Business category prediction method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention provides a service category prediction method and device, electronic equipment and a storage medium, and belongs to the field of service category prediction and the field of financial science and technology. The method comprises the following steps: acquiring a service sample with a service category label; classifying the service samples to obtain service training samples and service verification samples; generating a training association graph of service attribute features based on the service feature data of the service training samples and the service category labels; generating a verification association graph of the service attribute features based on the service feature data of the service verification sample and the service category tag; detecting feature relevancy of the business attribute features based on the training association graph and the verification association graph so as to screen out target business features from the business attribute features; and training a service category prediction model based on the target service features, and performing service category prediction according to the trained model. According to the embodiment of the invention, the screening accuracy of the service attribute characteristics can be improved, so that the prediction accuracy of the model on the service types is improved.

Description

Technical Field

[0001] The present application relates to the fields of business category prediction and financial technology, and in particular to a business category prediction method and device, electronic device, and storage medium. Background Art

[0002] With the advancement of science and technology, acquiring data and using data modeling to solve corresponding problems has become a common technical approach. For example, insurance companies' claims platforms can collect business data from historical insurance claims requests and build claims prediction models based on this data. This trained model can then be used to predict claims for data uploaded by subsequent claim requesters, thereby determining whether the claim requester can access claim resources.

[0003] At present, in the claims business scenario, the claims business data contains a large number of business attribute features (such as the object attributes of the claim requester, the insured attributes, etc.), but some business attribute features have no effect on the training of the claims prediction model, so the business attribute features need to be screened. However, the process of filtering attribute features in business data by related technologies is relatively complicated and the screening accuracy is insufficient, thereby reducing the accuracy of the prediction of business categories. Therefore, how to improve the accuracy of screening business attribute features to improve the accuracy of predicting business categories in business scenarios has become a technical problem that needs to be solved urgently. Summary of the Invention

[0004] The main purpose of the embodiments of the present application is to propose a business category prediction method and device, electronic device, and storage medium, aiming to improve the accuracy of screening business attribute features to improve the accuracy of predicting business categories in business scenarios.

[0005] To achieve the above objectives, a first aspect of an embodiment of the present application provides a service category prediction method, the method comprising:

[0006] Acquire a business sample with a business category label in a target business scenario, wherein the business sample includes a business attribute feature and business feature data of the business attribute feature, and the business category label is used to characterize the category into which the business sample is classified in the target business scenario;

[0007] Classifying the service samples to obtain service training samples and service verification samples;

[0008] Generating a training association graph of the business attribute features based on the business feature data and business category labels of the business training samples;

[0009] Generating a verification association graph of the business attribute features based on the business feature data and business category labels of the business verification sample;

[0010] Performing correlation detection based on the training correlation graph and the verification correlation graph to obtain feature correlation of the business attribute feature;

[0011] Filtering target business features from the business attribute features based on the feature relevance;

[0012] The business category prediction model of the target business scenario is trained based on the target business characteristics, and business category prediction is performed according to the trained business category prediction model.

[0013] In some embodiments, performing correlation detection based on the training correlation graph and the verification correlation graph to obtain the feature correlation of the business attribute feature includes:

[0014] Acquire training sample points from the training association graph; wherein the training sample points include a first horizontal coordinate and a first vertical coordinate, and the verification association graph and the verification association graph corresponding to the same business attribute feature are constructed using the same coordinate system;

[0015] Acquire a verification sample point from the verification association graph, wherein the verification sample point includes a second abscissa and a second ordinate;

[0016] If the first horizontal coordinate and the second horizontal coordinate are the same, performing correlation calculation on the first vertical coordinate and the second vertical coordinate to obtain coordinate correlation;

[0017] The feature correlation of the business attribute feature is determined according to the correlation of at least two coordinates of the business attribute feature.

[0018] In some embodiments, determining the feature correlation of the business attribute feature according to the correlation of at least two coordinates of the business attribute feature includes:

[0019] Performing graph fluctuation correlation calculation on the training association graph and the verification association graph to obtain graph fluctuation correlation;

[0020] The feature correlation of the business attribute feature is determined according to the graph fluctuation correlation of the business attribute feature and at least two of the coordinate correlations.

[0021] In some embodiments, performing graph fluctuation correlation calculation on the training association graph and the verification association graph to obtain the graph fluctuation correlation includes:

[0022] extracting the number of training fluctuation reversals and the number of training fluctuation peak points from the training correlation graph;

[0023] Extracting the number of verified fluctuation reversals and the number of verified fluctuation peak points from the verified correlation graph;

[0024] Calculating the difference between the training fluctuation reversal number and the verification fluctuation reversal number to obtain a fluctuation reversal difference;

[0025] Calculating the difference between the number of training fluctuation peak points and the number of verification fluctuation peak points to obtain a fluctuation peak point difference;

[0026] The graph fluctuation correlation is calculated based on the fluctuation reversal difference and the fluctuation peak point difference to obtain the graph fluctuation correlation.

[0027] In some embodiments, generating the training association graph of the service attribute features based on the service feature data and service category labels of the service training samples includes:

[0028] If the business attribute feature is a continuous feature, performing feature value conversion on the business feature data of the business training sample according to a preset feature value conversion rule to obtain converted feature data;

[0029] A training association graph of the business attribute features is generated based on the converted feature data and the business category label.

[0030] In some embodiments, generating the training association graph of the service attribute features based on the converted feature data and the service category label includes:

[0031] quantizing the service category labels of the service training samples to obtain label quantization values;

[0032] Calculating the mean of all the label quantization values corresponding to the same converted feature data to obtain the label mean of the converted feature data;

[0033] Constructing training sample points based on the converted feature data and label mean;

[0034] A training association graph of the business attribute features is generated based on all the training sample points.

[0035] In some embodiments, the training of the business category prediction model of the target business scenario based on the target business characteristics includes:

[0036] Updating the service sample based on the target service feature to obtain a target sample;

[0037] A business category prediction model for the target business scenario is trained based on the target samples.

[0038] To achieve the above-mentioned purpose, a second aspect of an embodiment of the present application provides a service category prediction device, the device comprising:

[0039] An acquisition module is configured to acquire a business sample with a business category label in a target business scenario, wherein the business sample includes a business attribute feature and business feature data of the business attribute feature, and the business category label is used to characterize the category into which the business sample is classified in the target business scenario;

[0040] A classification module, configured to classify the service samples to obtain service training samples and service verification samples;

[0041] A first graph generating module, configured to generate a training association graph of the business attribute features based on the business feature data and business category labels of the business training samples;

[0042] A second graph generating module is configured to generate a verification association graph of the business attribute features based on the business feature data and business category labels of the business verification sample;

[0043] A detection module, configured to perform correlation detection based on the training correlation graph and the verification correlation graph to obtain a feature correlation of the business attribute feature;

[0044] A screening module, configured to screen target business features from the business attribute features based on the feature relevance;

[0045] A training module is used to train a business category prediction model of the target business scenario based on the target business characteristics, and perform business category prediction based on the trained business category prediction model.

[0046] To achieve the above-mentioned purpose, the third aspect of an embodiment of the present application proposes an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the method described in the first aspect when executing the computer program.

[0047] To achieve the above-mentioned purpose, the fourth aspect of the embodiments of the present application proposes a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the method described in the first aspect.

[0048] The present application proposes a business category prediction method and device, electronic device, and storage medium. First, a business sample with a business category label in a target business scenario is obtained, where the business sample includes a business attribute feature and business feature data of the business attribute feature, and the business category label is used to characterize the category into which the business sample is divided in the target business scenario; then, the business sample is classified to obtain a business training sample and a business verification sample; further, a training association graph of the business attribute feature is generated based on the business feature data and business category label of the business training sample; further, a verification association graph of the business attribute feature is generated based on the business feature data and business category label of the business verification sample; then, a correlation test is performed based on the training association graph and the verification association graph to obtain the feature correlation of the business attribute feature; based on the feature correlation, a target business feature is screened out from the business attribute feature; finally, a business category prediction model for the target business scenario is trained based on the target business feature, and business category prediction is performed based on the trained business category prediction model. The embodiment of the present application constructs a training association graph and a verification association graph corresponding to the business attribute features respectively. The constructed association graph can better reflect the overall association of the business attribute features between multiple samples and deepen the understanding of the features. In this way, the screening of attribute features based on the feature correlation determined by the training association graph and the verification association graph can improve the accuracy of the screening of attribute features, thereby improving the prediction accuracy of the business category by the business category prediction model trained based on the screened attribute features. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 This is a flow chart of the service category prediction method provided by an embodiment of the present application;

[0050] Figure 2 yes Figure 1 A flowchart of step S130 in FIG.

[0051] Figure 3 yes Figure 2 A flowchart of step S220 in FIG.

[0052] Figure 4 This is a schematic diagram of a training association diagram provided in an embodiment of the present application;

[0053] Figure 5 yes Figure 1 A flowchart of step S150 in FIG.

[0054] Figure 6 yes Figure 5 A flowchart of step S540 in FIG.

[0055] Figure 7 yes Figure 6 A flowchart of step S610 in FIG.

[0056] Figure 8 This is a flow chart of screening business attribute features provided by an embodiment of the present application;

[0057] Figure 9 This is a flow chart for training a business category prediction model provided by an embodiment of the present application;

[0058] Figure 10 This is a structural diagram of a service category prediction device provided by an embodiment of the present application;

[0059] Figure 11 This is a hardware structure diagram of the electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

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

[0061] It should be noted that although the device schematics illustrate functional module divisions and the flowcharts illustrate logical sequences, in certain circumstances, the steps shown or described may be performed in a sequence that differs from the module divisions in the device or the sequence in the flowcharts. The terms "first," "second," and so on, in the specification, claims, and drawings, are used to distinguish similar items and are not necessarily used to describe a specific sequence or precedence.

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

[0063] First, let’s analyze some of the terms used in this application:

[0064] Artificial intelligence (AI) is a new technical discipline that studies and develops theories, methods, technologies, and application systems for simulating, extending, and expanding human intelligence. A branch of computer science, AI seeks to understand the essence of intelligence and produce new intelligent machines that can respond in a manner similar to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing, and expert systems. AI can simulate the information processes of human consciousness and thinking. It also encompasses the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, to perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results.

[0065] Claim resources refer to the amount of resources paid by a resource payer (such as an insurance company or data provider) in response to a claim request. For example, in the insurance sector of FinTech, claim resources can be the premiums received by a claim requester. Because each claim requester's choice of claim items and individual circumstances vary, the specific claim information used in calculating claim resources, such as deductibles, payout ratios, and payout limits, varies.

[0066] With the advancement of science and technology, acquiring data and using data modeling to solve corresponding problems has become a common technical approach. For example, insurance companies' claims platforms can collect business data from historical insurance claims requests and build claims prediction models based on this data. This trained model can then be used to predict claims for data uploaded by subsequent claim requesters, thereby determining whether the claim requester can access claim resources.

[0067] Currently, in claims processing scenarios, for example, claims data contains numerous business attribute features (such as the claimant's attributes and policy attributes). However, some of these attributes are not useful for training claims prediction models, necessitating filtering. Feature engineering is a key aspect of machine learning. Currently, commonly used feature filtering methods include filtering, wrapping, and embedding. Of these three methods, filtering is simple to implement, the code is easy to understand, and it can quickly select commonly used features, but it cannot handle data with cyclical or trending changes. Wrapping and embedding have complex and difficult-to-understand code, and are more suited to large datasets and high-dimensional features, resulting in lower applicability and efficiency. Therefore, the filtering process for attribute features in business data is complex and inaccurate, resulting in reduced prediction accuracy for business categories.

[0068] Based on this, the embodiments of the present application provide a business category prediction method and device, an electronic device, and a storage medium, which can improve the accuracy of screening business attribute features to improve the accuracy of predicting business categories.

[0069] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology, and application system that 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 achieve optimal results.

[0070] Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interaction systems, and mechatronics. AI software technologies primarily encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0071] The business category prediction method provided in the embodiment of the present application relates to the field of artificial intelligence technology. The business category prediction method provided in the embodiment of the present application can be applied to a terminal, can be applied to a server side, or can be software running in a terminal or a server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or as a server cluster or distributed system composed of multiple physical servers, or can be configured as a cloud server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms and other basic cloud computing services; the software can be an application that implements the business category prediction method, etc., but is not limited to the above forms.

[0072] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and the like. 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, and the like that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments in which tasks are performed by remote processing devices connected via a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media, including storage devices.

[0073] It should be noted that in each specific embodiment of the present application, when it comes to the need to perform relevant processing based on data related to the identity or characteristics of the object, such as the object's insurance policy information, the object's claims data, and the object's attribute characteristics, the object's permission or consent will be obtained first, and the collection, use, and processing of such data will comply with relevant laws, regulations, and standards. In addition, when the embodiment of the present application needs to obtain the sensitive personal information of the object, the object's separate permission or consent will be obtained through a pop-up window or by jumping to a confirmation page. After clearly obtaining the object's separate permission or consent, the necessary object-related data for the normal operation of the embodiment of the present application will be obtained.

[0074] See also Figure 1 , Figure 1 This is an optional flow chart of the business category prediction method provided by the embodiment of the present application. In some embodiments, Figure 1 The method may include but is not limited to steps S110 to S170:

[0075] Step S110, obtaining a business sample with a business category label in the target business scenario;

[0076] Step S120, classifying the service samples to obtain service training samples and service verification samples;

[0077] Step S130, generating a training association graph of business attribute features based on the business feature data and business category labels of the business training samples;

[0078] Step S140, generating a verification association graph of business attribute features based on the business feature data and business category labels of the business verification sample;

[0079] Step S150, performing correlation detection based on the training correlation graph and the verification correlation graph to obtain feature correlation of the business attribute features;

[0080] Step S160, filtering out target business features from the business attribute features based on feature relevance;

[0081] Step S170 , training a business category prediction model for a target business scenario based on target business characteristics, and performing business category prediction based on the trained business category prediction model.

[0082] In step S110 of some embodiments, the present application may first obtain a business sample set, which includes multiple business samples. The business sample set is a sample set constructed based on different expected target business scenarios. The target business scenario corresponds to the business category prediction model. For example, when the target business scenario is a claim category prediction, that is, the claim category prediction model (equivalent to the business category prediction model) is used to predict whether a claim will be made to the claim object. The business category at this time may include successful claims and failed claims. For example, when the target business scenario is an object risk category prediction, that is, the risk category of the applicant is predicted through a risk category prediction model (equivalent to the business category prediction model). The business category at this time indicates the risk level, including level one, level two, level three, etc., and the higher the risk level, the lower the probability of repayment of the applicant. In the following embodiments, the present application will take the claim category prediction scenario as an example for explanation, and the business samples, business attribute features, category labels, business feature data, etc. can be flexibly replaced according to the transformation of the business scenario.

[0083] Business samples include business attribute features and business feature data of business attribute features. For example, in the claim category prediction scenario, each business sample at this time can be used to characterize the object claim information of the sample claim object, which can be called a claim sample. For each claim sample, the object claim information may include the claim application documents entered or uploaded by the sample claim object during the claim application stage, such as the historical insurance text of the sample claim object, the claim certificate text (such as hospital hospitalization settlement form, diagnosis text, medical examination text, etc.). Corresponding to the risk category prediction scenario, the corresponding business sample may include object attribute information, risk assessment report, etc., without limitation.

[0084] Business attribute features are used to characterize attribute features extracted from business samples. For example, attribute features extracted from the historical claim application documents of the sample claim subject may include the subject's personal information, number of insurance policies, insurance type, and number of historical claims. Business feature data refers to the specific numerical value corresponding to the business attribute feature. For example, when the business attribute feature is the number of insurance policies, the corresponding business feature data can be positive integer values such as 1 and 2; when the business attribute feature is the insurance type, the corresponding business feature data can be insurance types such as Type A and Type B. Business category labels are used to characterize business samples as successful claims or failed claims.

[0085] In the above embodiment, the present application can provide a supervised learning basis for the subsequent business category prediction model by obtaining business samples with business category labels, ensuring that the model can distinguish between valid and invalid business category prediction requests.

[0086] It should be noted that in order to deeply understand the relevant information for claim category prediction in the object claim information, this application can use a preset text parsing model to parse the claim application documents of the sample claim object, and extract the complete business attribute features and business feature data of the business attribute features, such as 50 business attribute features extracted. In other words, this application can use a pre-trained text parsing model to quickly scan the claim application documents of the sample claim object, and can identify and mark all relevant business attribute features. Among them, the text parsing model can be established based on historical data and actual cases, and different models can be used, such as machine learning models (such as random forests, support vector machines, etc.), deep learning models, large models, etc., without specific limitation.

[0087] In step S120 of some embodiments, in order to accurately determine the degree of influence of each business attribute feature on the claim category prediction, the present application can classify multiple business samples in the business sample set to obtain a training sample set and a verification sample set, where the training sample set includes multiple business training samples and the verification sample set includes multiple business verification samples.

[0088] The present application can partition multiple business samples according to a preset partitioning ratio. For example, if the preset partitioning ratio is 8:2, 80% of the business samples in the business sample set can be randomly selected as business training samples, and the remaining 20% of the business samples can be used as business verification samples. This random partitioning method based on the preset partitioning ratio is simple and efficient, improving the efficiency of model training, preventing model overfitting, and enhancing generalization capabilities.

[0089] In step S130 of some embodiments, the present application may generate a corresponding training association graph and verification association graph for each business attribute feature, thereby identifying the role of the business attribute feature in model training based on the training association graph and the verification association graph. The present application uses association graphs to visualize each business attribute feature, which can deepen understanding of the feature, is simple and easy to understand, and improves the efficiency of feature screening.

[0090] See Figure 2 , Figure 2 This is an optional flowchart of step S130 provided in an embodiment of the present application. In some embodiments of the present application, step S130 may specifically include steps S210 to S220:

[0091] Step S210: If the business attribute feature is a continuous feature, the business feature data of the business training sample is converted into feature values according to a preset feature value conversion rule to obtain converted feature data;

[0092] Step S220 , generating a training association graph of the business attribute features based on the converted feature data and the business category labels.

[0093] In step S210 of some embodiments, the business attribute feature may be a continuous feature, such as the cumulative insured amount, policy duration, and other numerical data that can take continuous values and can reflect the dynamic risk characteristics of the policyholder or the insured. The business attribute feature may also be a discrete feature, such as the disease type, the subject's blood type, and other features that cannot be represented by continuous numerical values and have clear classification or counting. If the business attribute feature is a continuous feature, the present application may discretize the continuous feature, that is, perform feature value conversion on the business feature data of the business training sample according to a preset feature value conversion rule to obtain the converted feature data.

[0094] Preset feature value conversion rules refer to data discretization rules set for different business attribute features. The converted feature data refers to the numerical value obtained by discretizing the business feature data. For example, for the continuous feature "policy duration," the corresponding preset feature value conversion rules are shown in Table 1 below. Based on the business feature data corresponding to the business training sample, the corresponding converted feature data can be obtained by searching Table 1.

[0095] Policy duration (years) Corresponding discrete values [0,0.5] 1 (0.5,1] 2 (1,3] 3 (3,5] 4 (5,+∞] 5

[0096] Table 1

[0097] Among them, the embodiment of the present application can convert the feature values corresponding to continuous features into discrete values through data conversion, which can be flexibly applied to various types of features, such as data with periodic or trend changes, thereby improving the flexibility and applicability of feature screening.

[0098] In step S220 of some embodiments, the training association graph of the present application may refer to a line graph generated based on the business feature data and business category labels of the business training samples, and the horizontal coordinate of the coordinate system corresponding to the training association graph is determined based on the converted feature data corresponding to the business feature data, and the vertical coordinate is determined based on the business category label.

[0099] It should be noted that if the business attribute feature is a discrete feature, the horizontal coordinate of the coordinate system corresponding to the training association graph is determined based on the business feature data, and the vertical coordinate is determined based on the business category label.

[0100] See Figure 3 , Figure 3 This is an optional flowchart of step S220 provided in an embodiment of the present application. In some embodiments of the present application, step S220 may specifically include steps S310 to S340:

[0101] Step S310, quantizing the service category labels of the service training samples to obtain label quantization values;

[0102] Step S320, performing mean calculation on all label quantization values corresponding to the same converted feature data to obtain the label mean of the converted feature data;

[0103] Step S330, constructing training sample points based on the converted feature data and label mean;

[0104] Step S340: Generate a training association graph of business attribute features based on all training sample points.

[0105] In step S310 of some embodiments, when determining the vertical coordinate corresponding to each converted feature data in the training association graph, the business category labels of the business training samples may be quantized to obtain label quantization values. For example, if the business category labels corresponding to the current claim settlement category include "claim approval" and "claim disapproval," the quantization value of the label corresponding to "claim approval" may be set to 1, and the quantization value of the label corresponding to "claim disapproval" may be set to 0, without limitation.

[0106] In step S320 of some embodiments, since the business feature data of multiple business training samples may correspond to different converted feature data, and each business training sample may have a corresponding business category label, such as whether the claim was successful or failed, a mean of the label quantization values corresponding to all business training samples of the same converted feature data may be calculated to obtain a label mean of the converted feature data, and the label mean is used as the vertical coordinate corresponding to the converted feature data in the training association graph.

[0107] For example, Figure 4As shown, it is a schematic diagram of the training association graph provided by the embodiment of the present application. For example, in the claim category prediction scenario, there are 80 business training samples. For the continuous business attribute feature of the policy duration, the business feature data corresponding to the policy duration in each business training sample is converted according to the mapping relationship recorded in Table 1 to obtain the corresponding converted feature data. At this time, there are 30 business training samples in the interval [0,0.5], 30 business training samples in the interval (0.5,1], 20 business training samples in the interval (1,3], 0 business training samples in the interval (3,5], and 0 business training samples in the interval (5,+∞]. Since the business category labels of the current business scenario include successful claims and failed claims, the corresponding vertical coordinate in the training association graph is calculated for each converted feature data, and the label quantization value corresponding to the successful claim is set to 1, and the label quantization value corresponding to the successful claim is set to 0. In this way, for the interval [0,0.5] (that is, the corresponding converted feature data is 1), there are 2 in 30 business training samples. 7 examples are successful claims and 3 are failed claims. The corresponding label mean of the converted feature data is (27×1+3×0)30=0.9. Similarly, for the interval (0.5,1] (i.e., the corresponding converted feature data is 2), 24 of the 30 business training examples are successful claims and 6 are failed claims. The corresponding label mean of the converted feature data is 0.8. For the interval (1,3] (i.e., the corresponding converted feature data is 3), 18 of the 20 business training examples are successful claims and 2 are failed claims. The corresponding label mean of the converted feature data is 0.9. For the intervals (3,5] and (5,+∞], the corresponding label mean is 0.

[0108] In step S330 of some embodiments, after determining the label mean corresponding to each converted feature data, the converted feature data and the corresponding label mean can be used to construct a training sample point, that is, a coordinate point in the association graph, and the coordinate format can be: (converted feature data, label mean). Figure 4 As shown, for the converted feature data 2, the corresponding label mean is 0.8, and the corresponding constructed sample data point is (2, 0.8).

[0109] In step S340 of some embodiments, Figure 4 As shown, all training sample points are connected to generate a training association graph for each business attribute feature.

[0110] In step S140 of some embodiments, the verification association graph of the present application refers to a line graph generated based on the business feature data and business category labels of the business verification sample, and the horizontal coordinate of the coordinate system corresponding to the verification association graph is determined based on the converted feature data corresponding to the business feature data, and the vertical coordinate is determined based on the business category label. The verification association graph and the verification association graph corresponding to the same business attribute feature are constructed using the same coordinate system. Referring to the generation process of the training association graph described above, a verification association graph corresponding to each business attribute feature can be generated.

[0111] In step S150 of some embodiments, to determine the importance of each business attribute feature to model training, a correlation test may be performed on the training association graph and the validation association graph corresponding to each business attribute feature to obtain a feature relevance of the business attribute feature. The feature relevance evaluates the relevance of the business attribute feature to the training of the business category prediction model, and a greater feature relevance indicates a greater importance of the business attribute feature to model training.

[0112] See Figure 5 , Figure 5 This is an optional flowchart of step S150 provided in an embodiment of the present application. In some embodiments of the present application, step S150 may specifically include steps S510 to S540:

[0113] Step S510, obtaining a training sample point from a training association graph; wherein the training sample point includes a first abscissa and a first ordinate;

[0114] Step S520, obtaining a verification sample point from the verification association graph, where the verification sample point includes a second abscissa and a second ordinate;

[0115] Step S530: If the first horizontal coordinate and the second horizontal coordinate are the same, a correlation calculation is performed on the first vertical coordinate and the second vertical coordinate to obtain a coordinate correlation;

[0116] Step S540: determining the feature correlation of the business attribute feature according to the correlation of at least two coordinates of the business attribute feature.

[0117] In step S510 of some embodiments, the training association graph includes multiple training sample points, such as Figure 4 As shown, each training sample point includes a first abscissa and a first ordinate. The first abscissa represents the abscissa data corresponding to the training sample point in the training association graph, and the first ordinate represents the ordinate data corresponding to the training sample point in the training association graph.

[0118] In step S520 of some embodiments, the verification association graph includes a plurality of verification sample points, each of which includes a second abscissa and a second ordinate. The second abscissa represents the abscissa data corresponding to the verification sample point in the verification association graph, and the second ordinate represents the ordinate data corresponding to the verification sample point in the verification association graph.

[0119] In some embodiments, in step S530, a correlation calculation may be further performed on the first and second ordinates corresponding to the same abscissa to obtain the coordinate correlation corresponding to each transformed data point. This can measure the strength of the association between the training sample point and the validation sample point on the same feature dimension (abscissa). The method for calculating the ordinate correlation may be a Pearson coefficient, cosine similarity, or other calculation method, without limitation.

[0120] In step S540 of some embodiments, the contribution of business attribute features to the prediction target (i.e., the category prediction target of the target business scenario, such as predicting a successful or failed claim) is further evaluated by integrating the results of multiple coordinate correlation calculations. Specifically, the final correlation can be determined based on at least two coordinate correlations (e.g., different feature combinations) through a weighted or voting mechanism. This can improve the robustness of feature screening, reduce the risk of misjudgment of a single correlation, and enhance the model's ability to capture complex feature interactions.

[0121] See Figure 6 , Figure 6 This is an optional flowchart of step S540 provided in an embodiment of the present application. In some embodiments of the present application, step S540 may specifically include steps S610 to S620:

[0122] Step S610, performing graph fluctuation correlation calculation on the training association graph and the verification association graph to obtain graph fluctuation correlation;

[0123] Step S620 : determining the feature correlation of the business attribute feature according to the graph fluctuation correlation and at least two coordinate correlations of the business attribute feature.

[0124] In step S610 of some embodiments, the present application may also determine the feature correlation corresponding to each business attribute feature from the perspectives of both coordinate correlation and graph volatility. Graph volatility correlation measures the degree of dynamic difference in feature distribution or structure between the training association graph and the validation association graph, and can reflect the stability of feature correlation.

[0125] See Figure 7 , Figure 7 This is an optional flowchart of step S610 provided in an embodiment of the present application. In some embodiments of the present application, step S610 may specifically include steps S710 to S750:

[0126] Step S710, extracting the number of training fluctuation reversals and the number of training fluctuation peak points from the training association graph;

[0127] Step S720, extracting the number of verified fluctuation reversals and the number of verified fluctuation peak points from the verification correlation graph;

[0128] Step S730, calculating the difference between the number of training fluctuation reversals and the number of verification fluctuation reversals to obtain a fluctuation reversal difference;

[0129] Step S740, performing a difference calculation between the number of training fluctuation peak points and the number of verification fluctuation peak points to obtain a fluctuation peak point difference;

[0130] Step S750, calculating the graph fluctuation correlation based on the fluctuation reversal difference and the fluctuation peak point difference to obtain the graph fluctuation correlation.

[0131] In step S710 of some embodiments, the number of training fluctuation reversals refers to the number of fluctuation reversals that occur in the training correlation graph, that is, the number of trend shifts that occur in the training correlation graph. For example, if the ordinate increases as the horizontal axis moves to the right and then suddenly decreases, this corresponds to a trend shift, or if the ordinate first decreases and then suddenly increases. The number of training fluctuation peak points refers to the number of local maximum points in the training correlation graph (e.g., local peaks of the label means).

[0132] In step S720 of some embodiments, verifying the number of fluctuation reversals refers to verifying the number of fluctuation reversals that appear in the correlation graph, that is, verifying the number of trend transitions that appear in the correlation graph. Verifying the number of fluctuation peak points refers to verifying the number of local maximum points in the correlation graph (e.g., local peaks of the label means).

[0133] In step S730 of some embodiments, the fluctuation reversal difference refers to the absolute difference between the number of fluctuation reversals during training and the number of fluctuation reversals during validation, and can be used to reflect the model's sensitivity to reversal patterns. By considering the fluctuation reversal difference, it is possible to identify differences in the stability of reversal patterns during training and validation, helping to determine the risk of model overfitting.

[0134] In some embodiments, in step S740, the peak-to-peak difference refers to the absolute difference between the number of training peak points and the number of validation peak points, and can be used to reflect the difference in the model's ability to capture local extreme values. By considering the peak-to-peak difference, the consistency of the distribution of local extreme values during training and validation can be identified, improving the accuracy of the model's generalization ability assessment.

[0135] In some embodiments, step S750 can measure the overall fluctuation consistency of the training correlation graph and the validation correlation graph by combining the fluctuation reversal difference and the peak point difference. Specifically, the fluctuation reversal difference and the fluctuation peak point difference can be normalized to the same interval, and corresponding weights can be assigned to the two differences. The graph fluctuation correlation can be obtained by performing a weighted sum calculation on the differences and the corresponding weights.

[0136] In step S620 of some embodiments, the graph fluctuation correlation and at least two coordinate correlations of the business attribute feature can be further standardized to the same interval, and then the standardized graph fluctuation correlation and at least two coordinate correlations can be weighted and calculated to determine the feature correlation of the business attribute feature.

[0137] In step S160 of some embodiments, after obtaining the feature relevance corresponding to each business attribute feature, the feature relevance can be numerically compared with a preset relevance threshold, and business attribute features with a feature relevance greater than or equal to the preset relevance threshold are selected as target business features. Target business features are features selected from business attribute features that have a high degree of relevance to the training of the business category prediction model, which can improve the efficiency and accuracy of model training.

[0138] In other embodiments, the present application may also filter target business features from business attribute features based on feature correlation and fluctuation reversal difference. Specifically, when the feature correlation corresponding to a business attribute feature is greater than or equal to a preset correlation threshold, and the fluctuation reversal difference is less than or equal to a preset reversal difference threshold, the corresponding business attribute feature may be selected as the target business feature. In this case, it can be assumed that the trend transformation of the business attribute feature in the training sample set and the test sample set is consistent, and this feature can be retained.

[0139] like Figure 8 , which is a schematic diagram of the process of screening business attribute features provided in an embodiment of the present application.

[0140] Step S810, classifying multiple business samples in the business sample set to obtain a training sample set and a verification sample set;

[0141] Step S820, determining whether each business attribute feature to be screened in the training sample set and the verification sample set is a continuous feature, if so, executing step S831, if not, executing step S832;

[0142] In step S831 , the continuous features are discretized, the discretized feature values are sorted, and step S840 is executed.

[0143] Among them, step S831 corresponds to step S210. Since each business attribute feature is based on the same preset feature value conversion rule during the discretization process, a process of sorting these converted feature values is implicitly included.

[0144] Step S832: do not process the continuous features, sort the feature values of the continuous features, and execute step S840;

[0145] Step S840, calculating the label mean corresponding to each discrete feature value in each business attribute feature;

[0146] Step S850: generating a training association graph and a verification association graph based on the service feature data and the service category label;

[0147] Step S860 , determining whether the training association graph and the verification association graph are consistent, if so, executing step S871 , if not, executing step S872 .

[0148] The determination of whether the training association graph and the verification association graph are consistent corresponds to step S150 in the above embodiment, that is, calculating the feature correlation of the business attribute features to determine whether they are consistent, which will not be described in detail.

[0149] Step S871, retaining the corresponding business attribute characteristics;

[0150] Step S872: Eliminate the corresponding business attribute features.

[0151] Based on this, this application can be applied to a variety of different data types (such as numerical and categorical types) and application scenarios (that is, it can be applied in all supervised learning), and has wide application value.

[0152] In step S170 of some embodiments, after obtaining multiple target business features, a business category prediction model can be trained based on the target business features. The trained business category prediction model can then be used to predict the business category of the input target subject's application document, obtaining the current business category. For example, if the business category prediction model is a claim category prediction model, the target subject's application document is a claim application document, and the current claim category represents the claim settlement conclusion for the claim application document, i.e., whether the claim is approved or not. For example, if there are 50 original business attribute features, and after screening in the above steps, 32 features are obtained as target business features, a model can be trained based on these 32 target business features, ultimately obtaining a model that automatically predicts whether a claim can be automatically settled. When the target subject needs to make a claim, they can submit a target claim request. At this point, the claims system can respond to the target claim request submitted by the target subject and, based on the target claim request, obtain the target claim data for the target subject. The target claim data refers to claims data related to the target claim request, such as accident details, loss details, hospitalization records, and records of involved claim resources. Furthermore, the claim category prediction model can perform claim prediction on the input target claim data to obtain the claim category of the target claim data.

[0153] See Figure 9 , Figure 9 This is an optional flow chart for training a business category prediction model provided by an embodiment of the present application. In some embodiments of the present application, the step of training a business category prediction model for a target business scenario based on target business features may specifically include steps S910 to S920:

[0154] Step S910, updating the service sample based on the target service characteristics to obtain a target sample;

[0155] Step S920: training a business category prediction model for a target business scenario based on the target sample.

[0156] In steps S910 and S920 of some embodiments, the present application can update the business samples based on the target business characteristics, that is, remove relevant data that are not target business characteristics from the business samples to obtain new target samples, and train the business category prediction model of the target business scenario based on multiple target samples.

[0157] It should be noted that the non-Company's software tools or components that appear in the embodiments of this application are merely examples and do not represent actual use.

[0158] The embodiment of the present application provides a method for predicting a business category. By constructing a training association graph and a verification association graph corresponding to business attribute features respectively, the constructed association graph can better reflect the overall association between business attribute features in multiple samples and deepen the understanding of features. In this way, the screening of attribute features based on the feature correlation determined by the training association graph and the verification association graph can improve the accuracy of the screening of attribute features, thereby improving the accuracy of the prediction of business categories by the business category prediction model trained based on the screened attribute features. For example, the selectivity of features can be improved by judging the trend of feature data and the number of trend changes, and can be applied to a variety of different data types and application scenarios. In addition, the method provided by the embodiment of the present application can improve the accuracy and efficiency of feature screening, thereby improving the performance of the model. Compared with the filtering method used in the related art, it can discover the trend of data; and compared with the embedding method and packaging method used in the related art, it is simpler and easier to understand, and the model trained after feature filtering can achieve a good recall rate and accuracy.

[0159] See also Figure 10 The embodiment of the present application further provides a service category prediction device that can implement the above service category prediction method. The service category prediction device 1000 includes:

[0160] An acquisition module 1010 is configured to acquire a business sample with a business category label in a target business scenario. The business sample includes business attribute features and business feature data of the business attribute features. The business category label is used to characterize the category of the business sample in the target business scenario.

[0161] The classification module 1020 is used to classify the service samples to obtain service training samples and service verification samples;

[0162] A first graph generating module 1030 is configured to generate a training association graph of business attribute features based on business feature data and business category labels of business training samples;

[0163] The second graph generating module 1040 is configured to generate a verification association graph of business attribute features based on the business feature data and business category labels of the business verification samples;

[0164] A detection module 1050 is configured to perform correlation detection based on the training correlation graph and the verification correlation graph to obtain a feature correlation of the business attribute feature;

[0165] A screening module 1060 is configured to screen target business features from business attribute features based on feature relevance;

[0166] The training module 1070 is used to train a business category prediction model of a target business scenario based on target business features, and perform business category prediction based on the trained business category prediction model.

[0167] The specific implementation of the service category prediction device is substantially the same as the specific embodiment of the service category prediction method described above, and will not be described in detail here.

[0168] The present application also provides an electronic device comprising a memory and a processor. The memory stores a computer program, and the processor implements the aforementioned service category prediction method when executing the computer program. The electronic device, which may also be referred to as a computer device, may be any intelligent terminal, including a tablet computer and an in-vehicle computer.

[0169] See also Figure 11 , Figure 11 The hardware structure of an electronic device according to another embodiment is shown. The electronic device includes:

[0170] The processor 1110 may be implemented as a general-purpose central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is configured to execute relevant programs to implement the technical solutions provided in the embodiments of the present application.

[0171] The memory 1120 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 1120 can store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 1120 and is called by the processor 1110 to execute the service category prediction method of the embodiments of this application.

[0172] Input / output interface 1130, used for information input and output;

[0173] Communication interface 1140, used to implement communication interaction between this device and other devices, which can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WiFi, Bluetooth, etc.);

[0174] bus 1150 , which transmits information between various components of the device (e.g., processor 1110 , memory 1120 , input / output interface 1130 , and communication interface 1140 );

[0175] The processor 1110 , the memory 1120 , the input / output interface 1130 , and the communication interface 1140 are communicatively connected to each other within the device via a bus 1150 .

[0176] An embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned business category prediction method is implemented.

[0177] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0178] The embodiments described in the embodiments of this application are intended to more clearly illustrate the technical solutions of the embodiments of this application and do not constitute a limitation on the technical solutions provided by the embodiments of this application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0179] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0180] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, i.e., they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of this embodiment.

[0181] Those skilled in the art will appreciate that all or some of the steps in the methods, systems, and functional modules / units in the devices disclosed above may be implemented as software, firmware, hardware, or appropriate combinations thereof.

[0182] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0183] It should be understood that in this application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least 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.

[0184] In the several embodiments provided in this 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 schematic. For example, the division of the above-mentioned units is only a logical function division. In actual implementation, there may be other division methods, such as 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 mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0185] The units described above as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0186] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0187] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes multiple instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store programs.

[0188] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but are not intended to limit the scope of the present invention. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and essence of the present invention should be within the scope of the present invention.

Claims

1. A business category prediction method, characterized in that: The method comprises: Acquire a business sample with a business category label in a target business scenario, wherein the business sample includes a business attribute feature and business feature data of the business attribute feature, and the business category label is used to characterize the category into which the business sample is classified in the target business scenario; Classifying the service samples to obtain service training samples and service verification samples; Generating a training association graph of the business attribute features based on the business feature data and business category labels of the business training samples; Generating a verification association graph of the business attribute features based on the business feature data and business category labels of the business verification sample; Performing correlation detection based on the training correlation graph and the verification correlation graph to obtain feature correlation of the business attribute feature; Filtering target business features from the business attribute features based on the feature relevance; The business category prediction model of the target business scenario is trained based on the target business characteristics, and business category prediction is performed according to the trained business category prediction model.

2. The method according to claim 1, characterized in that The performing correlation detection based on the training correlation graph and the verification correlation graph to obtain the feature correlation of the business attribute feature includes: Acquire training sample points from the training association graph; wherein the training sample points include a first horizontal coordinate and a first vertical coordinate, and the verification association graph and the verification association graph corresponding to the same business attribute feature are constructed using the same coordinate system; Acquire a verification sample point from the verification association graph, wherein the verification sample point includes a second abscissa and a second ordinate; If the first horizontal coordinate and the second horizontal coordinate are the same, performing correlation calculation on the first vertical coordinate and the second vertical coordinate to obtain coordinate correlation; The feature correlation of the business attribute feature is determined according to the correlation of at least two coordinates of the business attribute feature.

3. The method according to claim 2, characterized in that The determining the feature correlation of the business attribute feature according to the correlation of at least two coordinates of the business attribute feature includes: Performing graph fluctuation correlation calculation on the training association graph and the verification association graph to obtain graph fluctuation correlation; The feature correlation of the business attribute feature is determined according to the graph fluctuation correlation of the business attribute feature and at least two of the coordinate correlations.

4. The method according to claim 3, characterized in that The performing graph fluctuation correlation calculation on the training association graph and the verification association graph to obtain the graph fluctuation correlation includes: extracting the number of training fluctuation reversals and the number of training fluctuation peak points from the training correlation graph; Extracting the number of verified fluctuation reversals and the number of verified fluctuation peak points from the verified correlation graph; Calculating the difference between the training fluctuation reversal number and the verification fluctuation reversal number to obtain a fluctuation reversal difference; Calculating the difference between the number of training fluctuation peak points and the number of verification fluctuation peak points to obtain a fluctuation peak point difference; The graph fluctuation correlation is calculated based on the fluctuation reversal difference and the fluctuation peak point difference to obtain the graph fluctuation correlation.

5. The method according to any one of claims 1 to 4, characterized in that The generating of the training association graph of the business attribute features based on the business feature data and the business category label of the business training sample includes: If the business attribute feature is a continuous feature, performing feature value conversion on the business feature data of the business training sample according to a preset feature value conversion rule to obtain converted feature data; A training association graph of the business attribute features is generated based on the converted feature data and the business category label.

6. The method according to claim 5, characterized in that Generating the training association graph of the business attribute features according to the converted feature data and the business category label includes: quantizing the service category labels of the service training samples to obtain label quantization values; Calculating the mean of all the label quantization values corresponding to the same converted feature data to obtain the label mean of the converted feature data; Constructing training sample points based on the converted feature data and label mean; A training association graph of the business attribute features is generated based on all the training sample points.

7. The method according to any one of claims 1 to 4, characterized in that The training of the business category prediction model for the target business scenario based on the target business characteristics includes: Updating the service sample based on the target service feature to obtain a target sample; A business category prediction model for the target business scenario is trained based on the target samples.

8. A business category prediction device, characterized in that: The device comprises: An acquisition module is configured to acquire a business sample with a business category label in a target business scenario, wherein the business sample includes a business attribute feature and business feature data of the business attribute feature, and the business category label is used to characterize the category into which the business sample is classified in the target business scenario; A classification module, configured to classify the service samples to obtain service training samples and service verification samples; A first graph generating module, configured to generate a training association graph of the business attribute features based on the business feature data and business category labels of the business training samples; A second graph generating module is configured to generate a verification association graph of the business attribute features based on the business feature data and business category labels of the business verification sample; A detection module, configured to perform correlation detection based on the training correlation graph and the verification correlation graph to obtain a feature correlation of the business attribute feature; A screening module, configured to screen target business features from the business attribute features based on the feature relevance; A training module is used to train a business category prediction model of the target business scenario based on the target business characteristics, and perform business category prediction based on the trained business category prediction model.

9. An electronic device, characterized in that: The electronic device includes 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