WO2022121217A1

WO2022121217A1 - Quota prediction method and device, and computer-readable storage medium

Info

Publication number: WO2022121217A1
Application number: PCT/CN2021/090569
Authority: WO
Inventors: 郝晓丽; 陈吕; 张乐婷; 洪霞; 袁丽乔
Original assignee: 平安科技（深圳）有限公司
Priority date: 2020-12-07
Filing date: 2021-04-28
Publication date: 2022-06-16
Also published as: CN112529716A; CN112529716B

Abstract

A quota prediction method and device, and a computer-readable storage medium, which relate to classification model and blockchain technology. According to the method, a follow-up insurance quota estimation is performed only when a risk level is detected to be within a certain range, so as to prevent negative impact that can be caused by continually insuring customers of a high risk level; by means of selecting a risk assessment model corresponding to a current insurance type for risk assessment, an assessment error caused by different insurance types is avoided; the risk assessment model is obtained by means of using feature engineering and machine learning training, and target features are obtained by using the risk assessment model, so that risk assessment results finally obtained by the risk assessment model are more accurate; by means of synthesizing previous insurance information and quota adjustment factors to predict a coverage range that customers can currently underwrite, insurance quota information consistent with the actual situation can be obtained at the time of the current insurance, wherein the previous insurance information can be stored in a blockchain.

Description

Quota prediction method, device and computer-readable storage medium

This application claims the priority of a Chinese patent application with an application number of 202011420120.9 and an invention title of "Amount Prediction Method, Device and Computer-readable Storage Medium" filed with the China Patent Office on December 7, 2020, the entire contents of which are incorporated by reference in this application.

technical field

The present application relates to the technical field of artificial intelligence, and in particular, to a quota prediction method, device, and computer-readable storage medium.

Background technique

In the insurance industry, for customers to apply for insurance, the common practice in the industry is that after the customer fills in the insurance application information, the company makes a decision to underwrite or not to underwrite after review. However, the inventor realized that it is difficult for the customer or the agent in charge of insurance sales to directly obtain the accurate information on the amount of insurance that can continue to be insured after this policy is insured. Considering that if the amount of re-insurance is too large, the customer or the agent will face the risk of surrender or refusal of insurance by the insurance company. When re-insuring, they tend to be more conservative and invest a smaller amount in order to be successful in insurance. It is easy to happen that the amount of insurance that the customer purchases again is not enough to meet their protection needs, which leads to the technical problem of low accuracy in predicting the amount of insurance when adding insurance.

technical problem

The main purpose of the present application is to provide an amount prediction method, device and computer-readable storage medium, which aims to solve the technical problem of low accuracy in predicting the insurance amount when adding insurance.

technical solutions

In order to achieve the above purpose, the present application provides a quota prediction method, the quota prediction method includes: obtaining the insurance application information to be evaluated of the target insurance applicant, and obtaining the previous insurance application information of the target insurance applicant according to the insurance application information to be evaluated; Type of insurance application this time; select a pre-trained risk assessment model corresponding to the type of insurance application this time, use the risk assessment model to extract text features and numerical features in the insurance application information to be assessed, and combine the text features with The target feature is obtained by combining the digital features, wherein the risk assessment model is obtained by using the training data set screened by feature engineering to be trained by a machine learning algorithm; the risk level of the insurance application information to be evaluated is determined according to the target feature, and the When the risk level is lower than high risk, obtain the limit adjustment factor of the insurance application information to be evaluated; combine the previous insurance application information and the limit adjustment factor to predict the insurance addition limit information of the target insurance applicant.

In addition, in order to achieve the above purpose, the present application also provides a quota prediction device, the quota prediction device includes: a past information acquisition module for acquiring the insurance application information to be evaluated of the target insurance applicant, and obtaining the insurance application information to be evaluated according to the to-be-evaluated insurance application information The previous insurance application information of the target applicant and the type of this insurance application; the target feature acquisition module is used to select a pre-trained risk assessment model corresponding to the current insurance application type, and use the risk assessment model to extract the to-be-evaluated The text features and digital features in the insurance application information, and the target features are obtained by combining the text features and digital features, wherein the risk assessment model is obtained by using the training data set screened by feature engineering to be trained by a machine learning algorithm; risk level A determination module, configured to determine the risk level of the insurance application information to be evaluated according to the target characteristics, and obtain the quota adjustment factor of the insurance application information to be evaluated when it is detected that the risk level is lower than high risk; add insurance The limit prediction module is used for predicting the added insurance limit information of the target insured in combination with the previous insurance application information and the limit adjustment factor.

In addition, in order to achieve the above object, the present application also provides a quota prediction device, the quota prediction device includes a processor, a memory, and a quota prediction program stored on the memory and executable by the processor, wherein the When the above-mentioned quota prediction program is executed by the processor, the following method is implemented: obtaining the insurance application information to be evaluated of the target insurance applicant, and obtaining the previous insurance application information and the current insurance application type of the target insurance applicant according to the to-be-evaluated insurance application information; Selecting the pre-trained risk assessment model corresponding to the type of insurance this time, using the risk assessment model to extract the text features and numerical features in the insurance application information to be assessed, and combining the text features and numerical features to obtain a target feature, wherein the risk assessment model is obtained by using the training data set screened by feature engineering to be trained by a machine learning algorithm; the risk level of the insurance application information to be evaluated is determined according to the target feature, and the risk level is detected when the risk level is detected. When the level is below high risk, obtain the limit adjustment factor of the insurance application information to be evaluated; combine the previous insurance application information and the limit adjustment factor to predict the insurance addition limit information of the target insured.

In addition, in order to achieve the above purpose, the present application also provides a computer-readable storage medium, where a quota prediction program is stored on the computer-readable storage medium, wherein when the quota prediction program is executed by a processor, the following method is implemented: obtaining The insurance application information to be evaluated of the target insurance applicant, and the previous insurance application information and the current insurance application type of the target insurance applicant are obtained according to the to-be-evaluated insurance application information; the pre-trained risk assessment model corresponding to the current insurance application type is selected, Use the risk assessment model to extract text features and numerical features in the insurance application information to be assessed, and combine the text features and numerical features to obtain target features, wherein the risk assessment model is a training program selected by feature engineering The data set is obtained by training the machine learning algorithm; the risk level of the insurance application information to be evaluated is determined according to the target characteristics, and when it is detected that the risk level is lower than high risk, the quota adjustment of the insurance application information to be evaluated is obtained factor; combining the previous insurance application information and the limit adjustment factor, predict the insurance addition limit information of the target insured.

beneficial effect

The present application realizes that the insurable amount information consistent with the actual situation can be obtained at the time of applying for insurance, thereby solving the technical problem of low accuracy in predicting the insured amount when applying for insurance. At the same time, it also saves the customer from being disturbed by the insurance salesman who does not know the amount of insurance that can be added to him, such as physical examination and coordination.

Description of drawings

FIG. 1 is a schematic diagram of the hardware structure of the quota prediction device involved in the solution of the embodiment of the present application.

FIG. 2 is a schematic flowchart of the first embodiment of the quota prediction method of the application.

FIG. 3 is a schematic diagram of functional modules of the quota prediction device of the present application.

The realization, functional characteristics and advantages of the purpose of the present application will be further described with reference to the accompanying drawings in conjunction with the embodiments.

Embodiments of the present invention

It should be understood that the specific embodiments described herein are only used to explain the present application, but not to limit the present application.

The technical solution of the present application relates to the field of artificial intelligence technology, and can be applied to scenarios such as financial technology such as insurance limit prediction, so as to promote the construction of smart cities. Optionally, the data involved in this application, such as insurance application information and/or insurance limit information, may be stored in a database, or may be stored in a blockchain, which is not limited in this application.

The quota prediction method involved in the embodiment of the present application is mainly applied to a quota prediction device, and the quota prediction device may be a device with display and processing functions, such as a PC, a portable computer, and a mobile terminal.

Referring to FIG. 1 , FIG. 1 is a schematic diagram of the hardware structure of the quota prediction device involved in the solution of the embodiment of the present application. In this embodiment of the present application, the quota prediction device may include a processor 1001 (for example, a CPU), a communication bus 1002 , a user interface 1003 , a network interface 1004 , and a memory 1005 . Wherein, the communication bus 1002 is used to realize the connection and communication between these components; the user interface 1003 may include a display screen (Display), an input unit such as a keyboard (Keyboard); the network interface 1004 may optionally include a standard wired interface, a wireless interface (such as a WI-FI interface); the memory 1005 can be a high-speed RAM memory, or a stable memory (non-volatile memory), such as a disk memory, and the memory 1005 can optionally be a storage device independent of the aforementioned processor 1001 .

Those skilled in the art can understand that the hardware structure shown in FIG. 1 does not constitute a limitation on the quota prediction device, and may include more or less components than the one shown, or combine some components, or arrange different components.

Continuing to refer to FIG. 1 , the memory 1005 as a computer-readable storage medium in FIG. 1 may include an operating system, a network communication module, and a quota prediction program.

In FIG. 1 , the network communication module is mainly used to connect to the server and perform data communication with the server; and the processor 1001 can call the quota prediction program stored in the memory 1005 and execute the quota prediction method provided by the embodiment of the present application.

Based on the above hardware structure, various embodiments of the quota prediction method of the present application are proposed.

In the insurance industry, for customers to apply for insurance, the common practice in the industry is that after the customer fills in the insurance application information, the company makes a decision to underwrite or not to underwrite after review. However, customers cannot know the amount of coverage they can still purchase. At the same time, in the process of insurance sales, agents are also confused. For old customers, they do not know how much insurance they have and how much insurance they can buy; for new customers, they are not sure how much insurance should be matched for new customers. In addition, because the purchase of an excessively high amount of insurance will be subject to an underwriting limit, agents are usually conservative when designing insurance plans, which leads to the fact that the amount of insurance purchased by many customers is not enough to meet their protection needs. All of the above situations reflect the current technical problem of low efficiency in obtaining information on the added insurance amount.

In order to solve the above-mentioned problems, the present application provides a quota prediction method, that is, by obtaining the customer's insurance application information to obtain their previous insurance application information, using a pre-trained risk assessment model to determine the corresponding risk level, and only when the detected risk level is in the Subsequent reinsurance limit estimation is carried out only when it is within a certain range (below the high risk level), which avoids the negative impact that may be caused by customers with higher risk levels continuing to apply for insurance. The evaluation avoids the evaluation error caused by different types of insurance; the risk evaluation model is obtained by using feature engineering and machine learning training, and the target characteristics of the insurance information to be evaluated are obtained by using the risk evaluation model, so that the risk evaluation model finally obtained by the risk evaluation model The results are more accurate; by synthesizing the previous insurance information and the limit adjustment factor, the customer's current insured amount can be calculated intelligently, and the insurable amount information that is consistent with the actual situation can be obtained at the time of the current insurance application, so as to solve the problem. It solves the technical problem of low accuracy in predicting the insurance amount when adding insurance. At the same time, it also saves the customer from being disturbed by the insurance salesman who does not know the amount of insurance that can be added to him, such as physical examination and coordination.

Referring to FIG. 2 , FIG. 2 is a schematic flowchart of the first embodiment of the quota prediction method of the present application.

The first embodiment of the present application provides a quota prediction method, and the quota prediction method includes the following steps.

Step S10 , obtaining the insurance application information to be evaluated of the target insurance applicant, and obtaining the previous insurance application information and the current insurance application type of the target insurance applicant according to the to-be-evaluated insurance application information.

In this embodiment, the target insurance applicant, that is, the insurance application user, can fill in the information required for this insurance application on a personal terminal, a counter terminal of an insurance company, or other terminal equipment. The insurance application information to be evaluated may include the identity information of the insured user, the type of insurance this time, the amount of this insurance application, the date of application for insurance, and information about personal and family medical history. The previous insurance application information is the relevant information that the insured user has applied for insurance before this application, which may specifically include the historical insurance type, historical insurance amount, historical insurance time, etc. The previous acquisition method of insurance application information can be obtained directly from the information to be evaluated, or can be accessed to a database storing the user's insurance application information according to the personal identity information in the insurance application information to be evaluated. The type of insurance applied for this time can be the actual type of insurance such as life insurance type, critical illness insurance type, etc.

Specifically, if the current user fills in the corresponding insurance application questionnaire on the counter terminal of the insurance company, after completing the filling and submitting, the server receives the questionnaire data, uses the questionnaire data as the above-mentioned insurance application information to be evaluated, and then accesses the insurance application within the company. Information database, query to obtain the previous insurance information of the insured user and the type of insurance purchased this time.

Step S20, select a pre-trained risk assessment model corresponding to the current insurance application type, use the risk assessment model to extract text features and numerical features in the insurance application information to be assessed, and combine the text features and numerical features. The target features are obtained by combining, wherein the risk assessment model is obtained by using the training data set screened by feature engineering to be trained by a machine learning algorithm.

In this embodiment, it should be noted that a risk assessment model corresponding to common insurance types on the market is pre-trained on the server. When the server obtains the current insurance type, the corresponding risk assessment model can be directly determined according to the keyword of the insurance type. Different types of risk assessment models are trained with corresponding types of sample data. The server can collect a large amount of real sample data of related businesses such as insurance business from various channels, add labels to it manually or in other ways, and then use feature engineering to filter out a more effective part of the sample data as the final Finally, use machine learning algorithms, such as decision trees, clustering, deep neural networks, or XGBoost and other model algorithms to train the selected sample data, and finally obtain a trained risk assessment model.

The server takes the insurance application information to be evaluated as the input of the model, and the model first extracts the text information and digital information filled in by the user in the insurance application information to be evaluated. Text information such as family medical history, place of residence, etc.; digital information such as age, income, etc. The model can encode text information to make it a discrete numerical feature (text feature), and then perform some numerical validity processing on the digital information to obtain digital features. After the model obtains the text features and digital features, the features with high correlation are combined, and the combined features and the original features are combined to obtain the final target features.

Step S30: Determine the risk level of the insurance application information to be evaluated according to the target feature, and obtain a quota adjustment factor of the insurance application information to be evaluated when it is detected that the risk level is lower than high risk.

In this embodiment, a scoring criterion for multi-dimensional risk control indicators is set in the preset risk assessment model. Risk levels can be divided according to actual needs, such as low risk level, medium risk level and high risk level. Very high risk level, which is not limited in this embodiment. The quota adjustment factor may be related adjustment parameters, regional development levels, corresponding upward adjustment strategies for lower risk levels, and/or salesperson quality, etc. The risk level determination method can be that the model obtains the risk score corresponding to each risk control index of the current insurance application information to be evaluated, and then determines the specific risk level according to the interval in which the risk score falls.

Specifically, the model maps the currently obtained target characteristics to each risk control index, evaluates the questionnaire data based on each risk control index, and outputs the final risk level of the insured user this time. The server obtains the quota adjustment factor corresponding to the sub-questionnaire data when the risk level of the current insured user is low risk, medium risk and other non-high risk levels and above.

Step S40, combining the previous insurance application information and the limit adjustment factor, predict the insurance addition limit information of the target insurance applicant.

In this embodiment, the added insurance limit information may include the name of the insurance application, the code number, the range of the added insurance limit, and the like. The calculation method of the insurable amount can be obtained by first obtaining the original insurable amount corresponding to the user, then subtracting the insured amount from the original insurable amount, that is, the previous insured amount above, to obtain the remaining insurable amount, and finally adjusting the amount. The factor continues to adjust the remaining insurable amount, and the final value obtained is the maximum insurable amount, and the insured amount range is from 0 to the maximum insurable amount. In addition, after S30, the method may further include: generating a prompt message including the information on the added insurance amount, and sending the prompt message to the user terminal of the target insurance applicant. After the server calculates the reinsurable limit of the insured user's insurance this time, a notification message can be generated separately or displayed synchronously on the self-check result feedback interface, so that the insured user can know the current reinsurable limit.

Specifically, if the user purchased life insurance this time, the risk level output by the model is a very low risk level. The server can first obtain the user's standard maximum life insurance amount in life insurance according to general standards, and then adjust the standard maximum insurance amount according to the regional examination-free standard in the region where the user is located to obtain the user's maximum life insurance amount applicable to the region. Subtract the previous life insurance coverage to get the remaining life insurance coverage that can be insured. After the server multiplies the remaining amount of life insurance that can be insured by the preset adjustment coefficient corresponding to life insurance, and then adds the upward limit of the quality level of the salesperson corresponding to the insurance policy for this order and the upward limit corresponding to the extremely low risk, the final result is obtained. The limit value is the maximum limit of the user's current life insurance coverage.

In this embodiment, the application obtains the insurance application information to be evaluated of the target insurance applicant, and obtains the previous insurance application information and the current insurance application type of the target insurance applicant according to the to-be-evaluated insurance application information; A pre-trained risk assessment model corresponding to the type, using the risk assessment model to extract text features and numerical features in the insurance application information to be assessed, and combining the text features and numerical features to obtain target features, wherein the risk The evaluation model is obtained by using the training data set screened by feature engineering to be trained by a machine learning algorithm; the risk level of the insurance application information to be evaluated is determined according to the target feature, and when it is detected that the risk level is a level below high risk, Obtain the limit adjustment factor of the insurance application information to be evaluated; combine the previous insurance application information and the limit adjustment factor to predict the insurance addition limit information of the target insurance applicant. Through the above method, the application obtains the customer's insurance application information by obtaining their previous insurance application information, uses the pre-trained risk assessment model to determine the corresponding risk level, and only detects that the risk level is within a certain range (below the high risk level) The follow-up insurance limit estimation is carried out only at the time of day, which avoids the negative impact that may be caused by customers with higher risk levels continuing to apply for insurance. The risk assessment model is obtained by using feature engineering and machine learning training, and the target characteristics of the insurance information to be assessed are obtained by using the risk assessment model, so that the final risk assessment result obtained by the risk assessment model is more accurate; by synthesizing the previous insurance information It can intelligently calculate the insurance amount that the customer can currently insure with the amount adjustment factor, so that the insurable amount information that is consistent with the actual situation can be obtained when the insurance is applied, thus solving the problem of predicting the insurance amount when adding insurance. technical issues with low accuracy. At the same time, it also saves the customer from being disturbed by the insurance salesman who does not know the amount of insurance that can be added to him, such as physical examination and coordination.

Further, based on the first embodiment shown in FIG. 2 above, a second embodiment of the quota prediction method of the present application is proposed. In this embodiment, step S20 includes: based on the risk assessment model, acquiring text information and numerical information in the insurance application information to be evaluated; processing the text information by one-hot encoding to obtain the text features, and converting the text information into The digital information is subjected to missing value processing and dense processing to obtain the digital feature; the mutual information value of the text feature and the digital feature is obtained, and the text feature and the digital feature are divided into two groups based on the mutual information value. A first feature that can be combined and a second feature that cannot be combined; a combined feature is obtained by combining the first feature, and the combined feature and the second feature are used as the target feature.

In this embodiment, one-hot encoding is One-Hot-coding, also known as one-bit effective encoding. The method is to use an N-bit state register to encode N states, and each state has its independent register bit. And at any time, only one of them is valid. The model performs one-hot encoding on the text information extracted from the insurance application information to be evaluated to convert it into discrete numerical features, and it can continue to be densely processed to further reduce the amount of data. The model then performs missing value processing on the digital information extracted from the insurance information to be evaluated, such as special value filling, average filling, hot card filling, and expected value maximization. Then the model is densely processed, such as compressing sparse rows and columns, using Principal Component Analysis (PCA, Principal Components Analysis), Singular Value Decomposition (SVD, Singular Value Decomposition) and other methods for dimensionality reduction. The model needs to calculate the mutual information value between the features, and the mutual information value represents the degree of correlation between the features, and the specific calculation method can refer to the prior art, which will not be repeated here. After the model calculates the mutual information value between the features, it can be compared with the preset standard mutual information threshold, and the features exceeding the threshold can be combined to obtain the combined feature, and finally the combined feature and the uncombined feature can be combined. common as the target feature.

Further, when it is detected that the risk level is below high risk, the step of acquiring the quota adjustment factor of the insurance application information to be assessed includes: when the risk level is below high risk, judging the target Whether the insured has passed the self-check; if the target insured has passed the self-check, obtain the regional medical examination-exemption standard, the insurance salesperson level and the insurance type adjustment coefficient of the insurance application type corresponding to the insurance application information to be evaluated as the Quota adjustment factor.

In this embodiment, for the regional medical examination exemption standard, the server can access the relevant government platform in the region to search, or use the specific extracted name as a keyword to search in the database storing the information on the medical examination exemption standard in each region, so as to Obtain the latest information on local exemption standards for medical examinations; for insurance salesperson levels, the server can directly access the information database within the insurance company, and search for the corresponding level information according to the specific insurance salesperson's name, employee number and other information; for insurance type adjustment The server can also directly access the internal information database of the insurance company, and find the corresponding insurance type adjustment coefficient in the database according to the name of the specific insurance type purchased this time.

When the server detects that the currently obtained risk level is not high risk or above, it further determines whether the insured user has passed the self-check. If the insured user passes the self-check, the server will determine the highest medical-exemption insurance amount in the region according to the region information filled in by the user in the questionnaire, the quality level of the salesperson responsible for the insurance application for this order, and the type of insurance purchased this time; If the user fails to pass the self-check, it will enter the regular insurance application process, and the additional insurance amount will not be automatically calculated and displayed for the insured user.

Further, step S30 includes: determining the maximum medical-exemption insurance amount according to the regional medical-exemption standard, and obtaining an initial insurance increase amount in combination with the previous insurance application information; using the insurance type adjustment coefficient to carry out the initial insurance increase amount. Adjustment to obtain the adjusted insurance amount; obtain the level adjustment amount corresponding to the insurance salesperson's level, and use the level adjustment amount to adjust the adjusted insurance amount to obtain the target insurance applicant's insurance amount information , wherein the information on the added insurance amount of the target insured is stored in the blockchain.

In this embodiment, the following describes the calculation formula of the insurable amount of life insurance and the calculation formula of the insurable amount of critical illness insurance as examples.

For the calculation of the insured amount of life insurance that can be added, the server needs to call the highest medical examination-free life insurance insurance amount applicable to the insured customer in the region where the user is located. The sum of the sum insured for the physical examination of the life insurance is called C, and the calculation formula for the additional insured amount of the initial life insurance can be set as (A-C)*X1, where X1 is the life insurance coefficient, which can be flexibly set according to actual needs. For the calculation of the insured amount of critical illness insurance, the server needs to call the highest medical-exemption critical illness insurance amount applicable to the insured customer in the area where the user is located. The sum of the historical risk insured amount and the critical illness medical insurance amount is called D. The calculation formula of the initial life insurance reinsurable amount can be set as life insurance reinsurable amount = (B-D)*X2, where X2 is the critical illness insurance The coefficient can be set flexibly according to actual needs, which can be set to be the same as X1 or different. After the sum of the initial reinsurable amount is calculated, the initial value can be adjusted slightly according to the level of the salesperson responsible for the insurance business of this order to obtain the final reinsurable amount. Relevant personnel can pre-set different increase quotas corresponding to different salesperson levels on the server.

It should be emphasized that, in order to further ensure the privacy and security of the above-mentioned target applicant's insurance coverage information, the above-mentioned target insurance applicant's insurance coverage information can also be stored in a blockchain node.

Further, step S10 includes: when an insurance application instruction is received, acquiring an insurance application questionnaire of the target insurance applicant based on the insurance application instruction as the insurance application information to be evaluated; acquiring the identity filled in by the target insurance applicant in the insurance application questionnaire Information and insurance name information, determine the type of insurance this time according to the insurance name information, and determine whether the target insurance applicant is an insured user according to the identity information; The past insurance application information of the target insurance applicant is found in the information database.

In this embodiment, an insurance application user usually fills in a user insurance application questionnaire on a personal terminal, a counter terminal of an insurance company, or other terminal equipment. The questionnaire may contain questions about the user's personal information and questions about the setting of risk control indicators set by the insurance company according to actual needs. The identity information may specifically include name, age, ID number, place of residence, etc. After the insurance application user completes and submits the questionnaire, the server receives an insurance application instruction sent by the insurance application user, obtains the questionnaire information filled in by the user pointed to in the insurance application instruction, and uses it as the above-mentioned insurance application information to be evaluated.

Specifically, after receiving the user's insurance application questionnaire data sent by the user's questionnaire filling terminal, the server extracts the question option data submitted by each questionnaire question user, and extracts the information that can indicate the identity of the insurance applicant as a retrieval keyword in the insurance application information database. Query, inquire whether there is a previous insurance application record of the insured user in the database, and obtain the record when there is a previous insurance application record in the information database. At the same time, the server can extract the name of the insurance that the user has filled in or selected in the questionnaire, and use it as the insurance type of this insurance. In addition, it can also be converted into a unique number for storage for subsequent data processing.

Further, after the step of judging whether the target insured is an insured user according to the identity information, the step further includes: if the target insured is not an insured user, executing the determination according to the target feature. The risk level of the insurance application information to be evaluated, and when it is detected that the risk level is lower than the high risk level, the step of obtaining the quota adjustment factor of the insurance application information to be evaluated; information on the added insurance amount of the target policyholder.

In this embodiment, if the server does not query the previous insurance application information of the current insurance application user in the insurance application information database, that is, the current insurance application user is a new insurance application user, it continues to input the question answer information in the insurance application questionnaire of the new insurance application user. The risk assessment model determines the risk level of the current new insured users through the risk assessment model. If the server determines through the model that the risk level of the current new insured user is below a high risk level, such as a medium risk level, a low risk level, or a very low risk level, the quota adjustment factor corresponding to the questionnaire information is obtained. The server calculates the reinsurable limit for newly insured users. It only needs to be based on the adjustment factor of the amount, and does not need to be calculated based on the previous insurance information. Taking life insurance as an example, the specific calculation method can be as follows: multiply the maximum amount of medical examination-free life insurance for new insured users by the adjustment coefficient of life insurance, and add the corresponding increase amount corresponding to the level of the salesperson responsible for this insurance, and finally get The result is the life insurance reinsurable limit of the new insured user. If the server determines through the model that the risk level of the current new insured user is high risk level, extremely high risk level, or higher, the subsequent steps will not be performed, and the routine self-checking process will be turned to.

Further, by extracting the characteristics of the insurance application information to be evaluated and combining the characteristics, it is possible to deeply mine the effective information in the insurance application information to be evaluated, thereby improving the accuracy of the model evaluation results; Real-time access to the insured amount space of insurable insurance based on multi-dimensional calculation, helping agents to expand their business, increasing the insured amount of customers, and meeting customer protection needs.

Further, based on the first embodiment shown in FIG. 2 above, a third embodiment of the quota prediction method of the present application is proposed. In this embodiment, the step of determining the risk level of the insurance application information to be evaluated based on the preset risk evaluation model includes: calculating a risk score corresponding to the target feature on the preset risk control index; The rule determines the risk level corresponding to the risk score as the risk level of the insurance application information to be evaluated.

In this embodiment, the related issues of the risk control indicators may include related indicators of the diseases involved in the insurance product insured by the user this time, and the like. For example, physical health index, dietary health, exercise, smoking, drinking, stress and family medical history, etc., can also be flexibly set according to specific risk control needs. Each questionnaire question can be set with multiple answer options, and different answer options correspond to corresponding option scores.

The model maps the currently obtained multiple target features to each risk control index one by one, and then marks the risk score of each target feature on the risk control index according to the preset corresponding rules, and then sets the corresponding risk control index according to each risk control index. The weight coefficient of the insured user is calculated to calculate the total risk score of the insured user this time. After the model calculates and obtains the final total risk score, it can locate the risk level of the risk score according to the preset score level threshold. For example, a score of 0-20 is set as a very low risk level; a score of 21-40 is set as a low risk level; a score of 41-60 is set as a medium risk level; a score of 61-80 is set as a high risk level; 81- A score of 100 is set as a very high risk level.

Further, before step S20, it also includes: collecting an initial sample data set including multiple pieces of sample data, wherein each piece of sample data includes sample features and corresponding sample labels; constructing a feature index according to the sample features and the sample labels set, and calculate the highest discrimination accuracy rate of the feature index set; compare the highest discrimination accuracy rate with the preset standard accuracy rate threshold, and compare the sample corresponding to the highest discrimination accuracy rate not less than the standard accuracy rate threshold The data is used as target sample data to filter out the target sample data set; the target sample data set is trained by using a preset machine learning model to obtain the risk assessment model.

In this embodiment, the feature engineering may be the maximum correlation minimum redundancy combined with the maximum mutual information coefficient feature selection strategy. The sample label is used to indicate the user's risk score. The server first calculates the maximum mutual information coefficient between the sample feature in each piece of sample data and its sample label, then constructs a feature index set, and calculates the discrimination accuracy rate of one feature index set per day, and selects the item with the highest value and the Preset standard accuracy thresholds for comparison. The server includes the sample data corresponding to the discrimination accuracy rate greater than or equal to the threshold into the target sample data set. Finally, the server uses decision tree, clustering, deep neural network, or XGBoost and other model algorithms to train it, and finally trains to obtain the final risk assessment model.

The blockchain referred to in this application is a new application mode of computer technologies such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm. Blockchain, essentially a decentralized database, is a series of data blocks associated with cryptographic methods. Each data block contains a batch of network transaction information to verify its Validity of information (anti-counterfeiting) and generation of the next block. The blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.

Further, by presetting risk control indicators to quantify the risk scores of insured customers, and then determine their corresponding risk levels, so that the model can quickly assess the risk levels of insured customers; , which reduces the data processing burden of the device and improves the efficiency and accuracy of model training.

In addition, as shown in FIG. 3 , in order to achieve the above purpose, the present application also provides a limit prediction device, the limit prediction device includes: a past information acquisition module 10 for obtaining the insurance application information to be evaluated of the target insured, and according to The to-be-evaluated insurance application information obtains the target insurance applicant's previous insurance application information and the current insurance application type; the target feature acquisition module 20 is used to select a pre-trained risk assessment model corresponding to the current insurance application type, using the The risk assessment model extracts the text features and numerical features in the insurance application information to be assessed, and combines the text features and numerical features to obtain the target features, wherein the risk assessment model uses the training data set screened by feature engineering to pass through. Obtained from machine learning algorithm training; the risk level determination module 30 is configured to determine the risk level of the insurance application information to be evaluated according to the target feature, and obtain the to-be-evaluated risk level when it is detected that the risk level is below high risk The amount adjustment factor of the insurance application information; the insurance addition amount prediction module 40 is configured to combine the previous insurance application information and the amount adjustment factor to predict the insurance addition amount information of the target insurance applicant.

The present application also provides a quota prediction device.

The credit prediction device includes a processor, a memory, and a credit prediction program stored on the memory and executable on the processor, wherein the credit prediction program, when executed by the processor, implements the above-mentioned The steps of the quota prediction method.

For the method implemented when the quota prediction program is executed, reference may be made to the various embodiments of the quota prediction method of the present application, which will not be repeated here.

In addition, the embodiments of the present application further provide a computer-readable storage medium.

A quota prediction program is stored on the computer-readable storage medium of the present application, wherein when the quota prediction program is executed by the processor, the steps of the above quota prediction method are implemented.

Optionally, the storage medium involved in this application, such as a computer-readable storage medium, may be non-volatile or volatile.

It should be noted that, herein, the terms "comprising", "comprising" or any other variation thereof are intended to encompass non-exclusive inclusion, such that a process, method, article or system comprising a series of elements includes not only those elements, It also includes other elements not expressly listed or inherent to such a process, method, article or system. Without further limitation, an element qualified by the phrase "comprising a..." does not preclude the presence of additional identical elements in the process, method, article or system that includes the element.

The above-mentioned serial numbers of the embodiments of the present application are only for description, and do not represent the advantages or disadvantages of the embodiments.

From the description of the above embodiments, those skilled in the art can clearly understand that the method of the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is better implementation. Based on this understanding, the technical solutions of the present application can be embodied in the form of software products in essence or the parts that make contributions to the prior art. The computer software products are stored in a storage medium (such as ROM/RAM) as described above. , magnetic disk, optical disk), including several instructions to make a terminal device (which may be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) execute the methods described in the various embodiments of the present application.

The above are only the preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made by using the contents of the description and drawings of the present application, or directly or indirectly applied in other related technical fields , are similarly included within the scope of patent protection of this application.

Claims

A quota prediction method, wherein the quota prediction method includes:

Obtaining the insurance application information to be evaluated of the target insured, and obtaining the previous insurance application information and the current insurance application type of the target insurance applicant according to the to-be-evaluated insurance application information;

Selecting the pre-trained risk assessment model corresponding to the type of insurance this time, using the risk assessment model to extract the text features and numerical features in the insurance application information to be assessed, and combining the text features and numerical features to obtain a target feature, wherein the risk assessment model is obtained by using the training data set screened by feature engineering to be trained by a machine learning algorithm;

Determine the risk level of the insurance application information to be evaluated according to the target feature, and obtain the quota adjustment factor of the insurance application information to be evaluated when it is detected that the risk level is lower than high risk;

Combined with the previous insurance application information and the limit adjustment factor, the insurance addition limit information of the target insurance applicant is predicted.
The limit prediction method according to claim 1, wherein the use of the risk assessment model to extract text features and numerical features in the insurance application information to be assessed, and combining the text features and numerical features to obtain a Steps include:

Based on the risk assessment model, obtain text information and numerical information in the insurance application information to be assessed;

The text information is processed by one-hot encoding to obtain the text features, and the digital information is processed with missing values and dense to obtain the digital features;

Obtaining the mutual information value of the text feature and the digital feature, and dividing the text feature and the digital feature into a combinable first feature and an uncombinable second feature based on the mutual information value;

The first feature is combined to obtain a combined feature, and the combined feature and the second feature are used as the target feature.
The limit prediction method according to claim 1, wherein, when it is detected that the risk level is lower than high risk, the step of acquiring the limit adjustment factor of the insurance application information to be evaluated comprises:

When the risk level is below the high risk level, determine whether the target insured has passed the self-check;

If the target insured passes the self-check, the regional medical exemption standard corresponding to the insurance application information to be evaluated, the insurance salesperson level, and the insurance type adjustment coefficient of the current insurance application type are obtained as the quota adjustment factor.
The limit prediction method according to claim 3, wherein the step of predicting the added insurance limit information of the target insured by combining the previous insurance application information and the limit adjustment factor comprises:

Determine the maximum medical-exemption insurance amount according to the regional medical-exemption standard, and obtain the initial additional insurance amount in combination with the previous insurance information;

Adjusting the initial insurance coverage amount using the insurance type adjustment coefficient to obtain an adjusted insurance coverage amount;

Obtain the level adjustment limit corresponding to the level of the insurance clerk, and use the level adjustment limit to adjust the adjustment plus insurance limit, and obtain the information on the increase insurance limit of the target insurance applicant, wherein the target insurance applicant The added insurance amount information is stored in the blockchain.
The quota prediction method according to claim 1, wherein the step of obtaining the insurance application information to be evaluated of the target insurance applicant, and obtaining the previous insurance application information and the current insurance application type of the target insurance applicant according to the to-be-evaluated insurance application information include:

When receiving the insurance application instruction, obtain the insurance application questionnaire of the target applicant based on the insurance application instruction as the insurance application information to be evaluated;

Obtain the identity information and insurance name information filled in by the target insurance applicant in the insurance application questionnaire, and determine the current insurance application type according to the insurance application name information;

Determine whether the target insured is an insured user according to the identity information;

If the target insurance applicant is an insured user, the previous insurance application information of the target insurance applicant is found in the insurance application information database.
The quota prediction method according to claim 5, wherein after the step of judging whether the target insured is an insured user according to the identity information, the method further comprises:

If the target insured is not an insured user, the risk level of the insurance application information to be evaluated is determined according to the target feature, and when it is detected that the risk level is lower than high risk, the to-be-evaluated information is obtained. The steps of adjusting the amount of insurance information;

Based on the limit adjustment factor, information on the added insurance limit of the target insured who is not an insured user is obtained.
The limit prediction method according to claim 1, wherein the step of determining the risk level of the insurance application information to be evaluated according to the target characteristic comprises:

calculating the risk score corresponding to the target feature on the preset risk control index;

The risk level corresponding to the risk score is determined according to a preset score level correspondence rule, as the risk level of the insurance application information to be evaluated.
The quota prediction method according to any one of claims 1-7, wherein before the step of selecting a pre-trained risk assessment model corresponding to the current insurance application type, the method further comprises:

Collect an initial sample data set containing multiple pieces of sample data, wherein each piece of sample data includes sample features and corresponding sample labels;

Build a feature index set according to the sample feature and the sample label, and calculate the highest discrimination accuracy rate of the feature index set;

Comparing the highest discrimination accuracy with a preset standard accuracy threshold, and using the sample data corresponding to the highest discrimination accuracy not less than the standard accuracy threshold as the target sample data, to filter out the target sample data set;

The target sample data set is trained using a preset machine learning model to obtain the risk assessment model.
A credit prediction device, wherein the credit prediction device includes a processor, a memory, and a credit prediction program stored on the memory and executable by the processor, wherein the credit prediction program is executed by the processor When executed, implement the following methods:

Obtaining the insurance application information to be evaluated of the target insured, and obtaining the previous insurance application information and the current insurance application type of the target insurance applicant according to the to-be-evaluated insurance application information;

Selecting the pre-trained risk assessment model corresponding to the type of insurance this time, using the risk assessment model to extract the text features and numerical features in the insurance application information to be assessed, and combining the text features and numerical features to obtain a target feature, wherein the risk assessment model is obtained by using the training data set screened by feature engineering to be trained by a machine learning algorithm;

Determine the risk level of the insurance application information to be evaluated according to the target feature, and obtain the quota adjustment factor of the insurance application information to be evaluated when it is detected that the risk level is lower than high risk;

Combined with the previous insurance application information and the limit adjustment factor, the insurance addition limit information of the target insurance applicant is predicted.
The limit prediction device according to claim 9, wherein the extraction of text features and numerical features in the insurance application information to be evaluated by using the risk assessment model is performed, and the text features and numerical features are combined to obtain the target feature The steps include:

Based on the risk assessment model, obtain text information and numerical information in the insurance application information to be assessed;

The text information is processed by one-hot encoding to obtain the text features, and the digital information is processed with missing values and dense to obtain the digital features;

Obtaining the mutual information value of the text feature and the digital feature, and dividing the text feature and the digital feature into a combinable first feature and an uncombinable second feature based on the mutual information value;

The first feature is combined to obtain a combined feature, and the combined feature and the second feature are used as the target feature.
The limit prediction device according to claim 9, wherein the step of obtaining the limit adjustment factor of the insurance application information to be evaluated when it is detected that the risk level is lower than the high risk level comprises:

When the risk level is below the high risk level, determine whether the target insured has passed the self-check;

If the target insured passes the self-check, the regional medical exemption standard corresponding to the insurance application information to be evaluated, the insurance salesperson level, and the insurance type adjustment coefficient of the current insurance application type are obtained as the quota adjustment factor.
The limit prediction device according to claim 11, wherein the step of combining the previous insurance application information and the limit adjustment factor to predict the insurance addition limit information of the target insurance applicant comprises:

Determine the maximum medical-exemption insurance amount according to the regional medical-exemption standard, and obtain the initial additional insurance amount in combination with the previous insurance information;

Using the insurance type adjustment coefficient to adjust the initial insurance coverage amount to obtain an adjusted insurance coverage amount;

Obtain the level adjustment limit corresponding to the level of the insurance clerk, and use the level adjustment limit to adjust the adjustment plus insurance limit, and obtain the information on the increase insurance limit of the target insurance applicant, wherein the target insurance applicant The insurance amount information is stored in the blockchain.
The limit prediction device according to claim 9, wherein the obtaining of the insurance application information to be assessed of the target insurance applicant is performed, and the previous insurance application information of the target insurance applicant and the information of the current insurance application type are obtained according to the to-be-evaluated insurance application information. Steps include:

When receiving the insurance application instruction, obtain the insurance application questionnaire of the target applicant based on the insurance application instruction as the insurance application information to be evaluated;

Obtain the identity information and insurance name information filled in by the target insurance applicant in the insurance application questionnaire, and determine the current insurance application type according to the insurance application name information;

Determine whether the target insured is an insured user according to the identity information;

If the target insurance applicant is an insured user, the previous insurance application information of the target insurance applicant is found in the insurance application information database.
The limit prediction device according to any one of claims 9-13, wherein the limit prediction program is processed by the process before the step of selecting a pre-trained risk assessment model corresponding to the current insurance application type It is also used to implement:

Collect an initial sample data set containing multiple pieces of sample data, wherein each piece of sample data includes sample features and corresponding sample labels;

Build a feature index set according to the sample feature and the sample label, and calculate the highest discrimination accuracy rate of the feature index set;

Comparing the highest discrimination accuracy with a preset standard accuracy threshold, and using the sample data corresponding to the highest discrimination accuracy not less than the standard accuracy threshold as the target sample data, to filter out the target sample data set;

The target sample data set is trained using a preset machine learning model to obtain the risk assessment model.
A computer-readable storage medium, wherein a quota prediction program is stored on the computer-readable storage medium, wherein when the quota prediction program is executed by a processor, the following method is implemented:

Obtain the insurance application information to be evaluated of the target insurance applicant, and obtain the previous insurance application information and the current insurance application type of the target insurance applicant according to the insurance application information to be evaluated;

Selecting a pre-trained risk assessment model corresponding to the type of insurance this time, using the risk assessment model to extract text features and numerical features in the insurance application information to be assessed, and combining the text features and numerical features to obtain a target feature, wherein the risk assessment model is obtained by using the training data set screened by feature engineering to be trained by a machine learning algorithm;

Determine the risk level of the insurance application information to be evaluated according to the target feature, and obtain the quota adjustment factor of the insurance application information to be evaluated when it is detected that the risk level is lower than high risk;

Combined with the previous insurance application information and the limit adjustment factor, the insurance addition limit information of the target insurance applicant is predicted.
The computer-readable storage medium according to claim 15, wherein the extraction of text features and numerical features in the insurance application information to be assessed by using the risk assessment model is performed, and the text features and numerical features are combined to obtain The steps to target features include:

Based on the risk assessment model, obtain text information and numerical information in the insurance application information to be assessed;

The text information is processed by one-hot encoding to obtain the text features, and the digital information is processed with missing values and dense to obtain the digital features;

Obtaining the mutual information value of the text feature and the digital feature, and dividing the text feature and the digital feature into a combinable first feature and an uncombinable second feature based on the mutual information value;

The first feature is combined to obtain a combined feature, and the combined feature and the second feature are used as the target feature.
The computer-readable storage medium according to claim 15, wherein the step of obtaining the limit adjustment factor of the insurance application information to be evaluated when it is detected that the risk level is lower than the high risk level comprises:

When the risk level is below high risk, determine whether the target insured has passed the self-check;

If the target insured passes the self-check, the regional medical exemption standard corresponding to the insurance application information to be evaluated, the insurance salesperson level, and the insurance type adjustment coefficient of the current insurance application type are obtained as the quota adjustment factor.
The computer-readable storage medium of claim 17, wherein the step of combining the previous insurance application information and the amount adjustment factor to predict the insurance addition amount information of the target applicant comprises:

Determine the maximum medical-exemption insurance amount according to the regional medical-exemption standard, and obtain the initial additional insurance amount in combination with the previous insurance information;

Using the insurance type adjustment coefficient to adjust the initial insurance coverage amount to obtain an adjusted insurance coverage amount;

Obtain the level adjustment limit corresponding to the level of the insurance clerk, and use the level adjustment limit to adjust the adjustment plus insurance limit, and obtain the information on the increase insurance limit of the target insurance applicant, wherein the target insurance applicant The insurance amount information is stored in the blockchain.
The computer-readable storage medium according to claim 15, wherein the obtaining of the insurance application information to be evaluated of the target insurance applicant is executed, and the previous insurance application information and the current insurance application of the target insurance applicant are obtained according to the to-be-evaluated insurance application information Types of steps include:

When receiving the insurance application instruction, obtain the insurance application questionnaire of the target applicant based on the insurance application instruction as the insurance application information to be evaluated;

Obtain the identity information and insurance name information filled in by the target insurance applicant in the insurance application questionnaire, and determine the current insurance application type according to the insurance application name information;

Determine whether the target insured is an insured user according to the identity information;

If the target insurance applicant is an insured user, the previous insurance application information of the target insurance applicant is found in the insurance application information database.
The computer-readable storage medium of any one of claims 15-19, wherein the quota prediction program is processed before the step of selecting a pre-trained risk assessment model corresponding to the current insurance type It is also used to implement:

Collect an initial sample data set containing multiple pieces of sample data, wherein each piece of sample data includes sample features and corresponding sample labels;

Build a feature index set according to the sample feature and the sample label, and calculate the highest discrimination accuracy rate of the feature index set;

Comparing the highest discrimination accuracy with a preset standard accuracy threshold, and using the sample data corresponding to the highest discrimination accuracy not less than the standard accuracy threshold as the target sample data, to filter out the target sample data set;

The target sample data set is trained using a preset machine learning model to obtain the risk assessment model.