User willingness assessment method and device, equipment, storage medium and program product
By using pre-trained feature extraction and willingness evaluation models, automated feature extraction and willingness evaluation of insurance business data is solved, and the problems of inefficient user willingness evaluation and inaccurate results in the prior art are solved, achieving more efficient and accurate user willingness evaluation.
Patent Information
- Application Number
- CN202411287168.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-13
- Publication Date
- 2025-03-14
AI Technical Summary
The existing user willingness evaluation methods are inefficient, the evaluation results are inaccurate, and the reliance on manual feature extraction leads to insufficient efficiency and accuracy.
The pre-trained feature extraction model is used to regularize the insurance business data, extract regular term features and feature global importance parameters, and input these features into the pre-trained willingness evaluation model to obtain the user's insurance willingness evaluation results.
Through the automated feature extraction process, the efficiency and accuracy of feature extraction are improved, and the accuracy and efficiency of user willingness evaluation are significantly improved.
Smart Images
Figure CN119624663A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a user willingness evaluation method, device, equipment, storage medium and program product. Background Art
[0002] With the development of the insurance market, insurance business is increasing. Insurance institutions need to understand the information of insured users, conduct insurance evaluation on the insured users' safety, insurance history and risk-bearing capacity, and predict users' willingness to take out insurance.
[0003] In some related technologies, machine learning methods can be used to evaluate users' willingness to insure: extract relevant features of users and input relevant features into the machine learning model to obtain the insurance willingness evaluation results output by the model. In the above process, the extracted relevant features determine the prediction upper limit of the machine learning model. The more relevant features are extracted, the more comprehensive the representation of user information is, and the more the features can represent the main information of the data. Therefore, in the process of practical application, feature engineering is the key to the success of machine learning.
[0004] However, traditional feature engineering solutions are extremely dependent on manual labor, requiring humans to spend nearly 70% of their time studying real data samples and thinking about the potential forms and data structures of the problems. In addition, feature extraction methods that rely on manual labor may not necessarily describe the data well and dig out the true meaning of the data. The massive features extracted manually may not necessarily have a positive effect on the model. In some cases, manually extracted features may even reduce the performance of the model, and the screening of these features will introduce additional workload.
[0005] Therefore, existing user willingness evaluation methods are inefficient and the evaluation results are inaccurate. Summary of the invention
[0006] The present invention provides a user willingness evaluation method, device, equipment, storage medium and program product, which are used to solve the defects of low efficiency and inaccurate evaluation results of user willingness evaluation methods in the prior art.
[0007] The present invention provides a user willingness assessment method, comprising: obtaining the user's insurance business data; inputting the insurance business data into a pre-trained feature extraction model to obtain regularization item features and feature global importance parameters output by the feature extraction model; inputting the regularization item features and feature global importance parameters into a pre-trained willingness assessment model to obtain the user's insurance willingness assessment result output by the willingness assessment model.
[0008] According to a user willingness evaluation method provided by the present invention, a feature extraction model is used to perform regularization processing on insurance business data to obtain regularization item features corresponding to the insurance business data; perform data balancing processing on the regularization item features to obtain reconstructed features; perform global importance evaluation and normalization processing on the reconstructed features to obtain feature global importance parameters; and output regularization item features and feature global importance parameters.
[0009] According to a user willingness evaluation method provided by the present invention, the willingness evaluation model includes a first network layer, a second network layer and a third network layer; the first network layer includes a first Feature transformer layer, a first split layer and a first Attentive transformer layer connected in sequence; the second network layer includes a first Mask layer, a second Feature transformer layer, a second split layer, an activation layer and a second Attentive transformer layer; the first Mask layer, the second Feature transformer layer, the second split layer and the first activation layer are connected in sequence, and the second split layer and the second Attentive transformer layer are connected; the third network layer includes a second Mask layer, a third Feature transformer layer, a third split layer, a second activation layer, a plurality of additive models and a FC layer connected in sequence; wherein the first Attentive transformer layer is connected to the second Feature transformer layer, the second Attentive transformer layer is connected to the second Mask layer, and the first activation layer and the second activation layer are connected to the FC layer through a plurality of additive models.
[0010] According to a user willingness assessment method provided by the present invention, before obtaining the user's insurance business data, it also includes: obtaining sample insurance business data, sample regularization item features corresponding to the sample insurance business data, and sample feature global importance parameters corresponding to the sample insurance business data; based on the sample insurance business data, sample regularization item features corresponding to the sample insurance business data, and sample feature global importance parameters corresponding to the sample insurance business data, self-supervised pre-training of an initial deep forest model is performed to obtain a feature extraction model.
[0011] According to a user willingness assessment method provided by the present invention, before obtaining the user's insurance business data, it also includes: obtaining sample regularization item features and sample feature global importance parameters; based on the sample regularization item features and the sample feature global importance parameters, performing self-supervised pre-training on the initial TabNet model to obtain a willingness assessment model.
[0012] According to a user willingness evaluation method provided by the present invention, obtaining the user's insurance business data includes: obtaining the user's initial insurance business data; performing data analysis and data preprocessing on the initial insurance business data to obtain the insurance business data.
[0013] The present invention also provides a user willingness assessment device, comprising: an acquisition module for acquiring the user's insurance business data; a feature extraction module for inputting the insurance business data into a pre-trained feature extraction model to obtain regularization item features and feature global importance parameters output by the feature extraction model; an assessment module for inputting the regularization item features and feature global importance parameters into a pre-trained willingness assessment model to obtain the user's insurance willingness assessment result output by the willingness assessment model.
[0014] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, any of the user willingness evaluation methods described above is implemented.
[0015] The present invention also provides a non-transitory computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the user willingness evaluation method as described above is implemented.
[0016] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the above-mentioned user willingness evaluation methods.
[0017] The user willingness assessment method, device, equipment, storage medium and program product provided by the present invention obtain the user's insurance business data; input the insurance business data into a pre-trained feature extraction model to obtain the regular item features and feature global importance parameters output by the feature extraction model; input the regular item features and feature global importance parameters into the pre-trained willingness assessment model to obtain the user's insurance willingness assessment result output by the willingness assessment model. Through the above method, the feature extraction model is used to extract features from the user's insurance business data to avoid artificial construction and feature extraction, which can improve the efficiency and accuracy of feature extraction, and then the willingness assessment model is used to evaluate and predict the regular item features and feature global importance parameters output by the feature extraction model to obtain the user's insurance willingness assessment result, which can effectively improve the accuracy of the assessment result. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0019] Figure 1 This is one of the flow charts of the user willingness evaluation method provided by the present invention.
[0020] Figure 2 It is a structural schematic diagram of the willingness assessment model provided by the present invention.
[0021] Figure 3 This is the second flow chart of the user willingness evaluation method provided by the present invention.
[0022] Figure 4 It is a structural schematic diagram of the user willingness evaluation device provided by the present invention.
[0023] Figure 5 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION
[0024] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0025] It should be noted that in the description of the present invention, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "include one..." do not exclude the existence of other identical elements in the process, method, article or device including the elements. The orientation or position relationship indicated by the terms "upper", "lower" and the like is based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present invention. Unless otherwise clearly specified and limited, the terms "installed", "connected" and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be an indirect connection through an intermediate medium, or it can be a connection between the two elements. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.
[0026] The terms "first", "second", etc. in the present invention are used to distinguish similar objects, and are not used to 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 invention can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first", "second", etc. are generally of the same type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" means at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.
[0027] It should be noted that all actions of acquiring signals, information or data in the present invention are performed in compliance with the corresponding data protection laws and policies of the country where the device is located and with the authorization given by the owner of the corresponding device.
[0028] See also Figure 1 , Figure 1 This is one of the flow charts of the user willingness evaluation method provided by the present invention. In this embodiment, the user willingness evaluation method includes steps S110 to S130, and each step is specifically as follows: S110: Obtain the user's insurance business data.
[0029] Optionally, the insurance business data is in tabular form.
[0030] S120: Input the insurance business data into the pre-trained feature extraction model to obtain the regularization item features and feature global importance parameters output by the feature extraction model.
[0031] Optionally, the data analysis tool matplotlib can be used to explore the data scale of insurance business data, and then the model can be used to complete feature engineering, realize feature selection and synthesis, and generate feature statistics.
[0032] Preferably, the feature extraction model adopts a deep forest encoding (DFE) model; deep forest encoding is a method based on ensemble learning, which uses a deep forest model to encode features and can improve the prediction performance of the model.
[0033] S130: Input the regularization term features and the feature global importance parameters into the pre-trained willingness assessment model to obtain the user's insurance willingness assessment result output by the willingness assessment model.
[0034] Preferably, the willingness assessment model adopts the TabNet model.
[0035] Considering that most insurance business data is in tabular form, the TabNet model is a prediction model based on the self-attention mechanism. It can better capture the complex relationship between features when processing tabular data by modeling the self-attention weights of features.
[0036] Therefore, in this embodiment, the deep forest model is selected for feature extraction, and the TabNet model is used for further feature processing and prediction. The combination of the deep forest model and the TabNet model can fully utilize the advantages of both in the prediction task, thereby improving the accuracy of the user's insurance willingness assessment.
[0037] Preferably, the deep forest encoding method is used to encode the original data, and the high-dimensional insurance business data is converted into a low-dimensional feature representation; further, the encoded low-dimensional feature representation is input into the TabNet model, and its self-attention mechanism can be used to learn the complex relationship between features and evaluate user willingness. In the prediction process, the deep forest model and the TabNet model can be fused by weighted averaging or stacking to further improve the prediction performance and model stability.
[0038] Preferably, the deep forest model and the TabNet model are trained using a cross-validation method to improve the model performance of the deep forest model and the TabNet model.
[0039] The user willingness assessment method provided in this embodiment obtains the user's insurance business data; inputs the insurance business data into a pre-trained feature extraction model to obtain the regularization item features and feature global importance parameters output by the feature extraction model; inputs the regularization item features and feature global importance parameters into the pre-trained willingness assessment model to obtain the user's insurance willingness assessment result output by the willingness assessment model. In the above manner, the feature extraction model is used to extract features from the user's insurance business data to avoid artificial construction and feature extraction, which can improve the efficiency and accuracy of feature extraction, and then the willingness assessment model is used to evaluate and predict the regularization item features and feature global importance parameters output by the feature extraction model to obtain the user's insurance willingness assessment result, which can effectively improve the accuracy of the assessment result.
[0040] In some embodiments, a feature extraction model is used to perform regularization processing on insurance business data to obtain regularization item features corresponding to the insurance business data; perform data balancing processing on the regularization item features to obtain reconstructed features; perform global importance evaluation and normalization processing on the reconstructed features to obtain feature global importance parameters; and output regularization item features and feature global importance parameters.
[0041] Specifically, after obtaining the user's initial insurance business data, data analysis and data preprocessing are performed on the initial insurance business data to obtain insurance business data; the insurance business data is input into a feature extraction model, and the feature extraction model can regularize the preprocessed insurance business data to enhance the sparse capability of the model and generate regularized item features corresponding to the insurance business data.
[0042] In this embodiment, the regularization feature is improved based on the user's insurance scenario. The expression is as follows: ; in, Represents the total time step of the feature extraction model; the Mask matrix (mask matrix) of the feature extraction model in the data processing process has dimensions B and D, dimension B corresponds to the number of rows of insurance business data, and dimension D corresponds to the characteristics of different insurance business data; the Mask matrix is actually composed of masks composed of 0 and 1. represents the insured data in the Mask matrix at the bth row and jth column and the attention weight of the i-th sample corresponding to the insured data. The attention weight can be calculated by the sparse probability activation function Sparsemax, which is a method similar to the sparse Softmax function. represents the Gaussian white noise introduced in the feature extraction model calculation process. The above regularization feature formula takes into account the average entropy value, which can reflect The sparsity of .
[0043] After the feature extraction model is trained, the parameters in the above regularization feature formula can be applied to the instance-wise parameters of the Attentive transformer layer in the TabNet model. In short, the regularization process is to calculate the corresponding attention weights based on the input data at different time steps through the insurance data and reference position of the Mask matrix, and then use the Sparsemax function for sparseness on this basis, and further consider Gaussian white noise and regularization terms to obtain sparsity. Finally, these parameters will act on the Attentive transformer layer in the TabNet model.
[0044] It should be noted that the goal of regularization is to establish a model in the insurance business scenario that can calculate attention weights based on different time steps of time series data, using the insurance business data and reference position information of the Mask matrix. These weights are processed by the Sparsemax function, taking into account noise and regularization terms, to obtain better sparsity, and are ultimately used in the Attentive transformer layer in the TabNet model, so that the TabNet model can better understand and process insurance business data, which helps the TabNet model better adapt to the characteristics and data characteristics of different time points, improves the application potential of the TabNet model in the insurance field, and thus better predicts the user's willingness to insure.
[0045] Furthermore, data balancing is performed on the regularization term features to obtain the reconstructed features.
[0046] Specifically, the feature extraction model uses self-supervised learning to train the insurance business data in tabular form, and uses the regularized mean square error as the error value; after obtaining the regularized item features, the feature extraction model can perform data balancing on the regularized item features to obtain the reconstructed features. , reconstruction features The expression is as follows: ; The dimension of the matrix S generated in the feature extraction model data processing process is B×D, B is the number of rows of the matrix S, and D is the number of columns of the matrix S; variables b and j represent the row index and column index in the matrix S, respectively, and the value ranges are b=1, 2, ..., B and j=1, 2, ..., D, respectively; represents the unregularized user insurance feature data at the bth row and jth column in the matrix S, represents the normalized user insurance feature data at the bth row and jth column in the matrix S. The feature data is transformed into (1-S) and After multiplication, it can be passed as input to the encoder model to generate feature output; generally, in the decoder model, the summed features that have not passed through the fully connected layer is used as input, these feature vectors are The sum of the time steps can generate the reconstructed feature fnew.
[0047] Preferably, during the model training process of the feature extraction model, these reconstructed features can also be added to the original insurance business data training set to complete the feature conversion process. That is, by constructing the matrix S and combining it with the feature data, after the encoding and decoding process, the reconstructed features can be generated and added to the training data set.
[0048] In summary, this process aims to transform the original user insurance feature data into a more informative representation through encoding and decoding models, which is conducive to the subsequent use of machine learning models for classification prediction to analyze and predict the user's willingness to insure. Balanced analysis processing can ensure the reliability and accuracy of the model, thereby providing strong support for insurance business prediction.
[0049] Furthermore, the reconstructed features are globally evaluated and normalized to obtain feature global importance parameters; and regularization term features and feature global importance parameters are output.
[0050] Specifically, the reconstructed features are globally evaluated and normalized to obtain feature global importance parameters, which can be interpreted by the TabNet model, where the feature global importance parameter The expression is as follows: ; in, Represents the total time step of the feature extraction model; the Mask matrix (mask matrix) of the feature extraction model in the data processing process has dimensions B and D, dimension B corresponds to the number of rows of insurance business data, and dimension D corresponds to the characteristics of different insurance business data; the Mask matrix is actually composed of masks composed of 0 and 1. Represents the insurance data in the Mask matrix at the bth row and jth column position and the attention weight of the i-th sample corresponding to the insurance data; Represents the model parameter in the Mask matrix at the bth row and jth column position.
[0051] Furthermore, the feature extraction model outputs regularization term features and feature global importance parameters.
[0052] In some embodiments, the willingness assessment model includes a first network layer, a second network layer and a third network layer; the first network layer includes a first Feature transformer layer, a first split layer and a first Attentive transformer layer connected in sequence; the second network layer includes a first Mask layer, a second Feature transformer layer, a second split layer, an activation layer and a second Attentive transformer layer; the first Mask layer, the second Feature transformer layer, the second split layer and the first activation layer are connected in sequence, and the second split layer and the second Attentive transformer layer are connected; the third network layer includes a second Mask layer, a third Feature transformer layer, a third split layer, a second activation layer, multiple additive models and a FC layer connected in sequence; wherein, the first Attentive transformer layer and the second Feature transformer layer are connected, the second Attentive transformer layer and the second Mask layer are connected, and the first activation layer and the second activation layer are connected to the FC layer through multiple additive models.
[0053] See also Figure 2 , Figure 2 It is a structural schematic diagram of the willingness assessment model provided by the present invention.
[0054] In this embodiment, the willingness assessment model adopts the TabNet model, and the TabNet model architecture includes multiple Attentive transformer layers, Mask layers, and Feature transformer layers.
[0055] Among them, the Attentive transformer layer can obtain the weight distribution required by the Mask layer of the current step according to the output result of the previous layer; the Mask layer screens the matrix vector; the Feature transformer layer can calculate and process the features selected in the current step, and then cut the vector output by the Feature transformer layer into two parts through the split layer, one of which is output to the activation layer (the activation layer uses the Relu activation function, in Figure 2 The other part is used for the next Mask layer calculation. The TabNet model can be finally constructed through multiple additive models (i.e. Figure 2 The circle containing “+” in the output.
[0056] Specifically, Figure 2 As shown, Figure 2 From top to bottom, they are the first network layer, the second network layer, and the third network layer of the willingness assessment model, which are separated by dotted lines; the standardized data output by the feature extraction model will be input into the first network layer, the second network layer, and the third network layer of the willingness assessment model respectively.
[0057] The first network layer includes a first Feature transformer layer, a first split layer, and a first Attentive transformer layer connected in sequence.
[0058] The second network layer includes the first Mask layer, the second Feature transformer layer, the second split layer, the activation layer, and the second Attentive transformer layer; the first Mask layer, the second Feature transformer layer, the second split layer, and the first activation layer are connected in sequence, and the second split layer and the second Attentive transformer layer are connected.
[0059] The third network layer includes the second Mask layer, the third Feature transformer layer, the third split layer, the second activation layer, multiple additive models and the FC layer, which are connected in sequence.
[0060] Among them, the first Attentive transformer layer is connected to the second Feature transformer layer, the second Attentive transformer layer is connected to the second Mask layer, and the first activation layer and the second activation layer are connected to the FC layer through multiple additive models.
[0061] To facilitate understanding, the willingness assessment model is explained here in conjunction with the insurance business scenario.
[0062] In the prediction of insurance business, the Attentive transformer layer in the willingness assessment model can be understood as a "feature selection and weight assignment layer". The Attentive transformer layer can select the most relevant features for the data of the current step based on the results of previous data analysis and assign weights to them, which helps to determine which factors are the most critical factors for predicting insurance decisions.
[0063] The “Mask layer” can be understood as the “risk assessment layer”. In the Mask layer, the matrix vector will be screened to assess the risk level of the insurance case, which helps determine whether the user accepts or rejects the insurance application.
[0064] The “Feature transformer layer” can be understood as a “feature processing layer”. The Feature transformer layer can perform various statistical and computational operations on the selected features to better understand the insurance data and prepare for predictive modeling.
[0065] The "split layer" can be understood as a "decision division layer". The split layer can split the output of the feature processing layer into two parts. One part is used for further decision making, such as determining the insurance premium or underwriting conditions, and the other part can be used to calculate the next risk assessment.
[0066] Multiple additive models can be understood as a “decision output layer”, which is used to integrate all information to make the final insurance decision and generate the user’s insurance willingness assessment results.
[0067] In some embodiments, before obtaining the user's insurance business data, it also includes: obtaining sample insurance business data, sample regularization item features corresponding to the sample insurance business data, and sample feature global importance parameters corresponding to the sample insurance business data; based on the sample insurance business data, sample regularization item features corresponding to the sample insurance business data, and sample feature global importance parameters corresponding to the sample insurance business data, self-supervised pre-training of the initial deep forest model is performed to obtain a feature extraction model.
[0068] See also Figure 3 , Figure 3 This is the second flow chart of the user willingness evaluation method provided by the present invention.
[0069] like Figure 3 As shown in the figure, model training is required before evaluating user willingness.
[0070] Specifically, sample insurance business data, sample regularization item features corresponding to the sample insurance business data, and sample feature global importance parameters corresponding to the sample insurance business data are obtained to construct a data set; data analysis and data preprocessing are performed on the data in the data set; based on the preprocessed sample insurance business data, sample regularization item features corresponding to the sample insurance business data, and sample feature global importance parameters corresponding to the sample insurance business data, an initial deep forest model is self-supervised pre-trained to obtain a feature extraction model.
[0071] In some embodiments, before obtaining the user's insurance business data, it also includes: obtaining sample regularization item features and sample feature global importance parameters; based on the sample regularization item features and the sample feature global importance parameters, performing self-supervised pre-training on the initial TabNet model to obtain a willingness assessment model.
[0072] Please continue reading Figure 3The sample insurance business data that has undergone data preprocessing, the sample regularization item features corresponding to the sample insurance business data, and the sample feature global importance parameters corresponding to the sample insurance business data in the data set are divided into a training set and a data set in a ratio of 7:3, which are used to perform self-supervised pre-training on the initial TabNet model to obtain the willingness evaluation model.
[0073] Preferably, the TabNet model is trained using a cross-validation method, and the model parameters of the TabNet model are optimized and adjusted to improve the model performance of the TabNet model.
[0074] The user willingness assessment method provided in this embodiment proposes a user insurance willingness assessment method combining the deep forest model and the TabNet model for the insurance scenario. In the deep forest model, the CNN sliding convolution kernel can be used for feature extraction. The original features are scanned through a multi-granularity scanning method and a sliding window to generate and output features. This feature extraction method effectively enhances the features, mines the relationship between features, and enhances the expressiveness of features, so that the deep forest model has achieved higher results in various classic data sets. Fusion of the output results of the deep forest enhancement into the TabNet model can achieve richer feature construction without complicated manual intervention, and has achieved good results in actual scenario tests.
[0075] In some embodiments, obtaining the user's insurance business data includes: obtaining the user's initial insurance business data; performing data analysis and data preprocessing on the initial insurance business data to obtain the insurance business data.
[0076] The present invention also provides a user willingness evaluation device. Figure 4 , Figure 4 4 is a schematic diagram of the structure of the user willingness evaluation device provided by the present invention. In this embodiment, the user willingness evaluation device includes an acquisition module 410 , a feature extraction module 420 and an evaluation module 430 .
[0077] The acquisition module 410 is used to acquire the user's insurance business data.
[0078] The feature extraction module 420 is used to input the insurance business data into the pre-trained feature extraction model to obtain the regularization item features and feature global importance parameters output by the feature extraction model.
[0079] The evaluation module 430 is used to input the regularization term features and the feature global importance parameters into the pre-trained willingness evaluation model to obtain the user's insurance willingness evaluation result output by the willingness evaluation model.
[0080] In some embodiments, a feature extraction model is used to perform regularization processing on insurance business data to obtain regularization item features corresponding to the insurance business data; perform data balancing processing on the regularization item features to obtain reconstructed features; perform global importance evaluation and normalization processing on the reconstructed features to obtain feature global importance parameters; and output regularization item features and feature global importance parameters.
[0081] In some embodiments, the willingness assessment model includes a first network layer, a second network layer and a third network layer; the first network layer includes a first Feature transformer layer, a first split layer and a first Attentive transformer layer connected in sequence; the second network layer includes a first Mask layer, a second Feature transformer layer, a second split layer, an activation layer and a second Attentive transformer layer; the first Mask layer, the second Feature transformer layer, the second split layer and the first activation layer are connected in sequence, and the second split layer and the second Attentive transformer layer are connected; the third network layer includes a second Mask layer, a third Feature transformer layer, a third split layer, a second activation layer, multiple additive models and a FC layer connected in sequence; wherein, the first Attentive transformer layer and the second Feature transformer layer are connected, the second Attentive transformer layer and the second Mask layer are connected, and the first activation layer and the second activation layer are connected to the FC layer through multiple additive models.
[0082] In some embodiments, the feature extraction module 420 is used to obtain sample insurance business data, sample regularization item features corresponding to the sample insurance business data, and sample feature global importance parameters corresponding to the sample insurance business data; based on the sample insurance business data, sample regularization item features corresponding to the sample insurance business data, and sample feature global importance parameters corresponding to the sample insurance business data, the initial deep forest model is self-supervised pre-trained to obtain a feature extraction model.
[0083] In some embodiments, the evaluation module 430, before being used to obtain the user's insurance business data, also includes: obtaining sample regularization item features and sample feature global importance parameters; based on the sample regularization item features and the sample feature global importance parameters, performing self-supervised pre-training on the initial TabNet model to obtain a willingness evaluation model.
[0084] In some embodiments, the acquisition module 410 is used to acquire the user's insurance business data, including: acquiring the user's initial insurance business data; performing data analysis and data preprocessing on the initial insurance business data to obtain the insurance business data.
[0085] The invention also provides an electronic device. Figure 5 is a schematic diagram of the structure of the electronic device provided by the present invention, such as Figure 5 As shown, the electronic device may include: a processor 510, a communication interface 520, a memory 530 and a communication bus 540, wherein the processor 510, the communication interface 520 and the memory 530 communicate with each other through the communication bus 540. The processor 510 may call the logic instructions in the memory 530 to execute the user willingness evaluation method.
[0086] In addition, the logic instructions in the above-mentioned memory 530 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.
[0087] The present invention also provides a non-transitory computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the user willingness evaluation method provided by the above methods is implemented.
[0088] The present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the user willingness evaluation method provided by the above methods.
[0089] The device embodiments described above are merely illustrative, wherein 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 may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.
[0090] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.
[0091] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A user willingness evaluation method, characterized in that: include: Obtain the user's insurance business data; Inputting the insurance business data into a pre-trained feature extraction model to obtain regularization term features and feature global importance parameters output by the feature extraction model; The regularization item feature and the feature global importance parameter are input into a pre-trained willingness assessment model to obtain the user's insurance willingness assessment result output by the willingness assessment model.
2. The user willingness evaluation method according to claim 1, characterized in that: The feature extraction model is used to perform regularization processing on the insurance business data to obtain regularization item features corresponding to the insurance business data; Performing data balancing processing on the regularization term features to obtain reconstruction features; Performing global importance evaluation and normalization processing on the reconstructed features to obtain global importance parameters of the features; Output the regularization term features and the feature global importance parameters.
3. The user willingness evaluation method according to claim 1, characterized in that: The willingness assessment model includes a first network layer, a second network layer and a third network layer; The first network layer includes a first Feature transformer layer, a first split layer, and a first Attentive transformer layer connected in sequence; The second network layer includes a first Mask layer, a second Feature transformer layer, a second split layer, an activation layer, and a second Attentive transformer layer; the first Mask layer, the second Feature transformer layer, the second split layer, and the first activation layer are connected in sequence, and the second split layer is connected to the second Attentive transformer layer; The third network layer includes a second Mask layer, a third Feature transformer layer, a third split layer, a second activation layer, a plurality of additive models and an FC layer connected in sequence; Among them, the first attentive transformer layer is connected to the second feature transformer layer, the second attentive transformer layer is connected to the second mask layer, and the first activation layer and the second activation layer are connected to the FC layer through multiple additive models.
4. The user willingness evaluation method according to claim 1, characterized in that: Before obtaining the user's insurance business data, the method further includes: Obtaining sample insurance business data, sample regularization item features corresponding to the sample insurance business data, and sample feature global importance parameters corresponding to the sample insurance business data; Based on the sample insurance business data, the sample regularization term features corresponding to the sample insurance business data and the sample feature global importance parameters corresponding to the sample insurance business data, the initial deep forest model is self-supervised pre-trained to obtain the feature extraction model.
5. The user willingness evaluation method according to claim 1, characterized in that: Before obtaining the user's insurance business data, the method further includes: Obtain sample regularization term features and sample feature global importance parameters; Based on the sample regularization term features and sample feature global importance parameters, the initial TabNet model is self-supervised pre-trained to obtain the willingness assessment model.
6. The user willingness evaluation method according to claim 1, characterized in that: The obtaining of the user's insurance business data includes: Obtain the user's initial insurance business data; The initial insurance business data is subjected to data analysis and data preprocessing to obtain the insurance business data.
7. A user willingness evaluation device, characterized in that: include: An acquisition module is used to acquire the user's insurance business data; A feature extraction module, used for inputting the insurance business data into a pre-trained feature extraction model to obtain regularization term features and feature global importance parameters output by the feature extraction model; An evaluation module is used to input the regularization item features and the feature global importance parameters into a pre-trained willingness evaluation model to obtain the user's insurance willingness evaluation result output by the willingness evaluation model.
8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the user willingness evaluation method according to any one of claims 1 to 6 is implemented.
9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the user willingness evaluation method according to any one of claims 1 to 6 is implemented.
10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the user willingness evaluation method according to any one of claims 1 to 6 is implemented.