sample.

[0090] In one example, if the carbon emission reference object is an enterprise, the historical claim settlement features include enterprise industry type, enterprise scale, historical claim settlement amount, safety measure strategy, etc. The carbon emission reference features include carbon emissions of carbon emissions, achievement of emission reduction targets, enterprise insurance handlers, etc.

[0091] In another embodiment, it is also possible to perform feature filtering on the above-mentioned at least one historical claim settlement feature and at least one carbon emission reference feature to filter out relatively sparse features, and the specific filtering conditions can be set according to requirements.

[0092] In one example, due to the different data distributions of the above features, a solution combining a hybrid discretization method and an adaptive bucketing strategy is proposed. Specifically for the multi-dimensional data of enterprise carbon accounts, by intelligently selecting and combining different discretization strategies (such as equal-width bucketing, logarithmic bucketing, clustering bucketing, etc.), to process the carbon emission reference data with complex distributions. This method can effectively reduce information loss, improve the accuracy of the insurance pricing model, and accelerate the training process. Perform preprocessing and statistical analysis on the carbon emission reference data, extract the distribution type of each feature, classify different features (such as annual carbon emissions, achievement status, enterprise scale, etc.), and identify the discretization strategy suitable for this feature. For features with different degrees of discreteness, select the discretization strategy. For example, for features with a uniform distribution of carbon emissions, use equal-width bucketing; for features with a skewed or logarithmic distribution of carbon emission data, use logarithmic transformation bucketing. For each feature, the system is configured to select the most suitable discretization method according to its data distribution. For multiple features that need to be processed, the system will combine different discretization strategies to achieve hybrid discretization. On the basis of hybrid discretization, the system will also dynamically adjust the bucketing strategy according to the feedback during the training process. For enterprises with large or volatile changes in carbon emissions, the system will automatically refine the bucketing granularity to make the discretization process more precise, so as to better capture the subtle changes in enterprise carbon emissions. For enterprises with relatively stable or low-risk carbon emissions, the system will automatically merge adjacent buckets to reduce waste of computing resources and accelerate the training process.

[0093] In step 103 of some embodiments, the preset initial decision tree model is trained according to the carbon emission sample to obtain a target decision tree model. The initial decision tree model is a model based on the decision tree algorithm, such as the LightGBM model.

[0094] In one example, a carbon emission sample set (including multiple carbon emission samples) can be randomly divided into a training set, a validation set, and a test set in a ratio of 7:2:1. The LightGBM and logistic regression models are trained, and the AUC index of the predicted values of the model on the validation set is used as the objective function of the model, and the optimal model is selected for evaluation. The AUC index of the LightGBM model on the validation set is 0.90, while the AUC index of the logistic regression model on the validation set is only about 0.78. The results show that the effect of using the LightGBM model is better than that of the logistic regression model.

[0095] In one embodiment, referring to Figure 2 , step 103 may include:

[0096] Step 201, perform risk prediction on the carbon emission samples through the initial decision tree model to obtain the sample risk prediction degree; wherein, the sample risk prediction degree indicates the risk degree of the carbon emission reference object having an insurance claim;

[0097] Step 202, adjust the parameters of the initial decision tree model according to the sample risk prediction degree and the risk label value to obtain the target decision tree model;

[0098] Specifically, the carbon emission samples have risk label values. Among them, the risk label value is the first value or the second value. The first value indicates that the first carbon emission object has a claim risk. The second value indicates that the first carbon emission object does not have a claim risk. The first value is determined based on the first claim label value and the influence degree label value. The first claim label value indicates that a claim has been made for the first carbon emission object. The influence degree label value indicates the influence degree of the carbon emission reference data on the first carbon emission object. The second value is determined based on the second claim label value and the influence degree label value. The second claim label value indicates that no claim has been made for the first carbon emission object.

[0099] The benefits of the above embodiments of steps 201 to 202 are that in the process of model training, not only the carbon emission reference data is added to the input data, but also the influence degree label value is added to the label value, which further improves the accuracy of model training and thus improves the accuracy of risk prediction.

[0100] Regarding the influence degree label value, the influence degrees of different carbon emission references on the first carbon emission object are generally different. For example, the carbon emission reference data includes carbon emissions and the completion situation of the emission reduction target. Since carbon emission insurance generally makes claims when carbon emissions exceed the standard, the influence degree label value of the emission reduction target completion situation will be greater than the influence degree label value of carbon emissions at this time. It should be noted that the influence degree label value can be set according to the actual situation.

[0101] In one embodiment, referring to Figure 3, before step 201, the object classification processing method may further include:

[0102] Step 301, obtaining the target carbon emission reduction task that the carbon emission reference object needs to execute;

[0103] Step 302, generating an influence degree label value according to the influence degree of the carbon emission reference data on the execution of the target carbon emission reduction task;

[0104] Step 303, updating the first claim label value of the carbon emission sample according to the influence degree label value to obtain a first value, or updating the second claim label value of the carbon emission sample according to the influence degree label value to obtain a second value.

[0105] In step 301, the target carbon emission reduction task specifically refers to the goals and actions of reducing carbon dioxide and other greenhouse gas emissions. For example, the target carbon emission reduction task includes the following: (1) Taking a certain year as a benchmark (such as 2025), committing to reducing greenhouse gas emissions by a certain percentage (such as 50%) within a certain period (such as by 2030 or 2050). (2) Actively participating in the carbon trading market and offsetting its own emissions by purchasing carbon emission allowances.

[0106] In step 302, the influence degree of the carbon emission reference data on the execution of the target carbon emission reduction task may refer to the contribution degree of the carbon emission data to achieving the target carbon emission reduction task. For example, the lower the carbon emission in the carbon emission reference data, the higher the contribution degree to achieving the target carbon emission reduction task, and the higher the influence degree label value. Also, for example, the higher the carbon trading volume in the carbon emission reference data, the higher the contribution degree to achieving the target carbon emission reduction task, and the higher the influence degree label value. As for the specific mapping relationship, it can be achieved by looking up a table and will not be elaborated here.

[0107] In step 303, if the risk label value of the carbon emission sample is the first claim label value, then update the first claim label value according to the influence degree label value to obtain a first value. If the risk label value of the carbon emission sample is the second claim label value, then update the second claim label value according to the influence degree label value to obtain a second value.

[0108] The benefits of the above steps 301 to 303 are that the risk label values of the carbon emission samples corresponding to different carbon emission reference objects can be flexibly set according to the carbon emission reduction tasks, and the applicability is higher.

[0109] In an embodiment, before step 201, the object classification processing method may further include:

[0110] Construct a first object graph network, where the first object graph network indicates the carbon emission reference data of at least two candidate carbon emission objects respectively and the carbon emission correlation relationship between any two candidate carbon emission objects; among them, the carbon emission reference object belongs to one of the candidate carbon emission reference objects;

[0111] Search for candidate carbon emission objects in the first object graph network according to the carbon emission reference object to extract relevant carbon emission reference data and carbon emission correlation relationships, and obtain the first carbon emission graph embedding feature;

[0112] Add the first carbon emission graph embedding feature to the carbon emission sample to update the carbon emission sample.

[0113] Specifically, in this embodiment, a graph neural network (GNN) is added, and specifically the improved graph attention network GATv2 model is used to construct the first object graph network, where: the object entity is a node, and each node contains a feature vector, such as annual carbon emissions, achievement status, enterprise scale, etc., and the carbon emission correlation relationship between objects is an edge, such as the upstream and downstream relationship between objects, industry collaboration, geographical association and other related association relationships; the embedding features of nodes and edges are extracted through the network to capture the industry-level risk diffusion path.

[0114] The benefit of the above embodiment is that on the basis of considering historical claim data and carbon emission reference data, the carbon emission correlation between different objects is further introduced, so that the model can learn the impact of the carbon emission correlation between different objects on the claim risk, and further improve the accuracy of risk prediction.

[0115] In one embodiment, referring to Figure 4 , before step 201, the object classification processing method may further include:

[0116] Step 401, construct a carbon emission sample set according to at least two carbon emission samples;

[0117] Step 402, calculate the gradient of each sample in the carbon emission sample set to obtain the sample gradient;

[0118] Step 403, sort the samples in the target carbon emission sample set from large to small according to the absolute value of the sample gradient to obtain the sample sorting data;

[0119] Step 404, starting from the first sample in the sample sorting data, extract the first preset number ratio of samples to obtain the first retained samples;

[0120] Step 405, randomly sample the second preset number ratio of samples from the sample sorting data except the first retained samples to obtain the second retained samples; where the second preset number ratio is greater than the first preset number ratio;

[0121] Step 406: Replace the samples in the carbon emission sample set according to the first reserved samples and the second reserved samples to update the carbon emission sample set.

[0122] In step 401, a carbon emission sample is constructed from the historical claim data and carbon emission reference data of each carbon emission reference object, so the carbon emission reference objects corresponding to any two carbon emission samples are different.

[0123] In step 402, the sample gradient refers to the degree of influence of the sample on the parameters of the model. For example, through the initial decision tree model, risk prediction is performed on each sample in the carbon emission sample set to obtain the test risk degree, and the test loss function is calculated based on the test risk degree and the risk label value; according to the test loss function, the sample gradient is determined. The closer the test risk degree is to the risk label value, the smaller the test loss function, and the smaller the sample gradient.

[0124] In step 403, the sample gradients have positive values and negative values, and they are sorted uniformly according to the absolute values of the sample gradients.

[0125] In step 404, the first preset number ratio can be set according to requirements, such as 10%.

[0126] In step 405, the second preset number ratio can be set according to requirements, but it is necessary to ensure that the second preset number ratio is greater than the first preset number ratio. For example, the second preset number ratio is 20%.

[0127] In step 406, the samples in the carbon emission sample set will be replaced by the first reserved samples and the second reserved samples. For example, if the carbon emission sample set originally has 100 samples, the number of the first reserved samples is 10, and the number of the second reserved samples is 20, then the carbon emission sample set is updated to have 30 samples.

[0128] The benefits of the embodiments of the above steps 401 to 406 are that considering that the participation of some carbon emission samples that are not sensitive to risk in model training in the insurance scenario will affect the model performance, sample set preprocessing is performed, which can further improve the model training effect and thus improve the risk prediction accuracy.

[0129] In one embodiment, referring to Figure 5 , step 202 may include:

[0130] Step 501: Use the sample risk prediction degree corresponding to the first reserved sample as the first sample risk prediction degree, and use the risk label value corresponding to the first reserved sample as the first risk label value;

[0131] Step 502: Perform sampling bias learning on the second reserved sample through a preset residual neural network to obtain the sampling compensation weight;

[0132] Step 503: Multiply the sample risk prediction degree corresponding to the second retained sample by the sampling compensation weight to obtain the second sample risk prediction degree, and use the risk label value corresponding to the second retained sample as the second risk label value.

[0133] Step 504: Adjust the parameters of the initial decision tree model according to the first sample predicted risk probability and the first risk label value, as well as the second sample predicted risk probability and the second risk label value, to obtain the target decision tree model.

[0134] Specifically, the residual neural network can be a RESNET network, which is used to perform sampling bias learning on the second retained sample to obtain the sampling compensation weight corresponding to the second retained sample.

[0135] The benefits of the embodiments of the above steps 501 to 504 are that, on the basis of the foregoing update of the carbon emission sample set, considering the possible bias of the second retained sample due to random sampling, and then allocating a sampling compensation weight to the second retained sample, which can further improve the model training effect and thus improve the risk prediction accuracy.

[0136] In step 104 of some embodiments, obtain the insured basic data and the current carbon emission data of the carbon emission prediction object. Among them, the carbon emission prediction object belongs to or does not belong to one of the carbon emission reference objects. The insured basic data refers to various information collected and recorded during the processing of insurance applications. It covers the entire process from the application for insurance to the current time point (i.e., it may not have been successfully insured or there has been no claim yet). The insured basic data includes the insurance application time, the applicant, the submitted insurance application materials (enterprise scale, industry type, etc.), the insurance type, the insurance period, etc. The current carbon emission data refers to the carbon emission data generated by the carbon emission prediction object within a certain period of time. For example, the certain period of time is from the application for insurance to the current time point.

[0137] In step 105 of some embodiments, perform risk prediction on the insured basic data and the current carbon emission data through the target decision tree model to obtain the target risk prediction degree. Among them, the target risk prediction degree indicates the degree of risk of insurance claims for the carbon emission prediction object.

[0138] In one embodiment, step 105 may include:

[0139] Extract features from the insured basic data to obtain at least one insured basic feature;

[0140] Extract features from the current carbon emission data to obtain at least one current carbon emission feature;

[0141] Concatenate at least one insured basic feature and at least one current carbon emission feature to obtain a target risk basic feature;

[0142] Perform risk prediction on the target risk basic feature through the target decision tree model to obtain the target risk prediction degree.

[0143] In an example, if the carbon emission reference object is an enterprise, the insured basic features include the enterprise industry type, enterprise scale, insurance type, insurance scope, safety measure strategy, etc. The current carbon emission features include the carbon emission amount of carbon emissions, the achievement of emission reduction targets, the enterprise insurance handler, etc.

[0144] The benefit of the above embodiment is that by performing risk prediction through the target decision tree model, the accuracy of risk prediction is improved, which helps to improve the accuracy of object classification.

[0145] In one embodiment, step 105 may include:

[0146] Extract carbon emission indicators and carbon emission correlation relationships related to the carbon emission prediction object from the preset second object graph network to obtain the second carbon emission graph embedding feature; wherein, the second object graph network indicates the carbon emission indicators of at least two candidate carbon emission objects and the carbon emission correlation relationships between any two candidate carbon emission objects; wherein, the carbon emission prediction object belongs to one of the candidate carbon emission reference objects;

[0147] Extract features from the current carbon emission data to obtain the current carbon emission feature;

[0148] Extract features from the insured basic data to obtain the insured basic feature;

[0149] Generate a risk prediction basic feature according to the second carbon emission graph embedding feature, the current carbon emission feature, and the insured basic feature;

[0150] Perform risk prediction on the risk prediction basic feature through the target decision tree model to obtain the target risk prediction degree.

[0151] The benefit of this embodiment is that on the basis of the insured basic feature and the current carbon emission feature, the second carbon emission graph embedding feature is further introduced, increasing the model's attention to carbon emission-related data and further improving the accuracy of risk prediction.

[0152] Step 106, classify the carbon emission prediction object according to the target risk prediction degree. For example, if the target risk prediction degree is greater than or equal to the preset risk threshold, classify the carbon emission prediction object as an object not suitable for insurance; if the target risk prediction degree is less than the risk threshold, classify the carbon emission prediction object as an object suitable for insurance.

[0153] In one example, after classifying the carbon emission prediction object as an object not suitable for insurance, the insurance rate is determined according to the target risk prediction level for formulating insurance pricing.

[0154] In one embodiment, to ensure the accuracy and timeliness of the model, a real-time feedback mechanism is provided, which can dynamically optimize the target decision tree model according to the latest enterprise carbon emission data or carbon credit records. When the enterprise's carbon emission data is updated or the carbon trading record changes, the risk prediction level and insurance pricing can be automatically recalculated. With the accumulation of historical data, the binning granularity and the parameters of the pricing model can be automatically adjusted according to the long-term carbon emission performance of the enterprise;

[0155] In summary, at the enterprise insurance stage, the target decision tree model is used to predict the risk of the enterprise. According to the predicted risk prediction level, the relative rate LRL can be deduced, which can help the underwriter judge the claim risk of the enterprise.

[0156] Combining the above embodiments, the present application has at least one of the following beneficial effects: 1. Encourage enterprises to actively participate in carbon emission reduction: By directly linking the enterprise's carbon account data with insurance pricing, a clear motivation for emission reduction is provided for enterprises. In order to achieve a lower insurance rate, enterprises will actively take more measures to reduce carbon emissions, which helps the sustainable development of the enterprises themselves. 2. Reduce the economic risk of enterprises: The coverage of the insurance plan includes the risk of carbon emission overrun and carbon trading losses caused by various unforeseen factors. This provides important risk protection for enterprises and reduces potential economic losses caused by environmental or market changes. 3. Promote financial institutions to participate in the construction of carbon accounts: Promote the attention and support of financial institutions for carbon accounts and carbon markets. 4. Improve the information security and fairness of carbon accounts: The requirements for the information security and fairness of carbon accounts are further improved.

[0157] Please refer to Figure 6 , the embodiment of the present application also provides an object classification processing device, which can implement the above object classification processing method, Figure 6It is a block diagram of the module structure of the object classification processing device provided by the embodiment of the present application. The device includes: a first data acquisition module 601, configured to acquire the carbon emission reference data of the carbon emission reference object; a sample expansion module 602, configured to perform sample expansion according to the historical claim settlement data and the carbon emission reference data of the carbon emission reference object to obtain carbon emission samples; a model training module 603, configured to perform model training on a preset initial decision tree model according to the carbon emission samples to obtain a target decision tree model; a second data acquisition module 604, configured to acquire the insured basic data and the current carbon emission data of the carbon emission prediction object; wherein, the carbon emission prediction object belongs to or does not belong to one of the carbon emission reference objects; a risk prediction module 605, configured to perform risk prediction on the insured basic data and the current carbon emission data through the target decision tree model to obtain a target risk prediction degree; wherein, the target risk prediction degree indicates the risk degree of insurance claim settlement of the carbon emission prediction object; an object classification module 606, configured to classify the carbon emission prediction object according to the target risk prediction degree.

[0158] In one embodiment, the model training module 603 is further configured to: acquire the target carbon emission reduction task to be executed by the carbon emission reference object; generate an influence degree label value according to the influence degree of the carbon emission reference data on the execution of the target carbon emission reduction task; update the first claim settlement label value of the carbon emission sample according to the influence degree label value to obtain a first value, or update the second claim settlement label value of the carbon emission sample according to the influence degree label value to obtain a second value.

[0159] In one embodiment, the model training module 603 is further configured to: construct a carbon emission sample set according to at least two carbon emission samples; perform gradient calculation on each sample in the carbon emission sample set to obtain a sample gradient; sort the samples in the target carbon emission sample set from largest to smallest according to the absolute value of the sample gradient to obtain sample sorting data; starting from the first sample in the sample sorting data, extract the first preset number ratio of samples to obtain first retained samples; randomly sample the second preset number ratio of samples from the sample sorting data except the first retained samples to obtain second retained samples; wherein, the second preset number ratio is greater than the first preset number ratio; replace the samples in the carbon emission sample set according to the first retained samples and the second retained samples to update the carbon emission sample set.

[0160] In one embodiment, before performing risk prediction on the carbon emission samples through the initial decision tree model to obtain the sample risk prediction degree, the device further includes a sample update module, configured to: construct a first object graph network; search for candidate carbon emission objects in the first object graph network according to the carbon emission reference object to extract relevant carbon emission reference data and carbon emission association relationships to obtain a first carbon emission graph embedding feature; add the first carbon emission graph embedding feature to the carbon emission samples to update the carbon emission samples.

[0161] It should be noted that the specific implementation of the object classification processing device is basically the same as the specific embodiments of the above object classification processing method, and will not be elaborated here.

[0162] The embodiments of the present application also provide an electronic device, which includes: a memory, a processor, a program stored on the memory and executable on the processor, and a data bus for realizing the connection and communication between the processor and the memory. When the program is executed by the processor, the above object classification processing method is implemented. The electronic device can be any intelligent terminal including a tablet computer, a vehicle-mounted computer, etc.

[0163] Please refer to Figure 7 , Figure 7 which shows the hardware structure of an electronic device in another embodiment. The electronic device includes:

[0164] A processor 701, which can be implemented in ways such as a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided by the embodiments of the present application;

[0165] A memory 702, which can be implemented in forms such as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 702 can store an operating system and other application programs. When implementing the technical solutions provided by the embodiments of this specification through software or firmware, the relevant program codes are stored in the memory 702, and the processor 701 is called to execute the object classification processing method of the embodiments of the present application;

[0166] An input / output interface 703, which is used to implement information input and output;

[0167] A communication interface 704, which is used to implement communication interaction between this device and other devices, and can implement communication through a wired method (such as USB, network cable, etc.) or through a wireless method (such as mobile network, WIFI, Bluetooth, etc.);

[0168] A bus 705, which transmits information between the various components of the device (such as the processor 701, the memory 702, the input / output interface 703, and the communication interface 704);

[0169] Among them, the processor 701, the memory 702, the input / output interface 703, and the communication interface 704 are communicatively connected to each other inside the device through the bus 705.

[0170] The embodiments of the present application also provide a storage medium. The storage medium is a computer-readable storage medium for computer-readable storage. The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the above object classification processing method.

[0171] As a non-transitory computer-readable storage medium, the memory can be used to store non-transitory software programs and non-transitory computer-executable programs. In addition, the memory can include high-speed random access memory, and can also include non-transitory memory, such as at least one disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories can be connected to the processor through a network. Examples of the above network include but are not limited to the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0172] The embodiments described in the embodiments of the present application are for more clearly illustrating the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. Those skilled in the art will know that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of the present application are equally applicable to similar technical problems.

[0173] Those skilled in the art can understand that the technical solutions shown in the figures do not limit the embodiments of the present application, and may include more or fewer steps than those shown in the figures, or combine some steps, or different steps.

[0174] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0175] Those of ordinary skill in the art can understand that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices can be implemented as software, firmware, hardware, and appropriate combinations thereof.

[0176] In the description of the present application and the above-mentioned accompanying drawings, the terms "first", "second", "third", "fourth", etc. (if any) are used to distinguish similar objects and do not necessarily describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances so that the embodiments of the present application described herein can be implemented in an order different from those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0177] It should be understood that in the present application, "at least one (item)" means one or more, and "a plurality" means two or more. "And / or" is used to describe the association relationship of associated objects and indicates that three relationships can exist. For example, "A and / or B" can mean: only A exists, only B exists, and both A and B exist at the same time. Among them, A and B can be singular or plural. The character " / " generally means that the associated objects before and after are in an "or" relationship. "At least one (one) of the following" or its similar expression means any combination of these items, including any combination of single items (one) or plural items (ones). For example, at least one (one) of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0178] In several embodiments provided by the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection to each other can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be in electrical, mechanical or other forms.

[0179] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0180] In addition, in each embodiment of the present application, each functional unit may be integrated into one processing unit, may exist separately as individual physical units, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of a software functional unit.

[0181] If the above-mentioned integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it may be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, may be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions for causing an electronic device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store programs such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical discs.

[0182] The preferred embodiments of the embodiments of the present application have been described above with reference to the accompanying drawings. However, this does not limit the scope of the rights of the embodiments of the present application. Any modifications, equivalent replacements, and improvements made by those skilled in the art without departing from the scope and essence of the embodiments of the present application shall fall within the scope of the rights of the embodiments of the present application.

Claims

1. An object classification processing method, characterized in that: include: Obtain carbon emission reference data of carbon emission reference objects; Expanding the sample according to the historical claims data of the carbon emission reference object and the carbon emission reference data to obtain a carbon emission sample; Performing model training on a preset initial decision tree model according to the carbon emission sample to obtain a target decision tree model; Obtaining insurance basic data and current carbon emission data of a carbon emission forecast object; wherein the carbon emission forecast object belongs to or does not belong to one of the carbon emission reference objects; Perform risk prediction on the insured basic data and the current carbon emission data through the target decision tree model to obtain a target risk prediction degree; wherein the target risk prediction degree indicates the risk degree of insurance claims for the carbon emission prediction object; The carbon emission prediction objects are classified according to the target risk prediction degree.

2. The method according to claim 1, characterized in that The carbon emission sample has a risk label value; The step of training a preset initial decision tree model according to the carbon emission sample to obtain a target decision tree model includes: Performing risk prediction on the carbon emission sample through the initial decision tree model to obtain a sample risk prediction degree; wherein the sample risk prediction degree indicates the risk degree of insurance claims for the carbon emission reference object; Adjusting the parameters of the initial decision tree model according to the sample risk prediction degree and the risk label value to obtain the target decision tree model; Among them, the risk label value is a first value or a second value, the first value indicates that the carbon emission reference object has a claim risk, and the second value indicates that the carbon emission reference object does not have a claim risk, the first value is determined based on the first claim label value and the impact label value, the first claim label value indicates that a claim has been made for the carbon emission reference object, and the impact label value indicates the degree of influence of the carbon emission reference data on the carbon emission reference object, and the second value is determined based on the second claim label value and the impact label value, and the second claim label value indicates that no claim has been made for the carbon emission reference object.

3. The method according to claim 2, characterized in that Before training a preset initial decision tree model according to the carbon emission sample to obtain a target decision tree model, the method further includes: Obtain the target carbon emission reduction task that the carbon emission reference object needs to perform; Generate the impact label value according to the impact of the carbon emission reference data on the execution of the target carbon emission reduction task; The first claim label value of the carbon emission sample is updated according to the impact label value to obtain the first value, or the second claim label value of the carbon emission sample is updated according to the impact label value to obtain the second value.

4. The method according to claim 2, characterized in that: Before training a preset initial decision tree model according to the carbon emission sample to obtain a target decision tree model, the method further includes: A carbon emission sample set is constructed based on at least two carbon emission samples; Performing gradient calculation on each sample in the carbon emission sample set to obtain a sample gradient; Sort the samples in the carbon emission sample set from large to small according to the absolute value of the sample gradient to obtain sample sorting data; Starting from the first sample in the sample sorting data, extracting a first preset number of samples to obtain a first reserved sample; Randomly sampling a second preset number ratio of samples from the sample sorting data except the first reserved sample to obtain a second reserved sample; wherein the second preset number ratio is greater than the first preset number ratio; The samples in the carbon emission sample set are replaced according to the first reserved samples and the second reserved samples to update the carbon emission sample set.

5. The method according to claim 4, characterized in that The step of adjusting the parameters of the initial decision tree model according to the sample risk prediction degree and the risk label value to obtain the target decision tree model includes: Using the sample risk prediction degree corresponding to the first reserved sample as the first sample risk prediction degree, and using the risk label value corresponding to the first reserved sample as the first risk label value; Performing sampling bias learning on the second retained sample through a preset residual neural network to obtain a sampling compensation weight; Multiplying the sample risk prediction degree corresponding to the second reserved sample by the sampling compensation weight to obtain a second sample risk prediction degree, and using the risk label value corresponding to the second reserved sample as a second risk label value; According to the first sample risk prediction degree and the first risk label value, and the second sample risk prediction degree and the second risk label value, the parameters of the initial decision tree model are adjusted to obtain the target decision tree model.

6. The method according to claim 2, characterized in that Before performing risk prediction on the carbon emission sample by using the initial decision tree model to obtain the sample risk prediction degree, the method further includes: Constructing a first object graph network, the first object graph network indicating carbon emission reference data of each of at least two candidate carbon emission objects and a carbon emission association relationship between any two of the candidate carbon emission objects; wherein the carbon emission reference object belongs to one of the candidate carbon emission reference objects; Searching for candidate carbon emission objects in the first object graph network according to the carbon emission reference object to extract relevant carbon emission reference data and the carbon emission association relationship to obtain a first carbon emission graph embedding feature; The first carbon emission map embedding feature is added to the carbon emission sample to update the carbon emission sample.

7. The method according to any one of claims 1 to 6, characterized in that: The step of performing risk prediction on the insurance basic data and the current carbon emission data through the target decision tree model to obtain a target risk prediction degree includes: Extracting carbon emission indicators and carbon emission associations related to the carbon emission prediction object from a preset second object graph network to obtain a second carbon emission graph embedding feature; wherein the second object graph network indicates the carbon emission indicators of at least two candidate carbon emission objects and the carbon emission association between any two of the candidate carbon emission objects; wherein the carbon emission prediction object belongs to one of the candidate carbon emission reference objects; Extracting features from the current carbon emission data to obtain current carbon emission features; Extracting features from the basic insurance data to obtain basic insurance features; Generate a risk prediction basic feature according to the second carbon emission graph embedded feature, the current carbon emission feature and the insurance basic feature; The target decision tree model is used to perform risk prediction on the risk prediction basic features to obtain the target risk prediction degree.

8. An object classification processing device, characterized in that: include: A first data acquisition module is used to acquire carbon emission reference data of a carbon emission reference object; A sample expansion module, used to expand the sample according to the historical claims data of the carbon emission reference object and the carbon emission reference data to obtain a carbon emission sample; A model training module, used to perform model training on a preset initial decision tree model according to the carbon emission sample to obtain a target decision tree model; A second data acquisition module is used to acquire the insurance basic data and current carbon emission data of the carbon emission prediction object; wherein the carbon emission prediction object belongs to or does not belong to one of the carbon emission reference objects; A risk prediction module, used to perform risk prediction on the insurance basic data and the current carbon emission data through the target decision tree model to obtain a target risk prediction degree; wherein the target risk prediction degree indicates the risk degree of insurance claims for the carbon emission prediction object; The object classification module is used to classify the carbon emission prediction objects according to the target risk prediction degree.

9. An electronic device, characterized in that: The electronic device comprises a memory and a processor, the memory stores a computer program, and the processor implements the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images