Insurance pricing method and device based on strategy learning, electronic equipment and medium

Abstract

The embodiment of the invention provides an insurance pricing method and device based on strategy learning, electronic equipment and a medium, and belongs to the technical field of financial science and technology. The method comprises the following steps: acquiring behavior data of a user, and acquiring target behavior data of the behavior data in a preset time period; determining an insurance premium initial pricing according to the data category of the target behavior data; performing feature extraction on the target behavior data through a multi-view learning model to obtain a feature vector; performing pricing prediction on the premium initial pricing and the feature vector through a strategy network; according to the pricing ratio, the premium variance and the premium initial pricing obtained by the pricing prediction, determining an estimated current pricing; inputting the estimated current pricing, the initial premium pricing and the historical feature vector of the previous time step of the current time step into a preset value network for pricing evaluation, and outputting an evaluated value; and adjusting and estimating the current price according to the evaluation value. According to the embodiment of the invention, the pricing can be dynamically adjusted according to the behavior data of the user.

Description

Technical Field

[0001] The present application relates to the field of financial technology, and in particular to an insurance pricing method, device, electronic device and medium based on policy learning. Background Art

[0002] Currently, most pricing and risk management systems in the insurance industry rely on historical data and use predefined models and rules to assess risks and determine premiums. These systems lack the ability to adjust dynamically in real time and are difficult to respond to changes in customer behavior. For example, auto insurance pricing is usually based on a customer's historical driving record, and health insurance relies on a customer's health history. Although these methods can assess risks to a certain extent, they cannot motivate customers to maintain good behavior (such as safe driving and healthy living) during the insurance cycle, and they are difficult to cope with dynamic changes in customer behavior.

[0003] Because the system cannot reflect the latest changes in customer behavior in real time, risk assessments may quickly become outdated, resulting in the inability of insurance companies to accurately predict and respond to emerging risks. At the same time, due to the lack of sensitivity to changes in customer behavior, insurance products may not be updated in a timely manner to cover new or changing risks, resulting in insufficient insurance coverage and further reducing the customer experience. Summary of the invention

[0004] The main purpose of the embodiments of the present application is to propose an insurance pricing method, device, electronic device and medium based on policy learning, which can realize dynamic adjustment of pricing according to user behavior data.

[0005] To achieve the above objectives, a first aspect of an embodiment of the present application proposes an insurance pricing method based on policy learning, the method comprising: Acquire user behavior data, and acquire target behavior data of the behavior data within a preset time period, wherein the behavior data includes data of multiple time steps; Determining initial premium pricing according to the data category of the target behavior data; Extract features from the target behavior data using a preset multi-view learning model to obtain a feature vector for the current time step; Performing pricing prediction on the initial premium pricing and the feature vector through a preset strategy network to obtain a pricing ratio and premium variance; determining an estimated current pricing based on the pricing ratio, the premium variance, and the initial premium pricing; Input the estimated current pricing, the initial premium pricing, and the historical feature vector of the previous time step of the current time step into a preset value network for pricing evaluation, and output an evaluation value corresponding to the estimated current pricing; The estimated current pricing corresponding to the data category is adjusted according to the assessed value.

[0006] In some embodiments, the multi-view learning model includes a convolutional neural network and a long short-term memory network; the feature extraction of the target behavior data by the preset multi-view learning model to obtain the feature vector of the current time step includes: Inputting the target behavior data into a preset multi-view learning model, so that the convolutional neural network extracts features from the target behavior data and outputs a first feature vector, and the long short-term memory network extracts features from the target behavior data and outputs a second feature vector; Feature concatenation is performed on the first feature vector and the second feature vector to obtain a feature vector of the data category at the current time step.

[0007] In some embodiments, determining the estimated current pricing according to the pricing ratio, the premium variance, and the initial premium pricing includes: Obtaining a normal probability distribution according to the pricing ratio, the premium variance, and the initial premium pricing; A sampling operation is performed on the normal probability distribution to obtain an estimated current price.

[0008] In some embodiments, the initial premium pricing and the feature vector are predicted by a preset policy network to obtain a pricing ratio and a premium variance, including: Obtaining the current pricing ratio of the data category at the current time step, wherein the estimated current pricing ratio is used to represent the ratio of the current pricing of the data category at the current time step to the initial pricing of the premium; The initial premium pricing, the feature vector and the current pricing ratio are input into a preset strategy network, so that the strategy network concatenates the initial premium pricing, the feature vector and the current pricing ratio, and outputs the pricing ratio and premium variance.

[0009] In some embodiments, the step of inputting the estimated current pricing, the initial premium pricing, and the historical feature vector of the previous time step of the current time step into a preset value network for pricing evaluation, and outputting the evaluation value corresponding to the estimated current pricing includes: Inputting the estimated current price, the initial premium price and the historical feature vector of the previous time step of the current time step into a preset value network, so that the value network splices the estimated current price, the initial premium price and the historical feature vector and outputs a spliced ​​vector; The concatenated vector is input into a fully connected layer of the value network, so that the fully connected layer scores the estimated current pricing according to the concatenated vector and outputs an assessed value.

[0010] In some embodiments, adjusting the estimated current price corresponding to the data category according to the assessed value includes: When the evaluation value is greater than or equal to a first preset threshold, increasing the estimated current price corresponding to the data category based on the evaluation value; When the evaluation value is less than a preset second preset threshold, reducing the estimated current price corresponding to the data category based on the evaluation value; Wherein, the first preset threshold is greater than the second preset threshold.

[0011] In some embodiments, after inputting the estimated current pricing, the initial premium pricing, and the historical feature vector of the previous time step of the current time step into a preset value network for pricing evaluation and outputting the evaluation value corresponding to the estimated current pricing, the method further includes: Calculating the reward value of the target behavior data at different time steps; Training the value network according to the reward value and the evaluation value to obtain a pre-trained value network; The calculating the reward value of the target behavior data at different time steps includes: Performing status detection on the target behavior data; When the target behavior data completes state detection and detects that the target behavior data has a preset behavior, determine the first time step at which the preset behavior occurs and the first pricing corresponding to the first time step, obtain the compensation amount and the first profit margin corresponding to the preset behavior at the first time step, and obtain the first reward value through the compensation amount, the first profit margin and the first pricing; When the target behavior data completes state detection and no preset behavior is detected in the target behavior data, a second time step at the last moment within the preset time period and a second profit margin corresponding to the second time step are determined, and a second price corresponding to the second time step are obtained, and a second reward value is obtained through the second profit margin and the second price; When the target behavior data has not completed the state detection, the preset value is used as the third reward value.

[0012] To achieve the above-mentioned purpose, a second aspect of an embodiment of the present application proposes an insurance pricing device based on policy learning, wherein the classification model includes a plurality of classification sub-models; the device includes: A data acquisition module, used to acquire an insurance product set of a financial system and user behavior data of the financial system within a preset time period, wherein the insurance product set includes a plurality of insurance products; A data generation module, used to generate data to be predicted based on the insurance product set and the user behavior data; A feature processing module, used for performing feature processing on the data to be predicted to obtain a data set to be predicted; A model training module, used for training each of the classification sub-models according to the data set to be predicted to obtain a prediction set, wherein the prediction set includes prediction scores corresponding to the classification sub-models; An index fusion module, used to perform index fusion on the prediction scores in the prediction set to obtain a product score; The product recommendation module is used to sort the insurance products in the insurance product set based on the product scores, obtain a recommendation list, and display the recommendation list to the user.

[0013] To achieve the above-mentioned purpose, the third aspect of an embodiment of the present application proposes an electronic device, an electronic device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the policy learning-based insurance pricing method as described in the first aspect when executing the computer program.

[0014] To achieve the above-mentioned purpose, the fourth aspect of an embodiment of the present application proposes a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the insurance pricing method based on policy learning as described in the first aspect.

[0015] The insurance pricing method, device, electronic device and storage medium based on policy learning proposed in the present application first obtain the user's behavior data and the target behavior data of the behavior data within a preset time period, thereby reducing the amount of calculation and being able to selectively predict the pricing of the data. Since different data categories are set with different initial prices, the embodiment of the present application determines the initial pricing of the premium according to the data category of the target behavior data, and then extracts features of the target behavior data through a preset multi-view learning model to analyze the target behavior data to obtain the feature vector of the current time step, thereby fully understanding the complexity of the user's behavior, extracting more comprehensive features, and further improving the accuracy of evaluation and pricing. Afterwards, the initial premium pricing and feature vector are predicted through a preset policy network to achieve dynamic analysis of the user's behavior data. The pricing ratio and premium variance are obtained, so that the pricing ratio can reflect the change of pricing over time, further reflect the risk change of the target behavior data, and at the same time, the volatility of premium income can be measured by the premium variance, which is convenient for quantifying and evaluating the risk indicators of the target behavior data, and then the estimated current pricing is determined according to the pricing ratio, premium variance and initial premium pricing, so as to achieve accurate estimation of the current pricing, input the estimated current pricing, initial premium pricing and the historical feature vector of the previous time step of the current time step into the preset value network for pricing evaluation, output the evaluation value corresponding to the estimated current pricing, achieve accurate evaluation of the value of the estimated current pricing, adjust the estimated current pricing corresponding to the data category according to the evaluation value, so as to ensure the personalization, accuracy and real-time nature of pricing, and achieve accurate pricing of the insurance products corresponding to the target behavior data. The embodiment of the present application can monitor the user's behavior data in real time, and can perform dynamic risk evaluation and pricing adjustment according to the user's behavior data, ensure the personalization, accuracy and real-time nature of pricing through accurate pricing and personalized services, so that users can obtain more accurate and convenient products and services, and improve the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a flow chart of an insurance pricing method based on policy learning provided in an embodiment of the present application; Figure 2 yes Figure 1 Flow chart of step S103 in FIG. Figure 3 yes Figure 1 Flow chart of step S105 in FIG. Figure 4 yes Figure 1 Flow chart of step S104 in FIG. Figure 5 yes Figure 1 Flow chart of step S106 in FIG. Figure 6 yes Figure 1Flow chart of step S107 in FIG. Figure 7 is a flow chart of an insurance pricing method based on policy learning provided by another embodiment of the present application; Figure 8 A flow chart of a method for calculating the reward value of target behavior data at different time steps provided in an embodiment of the present application; Fig. 9 It is a structural diagram of an insurance pricing device based on policy learning provided in an embodiment of the present application; Fig.10 It is a schematic diagram of the hardware structure of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0018] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0020] First, some nouns involved in this application are analyzed: Natural Language Processing (NLP): NLP uses computers to process, understand, and apply human languages ​​(such as Chinese, English, etc.). NLP is a branch of artificial intelligence and an interdisciplinary subject between computer science and linguistics. It is often referred to as computational linguistics. Natural language processing includes grammatical analysis, semantic analysis, and text understanding. Natural language processing is commonly used in technical fields such as machine translation, handwritten and printed character recognition, speech recognition and text-to-speech conversion, information intent recognition, information extraction and filtering, text classification and clustering, public opinion analysis, and opinion mining. It involves data mining related to language processing, machine learning, knowledge acquisition, knowledge engineering, artificial intelligence research, and linguistic research related to language computing.

[0021] Multi-view Learning (MVL): Multi-view learning is a machine learning method that learns a common feature space or shared patterns by combining information from multiple different features or data sources. A main approach of multi-view learning is to map multi-view data into a common feature space, maximizing the mutual consistency between multiple views.

[0022] The insurance pricing method and device based on policy learning, electronic device and storage medium provided in the embodiments of the present application can realize dynamic adjustment of pricing according to user behavior data.

[0023] The following examples are used to illustrate the specific aspects. First, the insurance pricing method based on policy learning in the examples of the present application is described.

[0024] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0025] The basic technologies of artificial intelligence generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. Artificial intelligence software technologies mainly include computer vision technology, robotics technology, biometrics technology, module management technology of online meeting systems, natural language processing technology, and machine learning / deep learning.

[0026] The insurance pricing method based on policy learning provided in the embodiment of the present application relates to the field of financial technology. The insurance pricing method based on policy learning provided in the embodiment of the present application can be applied to a terminal, can be applied to a server side, or can be software running in a terminal or a server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements an insurance pricing method based on policy learning, etc., but is not limited to the above forms.

[0027] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data information types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0028] It should be noted that in each specific implementation of the present application, when it comes to the need to perform relevant processing based on data related to user identity or characteristics such as user information, user behavior data, user historical data, and user location information, the user's permission or consent will be obtained first, and the collection, use, and processing of these data will comply with the relevant laws, regulations, and standards of the relevant countries and regions. In addition, when the embodiment of the present application needs to obtain the user's sensitive personal information, the user's separate permission or consent will be obtained through a pop-up window or by jumping to a confirmation page. After clearly obtaining the user's separate permission or consent, the necessary user-related data for the normal operation of the embodiment of the present application will be obtained.

[0029] Currently, most pricing and risk management systems in the insurance industry rely on historical data and use predefined models and rules to assess risks and determine premiums. These systems lack the ability to adjust dynamically in real time and are difficult to respond to changes in customer behavior. For example, auto insurance pricing is usually based on a customer's historical driving record, and health insurance relies on a customer's health history. Although these methods can assess risks to a certain extent, they cannot motivate customers to maintain good behavior (such as safe driving and healthy living) during the insurance cycle, and they are difficult to cope with dynamic changes in customer behavior.

[0030] Because the system cannot reflect the latest changes in customer behavior in real time, risk assessments may quickly become outdated, resulting in the inability of insurance companies to accurately predict and respond to emerging risks. At the same time, due to the lack of sensitivity to changes in customer behavior, insurance products may not be updated in a timely manner to cover new or changing risks, resulting in insufficient insurance coverage and further reducing the customer experience.

[0031] In order to solve the above problems, the present embodiment provides an insurance pricing method, device, electronic device and storage medium based on strategy learning. First, the user's behavior data is obtained, and the target behavior data of the behavior data within a preset time period is obtained, so as to reduce the amount of calculation and selectively predict the pricing of the data. Since different data categories are set with different initial prices, the embodiment of the present application determines the initial pricing of the premium according to the data category of the target behavior data, and then extracts features of the target behavior data through a preset multi-perspective learning model to realize the analysis of the target behavior data to the feature vector of the current time step, so as to fully understand the complexity of the user's behavior, extract more comprehensive features, and further improve the accuracy of evaluation and pricing. Afterwards, the initial pricing of the premium and the feature vector are predicted through a preset strategy network to realize the user's behavior data. Dynamic analysis is performed to obtain pricing ratios and premium variances, so that pricing changes over time can be reflected through pricing ratios, and further reflect risk changes in target behavior data. At the same time, the volatility of premium income can be measured through premium variances, which is convenient for quantifying and evaluating risk indicators of target behavior data. The estimated current pricing is determined based on pricing ratios, premium variances, and initial premium pricing, so as to achieve accurate estimation of current pricing. The estimated current pricing, initial premium pricing, and historical feature vectors of the previous time step of the current time step are input into a preset value network for pricing evaluation, and the evaluation value corresponding to the estimated current pricing is output to achieve accurate evaluation of the value of the estimated current pricing. The estimated current pricing corresponding to the data category is adjusted according to the evaluation value, so as to ensure the personalization, accuracy, and real-time nature of pricing, and achieve accurate pricing of insurance products corresponding to target behavior data. The embodiment of the present application can monitor the user's behavior data in real time, and can perform dynamic risk assessment and pricing adjustment based on the user's behavior data. Through accurate pricing and personalized services, the personalization, accuracy, and real-time nature of pricing are ensured, so that users can obtain more accurate and convenient products and services, and improve user experience.

[0032] The following is a detailed description with reference to the accompanying drawings.

[0033] Figure 1 is an optional flow chart of the insurance pricing method based on policy learning provided in the embodiment of the present application. Figure 1 The method may include but is not limited to steps S101 to S107.

[0034] Step S101, obtaining user behavior data, and obtaining target behavior data of the behavior data within a preset time period, wherein the behavior data includes data of multiple time steps.

[0035] In step S101 of some embodiments, the user's behavior data is obtained, and target behavior data of the behavior data within a preset time period is obtained, for example, target behavior data of the behavior data within a month is obtained, target behavior data of the behavior data within a week is obtained, and so on, to facilitate subsequent analysis of insurance pricing based on the target behavior data.

[0036] It should be noted that the behavior data in the embodiment of the present application may be data obtained from a user device, such as a vehicle-mounted sensor, a health bracelet, and the like. The behavior data includes but is not limited to the user's current driving data, health data, and the like. The driving data may be sudden braking, speeding, and the like, and the health data may be the number of steps, heart rate, and the like. The embodiment of the present application records the behavior data as . Include Features, including time steps (for example, the time step unit is "day", that is, The customer exists in days of behavior data), record the current time step as , the customer behavior data at the current time step is ,but .

[0037] Specifically, the embodiment of the present application takes the behavior data as vehicle driving data as an example for explanation, and takes the time from when a single user can record data (such as when the car is just bought) to the time of the last accident of the vehicle if the vehicle has had an accident. If the vehicle has not had an accident, take the time until the last record of the vehicle to obtain the target behavior data.

[0038] Step S102, determining the initial premium pricing according to the data category of the target behavior data.

[0039] In step S102 of some embodiments, since different data categories correspond to different initial premiums, the embodiments of the present application determine the initial premium pricing according to the data category of the target behavior data to facilitate subsequent adjustments to the pricing of the target behavior data.

[0040] Step S103, extracting features from the target behavior data using a preset multi-view learning model to obtain a feature vector for the current time step.

[0041] In step S103 of some embodiments, feature extraction is performed on the target behavior data through a preset multi-perspective learning model, and the features of the target behavior data are extracted by combining a convolutional neural network and a long short-term memory network. This enables the multi-perspective learning model to have better generalization capabilities when processing complex data, and to capture data features from different angles, thereby obtaining a feature vector for the current time step and further improving the accuracy of risk assessment and pricing.

[0042] Step S104, performing pricing prediction on the initial premium pricing and feature vector through a preset strategy network to obtain a pricing ratio and premium variance.

[0043] In step S104 of some embodiments, the initial premium pricing and feature vector are predicted through a preset strategy network to obtain a pricing ratio and premium variance, thereby reflecting the changes in insurance product pricing over time and measuring the volatility of premium pricing, reducing the uncertainty of premium income, and further obtaining real-time pricing of insurance, thereby realizing dynamic evaluation of insurance data.

[0044] Step S105, determining an estimated current price based on the pricing ratio, the premium variance, and the initial premium price.

[0045] In step S105 of some embodiments, the estimated current pricing is determined based on the pricing ratio, the premium variance, and the initial premium pricing, so as to ensure that the estimated current pricing matches the risk and avoid the estimated current pricing being too high or too low.

[0046] Step S106, input the estimated current pricing, the initial premium pricing, and the historical feature vector of the previous time step of the current time step into a preset value network for pricing evaluation, and output the evaluation value corresponding to the estimated current pricing.

[0047] In step S106 of some embodiments, the estimated current pricing, the initial premium pricing, and the historical feature vector of the previous time step of the current time step are input into a preset value network for pricing evaluation, so as to obtain the value of the user behavior characteristics at the current time step, and output the evaluation value, so as to perform dynamic risk assessment based on the user's real-time behavior data, accurately evaluate the value of the current behavior data, and facilitate subsequent adjustment of data pricing.

[0048] Step S107, adjusting the estimated current pricing corresponding to the data category according to the assessed value.

[0049] In step S107 of some embodiments, the estimated current pricing corresponding to the data category is adjusted according to the assessed value to achieve dynamic adjustment of data pricing, which can provide users with more accurate and flexible insurance plans.

[0050] See also Figure 2 In some embodiments, step S103 may also include but is not limited to steps S201 to S202.

[0051] It should be noted that the multi-view learning model includes convolutional neural networks and long short-term memory networks.

[0052] Step S201, input the target behavior data into a preset multi-view learning model, so that the convolutional neural network extracts features from the target behavior data and outputs a first feature vector, and the long short-term memory network extracts features from the target behavior data and outputs a second feature vector.

[0053] In step S201 of some embodiments, the convolutional neural network and the long short-term memory network in the multi-view learning model extract features in parallel. Specifically, the target behavior data is input into the preset multi-view learning model so that the convolutional neural network performs a one-dimensional convolution operation on the target behavior data, wherein the convolution kernel size is , and in the process of feature extraction by convolutional neural network, the embodiment of the present application also adopts the same filling strategy to ensure that the output time step remains ,pass The convolution kernel slides in the time dimension to extract features and output the feature matrix , that is, extract at each time step features and outputs the first feature vector. Similarly, the LSTM network extracts features from the target behavior data through a series of gating mechanisms to capture long-term dependencies in the time series. Specifically, after the target behavior data passes through each layer of the LSTM network, the hidden state of each time step t is generated. , forming the output feature matrix , which is the second eigenvector, where is the number of hidden units. By combining the convolutional neural network and the long short-term memory network to extract the characteristics of the target behavior data, the multi-view learning model has better generalization ability when processing complex data and can capture data features from different angles.

[0054] It should be noted that the embodiment of the present application also adopts the same padding strategy. Specifically, before the convolution operation, each dimension (time step) of the input data is padded. For example, if p is the calculated padding size, p zeros are added before and after the input data to ensure that the number of time steps of the output after convolution remains t, and then the convolution kernel is used to slide on the padded input data to perform the convolution operation. Due to the use of same padding, the convolution kernel can cover the entire input data (including the padded zeros), thereby ensuring that the number of time steps of the output is the same as the input.

[0055] Step S202, performing feature concatenation on the first feature vector and the second feature vector to obtain a feature vector of the data category at the current time step.

[0056] In step S202 of some embodiments, the first feature vector and the second feature vector are concatenated. By integrating these two feature extraction methods, the multi-view learning model can more comprehensively understand the complexity of user behavior, obtain the feature vector of the data category at the current time step, and further improve the accuracy of risk assessment and pricing. Specifically, the feature vector can be expressed as The fused features are expressed as .

[0057] See also Figure 3 In some embodiments, step S105 may also include but is not limited to steps S301 to S302.

[0058] Step S301, obtaining a normal probability distribution according to the pricing ratio, premium variance and initial premium pricing.

[0059] Step S302, performing sampling operation on the normal probability distribution to obtain an estimated current price.

[0060] In steps S301 to S302 of some embodiments, in the process of determining the estimated current pricing based on the pricing ratio, the premium variance and the initial pricing of the premium, a normal distribution is constructed based on the pricing ratio, the premium variance and the initial pricing of the premium. Specifically, the mean of the normal probability distribution is the product of the pricing ratio and the initial pricing of the premium. Through the normal distribution, it is possible to quantify risks and evaluate risk-taking at different premium levels. The normal probability distribution is then sampled and the sampled values ​​obtained are used as the estimated current pricing, thereby ensuring that the estimated current pricing matches the risk and avoiding estimating the current pricing too high or too low.

[0061] Specifically, the policy network Output Pricing Ratio , premium variance Combined with initial premium pricing , can get The normal probability distribution of Sample a value to get .in, , , , for Distribute the next sample value, Distribution by , The normal distribution calculated with mean is , the variance is .

[0062] See also Figure 4In some embodiments, step S104 may also include but is not limited to steps S401 to S402.

[0063] Step S401, obtaining the current pricing ratio of the data category at the current time step.

[0064] It should be noted that the estimated current pricing ratio is used to characterize the ratio of the current pricing of the data category at the current time step to the initial pricing of the premium.

[0065] Step S402, inputting the initial premium pricing, feature vector and current pricing ratio into a preset strategy network, so that the strategy network concatenates the initial premium pricing, feature vector and current pricing ratio, and outputs the pricing ratio and premium variance.

[0066] In steps S401 to S402 of some embodiments, in the process of performing pricing prediction on the initial premium pricing and feature vector through the preset strategy network to obtain the pricing ratio and premium variance, the embodiment of the present application obtains the current pricing ratio of the data category at the current time step, that is, ,in, A value representing the ratio of the current price at the current state (current time step) to the initial price of the premium , and then input the initial premium pricing, the feature vector of the current time step, and the current pricing ratio into the preset strategy network, so that the strategy network can make predictions in chronological order, splice the initial premium pricing, the feature vector, and the current pricing ratio, and output the pricing ratio and premium variance, so as to reflect the changes in insurance product pricing over time, measure the volatility of premium pricing, reduce the uncertainty of premium income, and further obtain real-time pricing of insurance, realizing dynamic evaluation of insurance data.

[0067] It should be noted that the embodiment of the present application converts the feature vector of the current time step ,in, , initial pricing of premiums , pricing ratio , input policy network , output the pricing ratio for the next step , premium variance .in, It is a numerical value determined by the type of insurance paid by the user. is a value representing the ratio of the current price at the current state (step t) to the initial price , premium variance is also a value used to describe the distribution of current pricing. When forecasting, forecast in chronological order, that is, first , , Input Policy Network ,get , . You can then , Input to the policy network again ,get and .

[0068] It is understandable that the pricing ratio in the embodiments of the present application is Initially it is 1, which is obtained by the strategy network at the previous step t-1. Predicted. is the user's behavior feature at time t, obtained by the multi-view learning model. , , After splicing together (dimension 2+ ), output two values ​​through three layers of full connection (dimensions are 100, 10, and 2 respectively) and .

[0069] See also Figure 5 In some embodiments, step S106 may also include but is not limited to steps S501 to S502.

[0070] Step S501, input the estimated current price, the initial premium price and the historical feature vector of the previous time step of the current time step into a preset value network, so that the value network splices the estimated current price, the initial premium price and the historical feature vector and outputs a spliced ​​vector.

[0071] Step S502, input the concatenated vector into the fully connected layer of the value network, so that the fully connected layer scores the estimated current price according to the concatenated vector and outputs the assessed value.

[0072] In some embodiments, in steps S501 to S502, in the process of inputting the estimated current pricing, the initial premium pricing, and the historical feature vector of the previous time step of the current time step into the preset value network for pricing evaluation, the pricing ratio and premium variance output by the strategy network are scored through the value network, and the pricing ratio and premium variance output by the strategy network combined with the initial premium pricing can obtain the estimated current pricing, and the estimated current pricing, the initial premium pricing, and the historical feature vector of the previous time step of the current time step are input into the preset value network, so that the value network splices the estimated current pricing, the initial premium pricing, and the historical feature vector, and outputs the splicing vector, so as to achieve a comprehensive analysis of the data and improve the accuracy of the data analysis, and then input the splicing vector into the fully connected layer of the value network, and the estimated current pricing is scored according to the splicing vector through two layers of full connection, so as to obtain the value of the user behavior characteristics at the current time step, and output the evaluation value, so as to perform dynamic risk assessment based on the user's real-time behavior data, accurately assess the value of the current behavior data, and facilitate subsequent adjustment of the data pricing.

[0073] It is understandable that the value network will , , after being spliced ​​together (dimension is 3), a value is obtained through two layers of full connection (dimensions are 100 and 1 respectively) , which represents the user's historical feature vector at time t In the case of policy network Output Scoring, if The more accurate, the The higher.

[0074] See also Figure 6 In some embodiments, step S107 may also include but is not limited to steps S601 to S602.

[0075] Step S601, when the evaluation value is greater than or equal to a first preset threshold, the estimated current price corresponding to the data category is increased based on the evaluation value.

[0076] In step S601 of some embodiments, in the process of adjusting the estimated current pricing corresponding to the data category according to the evaluation value, the embodiments of the present application compare the evaluation value with the first preset threshold and the second preset threshold. When the evaluation value is greater than or equal to the first preset threshold, it means that the value of the current data is higher. The estimated current pricing corresponding to the data category can be increased based on the evaluation value, thereby increasing the pricing of data with higher value and realizing dynamic adjustment of data pricing.

[0077] Step S602: when the evaluation value is less than a preset second preset threshold, the estimated current price corresponding to the data category is reduced based on the evaluation value.

[0078] In step S602 of some embodiments, when the assessed value is less than a preset second preset threshold, indicating that the value of the current data is low, the estimated current pricing corresponding to the data category can be reduced based on the assessed value, thereby achieving dynamic adjustment of the data pricing, and providing users with more accurate and flexible insurance plans.

[0079] It should be noted that the first preset threshold is greater than the second preset threshold.

[0080] It is worth noting that when the evaluated price is in the interval formed by the first preset threshold and the second preset threshold, there is no need to adjust the estimated current price, and the estimated current price can be maintained.

[0081] See also Figure 7 , Figure 7 is a flow chart of an insurance pricing method based on policy learning provided by another embodiment of the present application, Figure 7 The method may include but is not limited to step S701 to step S702.

[0082] It should be noted that step S701 to step S702 occur after the estimated current pricing, the initial premium pricing, and the historical feature vector of the previous time step of the current time step are input into the preset value network for pricing evaluation, and the evaluation value corresponding to the estimated current pricing is output.

[0083] Step S701, calculating the reward value of the target behavior data at different time steps.

[0084] Step S702: training a value network according to the reward value and the evaluation value to obtain a pre-trained value network.

[0085] In steps S701 to S702 of some embodiments, after outputting the estimated evaluation value corresponding to the current pricing, the embodiments of the present application will also calculate the reward value of the target behavior data at different time steps, so as to obtain the reward value of the target behavior data at different time steps and under different states, and can motivate customers to maintain healthy living habits through a reward feedback mechanism. Then, the value network is trained according to the reward value and the evaluation value to obtain a pre-trained value network, which can quickly adapt to and optimize pricing strategies in a complex environment, achieve more efficient risk management, and improve the overall efficiency and service quality of the insurance industry.

[0086] Specifically, the embodiment of the present application obtains a loss function according to the reward value and the evaluation value, which is expressed as loss = mse( ),in, is the reward value, To assess value, improve the prediction accuracy of the value network for pricing.

[0087] See also Figure 8 , Figure 8 This is a flow chart of a method for calculating the reward value of target behavior data at different time steps provided in an embodiment of the present application. The method may include but is not limited to steps S801 to S804.

[0088] Step S801, performing status detection on target behavior data.

[0089] Step S802, when the target behavior data completes state detection and detects that the target behavior data has a preset behavior, determine the first time step at which the preset behavior occurs and the first pricing corresponding to the first time step, and obtain the compensation amount and the first profit margin corresponding to the preset behavior at the first time step, and obtain the first reward value through the compensation amount, the first profit margin and the first pricing.

[0090] In steps S801 to S802 of some embodiments, in the process of calculating the reward values ​​of the target behavior data at different time steps, the embodiments of the present application first perform a state detection on the target behavior data to determine the state of the target behavior data. For example, taking the target behavior data as vehicle driving data as an example, the state of the target behavior data at this time may be that an accident has occurred, no accident has occurred, etc. When the target behavior data completes the state detection and detects that the target behavior data exhibits a preset behavior, it is necessary to determine the first time step at which the preset behavior occurs and the first pricing corresponding to the first time step, thereby determining the first time step at which the preset behavior occurs, and obtaining the compensation amount and the first profit margin corresponding to the preset behavior at the first time step, and obtaining the first reward value through the compensation amount, the first profit margin and the first pricing. Specifically, the embodiments of the present application first determine the standard premium amount through the compensation amount and the first profit margin, and then determine the first reward value through the standard premium amount and the first pricing, so as to obtain the reward value when the preset behavior occurs.

[0091] It should be noted that the preset behavior in the embodiment of the present application can be set according to the data type of the target behavior data. Specifically, take the target behavior data as vehicle driving data and the preset behavior as the vehicle having an accident as an example. When it is determined that an accident has occurred during vehicle driving, the first time step at which the preset behavior occurs and the first pricing corresponding to the first time step are determined, and the compensation amount corresponding to the preset behavior at the first time step is obtained. and the first profit margin , through the amount of compensation and the first profit margin Calculate standard premium amount , where the standard premium amount is calculated through The first reward value is calculated based on the standard premium amount and the first pricing. The calculation process is as follows: ; in, For the first price, is the standard premium amount, express and The reciprocal of the absolute value of the difference, specifically, and The smaller the difference, the greater the reward. Indicates that The maximum and minimum values ​​are limited to between -100 and 100.

[0092] Step S803, when the target behavior data completes state detection and no preset behavior is detected in the target behavior data, determine the second time step at the last moment within the preset time period and the second profit margin corresponding to the second time step, and obtain the second price corresponding to the second time step, and obtain the second reward value through the second profit margin and the second price.

[0093] In step S803 of some embodiments, when the target behavior data completes status detection and no preset behavior is detected in the target behavior data, the second time step at the last moment within the preset time period and the second profit margin corresponding to the second time step are directly determined, and the second price corresponding to the second time step is obtained. The second reward value is obtained through the second profit margin and the second price, so that different reward values ​​can be set according to the status of different behavior data.

[0094] Specifically, take the target behavior data as vehicle driving data and the preset behavior as the vehicle having an accident as an example. When it is determined that no accident has occurred during the vehicle driving process, the second profit margin is determined as the standard premium amount at the last step T, that is, The reward value calculation process is shown in step S802, and the embodiment of the present application will not be repeated here.

[0095] Step S804: When the target behavior data has not completed the state detection, the preset value is used as the third reward value.

[0096] In step S804 of some embodiments, when the target behavior data has not completed status detection, it means that all time steps within the preset time period have not been detected. The preset value can be directly used as the third reward value, thereby motivating customers to maintain healthy living habits through a reward feedback mechanism, thereby improving the overall efficiency and service quality of the insurance industry.

[0097] See also Fig. 9 The embodiment of the present application also provides an insurance pricing device based on policy learning, the device comprising: The data acquisition module 901 is used to acquire the user's behavior data and acquire the target behavior data of the behavior data within a preset time period, wherein the behavior data includes data of multiple time steps; A pricing determination module 902, for determining an initial premium pricing according to the data category of the target behavior data; A feature extraction module 903 is used to extract features from the target behavior data through a preset multi-view learning model to obtain a feature vector of the current time step; The pricing prediction module 904 is used to perform pricing prediction on the initial premium pricing and feature vector through a preset strategy network to obtain a pricing ratio and premium variance; A pricing estimation module 905, for determining an estimated current pricing based on the pricing ratio, the premium variance, and the initial premium pricing; The value evaluation module 906 is used to input the estimated current pricing, the initial premium pricing, and the historical feature vector of the previous time step of the current time step into a preset value network for pricing evaluation, and output the evaluation value corresponding to the estimated current pricing; The pricing adjustment module 907 is used to adjust the estimated current pricing corresponding to the data category according to the assessed value.

[0098] In some embodiments, the insurance pricing device based on policy learning also includes a training network module, which is used to calculate the reward value of the target behavior data at different time steps; train the value network according to the reward value and the evaluation value to obtain a pre-trained value network; and perform state detection on the target behavior data; when the target behavior data completes state detection and detects that the target behavior data has a preset behavior, determine the first time step at which the preset behavior occurs and the first pricing corresponding to the first time step, and obtain the compensation amount and the first profit margin corresponding to the preset behavior at the first time step, and obtain the first reward value through the compensation amount, the first profit margin and the first pricing; when the target behavior data completes state detection and the target behavior data is not detected to have the preset behavior, determine the second time step at the last moment of the preset time period and the second profit margin corresponding to the second time step, and obtain the second pricing corresponding to the second time step, and obtain the second reward value through the second profit margin and the second pricing; and when the target behavior data has not completed state detection, use the preset value as the third reward value.

[0099] The specific implementation of the insurance pricing device based on policy learning is basically the same as the specific implementation of the insurance pricing method based on policy learning described above, and will not be repeated here.

[0100] The embodiment of the present application also provides an electronic device, the electronic device comprising: a memory, a processor, a program stored in the memory and executable on the processor, and a data bus for realizing connection and communication between the processor and the memory, and the program is executed by the processor to realize the above-mentioned insurance pricing method based on policy learning. The electronic device can be any intelligent terminal including a tablet computer, a car computer, etc.

[0101] See also Fig.10 , Fig.10 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes: The processor 1001 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application; The memory 1002 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 1002 can store an operating system and other application programs. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 1002, and the processor 1001 calls and executes the insurance pricing method based on policy learning in the embodiment of this application; Input / output interface 1003, used to implement information input and output; The communication interface 1004 is used to realize the communication interaction between the device and other devices. The communication can be realized through a wired manner (such as USB, network cable, etc.) or a wireless manner (such as mobile network, WIFI, Bluetooth, etc.); A bus 1005 , which transmits information between various components of the device (e.g., the processor 1001 , the memory 1002 , the input / output interface 1003 , and the communication interface 1004 ); The processor 1001 , the memory 1002 , the input / output interface 1003 and the communication interface 1004 are connected to each other in communication within the device via the bus 1005 .

[0102] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the above-mentioned insurance pricing method based on policy learning.

[0103] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0104] The insurance pricing method, device, electronic device and storage medium based on policy learning provided in the embodiments of the present application first obtain the user's behavior data and obtain the target behavior data of the behavior data within a preset time period, thereby reducing the amount of calculation and being able to selectively predict the pricing of the data. Since different data categories are set with different initial prices, the embodiments of the present application determine the initial pricing of the premium according to the data category of the target behavior data, and then extract features of the target behavior data through a preset multi-perspective learning model to analyze the target behavior data to obtain the feature vector of the current time step, thereby fully understanding the complexity of the user's behavior, extracting more comprehensive features, and further improving the accuracy of evaluation and pricing. Afterwards, the initial pricing of the premium and the feature vector are predicted through a preset policy network to achieve dynamic analysis of the user's behavior data. Analysis is performed to obtain pricing ratios and premium variances, so that pricing ratios can reflect changes in pricing over time, further reflecting changes in the risk of target behavior data, and premium variances can be used to measure the volatility of premium income, which is convenient for quantifying and evaluating the risk indicators of target behavior data, and then the estimated current pricing is determined based on the pricing ratio, premium variance, and initial premium pricing, so as to achieve an accurate estimate of the current pricing, and the estimated current pricing, initial premium pricing, and the historical feature vector of the previous time step of the current time step are input into a preset value network for pricing evaluation, and the evaluation value corresponding to the estimated current pricing is output, so as to achieve an accurate evaluation of the value of the estimated current pricing, and the estimated current pricing corresponding to the data category is adjusted according to the evaluation value, so as to ensure the personalization, accuracy, and real-time nature of pricing, and achieve accurate pricing of insurance products corresponding to the target behavior data. The embodiment of the present application can monitor the user's behavior data in real time, and can perform dynamic risk evaluation and pricing adjustment based on the user's behavior data, and ensure the personalization, accuracy, and real-time nature of pricing through accurate pricing and personalized services, so that users can obtain more accurate and convenient products and services, and improve user experience.

[0105] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0106] It can be understood by those skilled in the art that Figure 1-9 The technical solutions shown in the figure do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than those shown in the figure, or a combination of certain steps, or different steps.

[0107] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0108] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0109] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0110] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0111] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0112] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0113] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0114] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0115] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. An insurance pricing method based on policy learning, characterized in that: The method comprises: Acquire user behavior data, and acquire target behavior data of the behavior data within a preset time period, wherein the behavior data includes data of multiple time steps; Determining initial premium pricing according to the data category of the target behavior data; Extract features from the target behavior data using a preset multi-view learning model to obtain a feature vector for the current time step; Performing pricing prediction on the initial premium pricing and the feature vector through a preset strategy network to obtain a pricing ratio and premium variance; determining an estimated current pricing based on the pricing ratio, the premium variance, and the initial premium pricing; Input the estimated current pricing, the initial premium pricing, and the historical feature vector of the previous time step of the current time step into a preset value network for pricing evaluation, and output an evaluation value corresponding to the estimated current pricing; The estimated current pricing corresponding to the data category is adjusted according to the assessed value.

2. The insurance pricing method based on policy learning according to claim 1, characterized in that: The multi-view learning model includes a convolutional neural network and a long short-term memory network; the feature extraction of the target behavior data by the preset multi-view learning model to obtain the feature vector of the current time step includes: Inputting the target behavior data into a preset multi-view learning model, so that the convolutional neural network extracts features from the target behavior data and outputs a first feature vector, and the long short-term memory network extracts features from the target behavior data and outputs a second feature vector; Feature concatenation is performed on the first feature vector and the second feature vector to obtain a feature vector of the data category at the current time step.

3. The insurance pricing method based on policy learning according to claim 1, characterized in that: The step of determining the estimated current pricing according to the pricing ratio, the premium variance and the initial premium pricing includes: Obtaining a normal probability distribution according to the pricing ratio, the premium variance, and the initial premium pricing; A sampling operation is performed on the normal probability distribution to obtain an estimated current price.

4. The insurance pricing method based on policy learning according to claim 1, characterized in that: The method of performing pricing prediction on the initial premium pricing and the feature vector through a preset strategy network to obtain a pricing ratio and a premium variance includes: Obtaining a current pricing ratio of the data category at a current time step, wherein the estimated current pricing ratio is used to represent a ratio of the current pricing of the data category at a current time step to the initial pricing of the premium; The initial premium pricing, the feature vector and the current pricing ratio are input into a preset strategy network, so that the strategy network concatenates the initial premium pricing, the feature vector and the current pricing ratio, and outputs the pricing ratio and premium variance.

5. The insurance pricing method based on policy learning according to claim 1, characterized in that: The step of inputting the estimated current pricing, the initial premium pricing, and the historical feature vector of the previous time step of the current time step into a preset value network for pricing evaluation, and outputting the evaluation value corresponding to the estimated current pricing includes: Inputting the estimated current price, the initial premium price and the historical feature vector of the previous time step of the current time step into a preset value network, so that the value network splices the estimated current price, the initial premium price and the historical feature vector and outputs a spliced ​​vector; The concatenated vector is input into a fully connected layer of the value network, so that the fully connected layer scores the estimated current pricing according to the concatenated vector and outputs an assessed value.

6. The insurance pricing method based on policy learning according to claim 1, characterized in that: The adjusting the estimated current price corresponding to the data category according to the assessed value includes: When the evaluation value is greater than or equal to a first preset threshold, increasing the estimated current price corresponding to the data category based on the evaluation value; When the evaluation value is less than a preset second preset threshold, reducing the estimated current price corresponding to the data category based on the evaluation value; Wherein, the first preset threshold is greater than the second preset threshold.

7. The insurance pricing method based on policy learning according to claim 6, characterized in that: After inputting the estimated current pricing, the initial premium pricing, and the historical feature vector of the previous time step of the current time step into a preset value network for pricing evaluation and outputting the evaluation value corresponding to the estimated current pricing, the method further includes: Calculating the reward value of the target behavior data at different time steps; Training the value network according to the reward value and the evaluation value to obtain a pre-trained value network; The calculating the reward value of the target behavior data at different time steps includes: Performing status detection on the target behavior data; When the target behavior data completes state detection and detects that the target behavior data has a preset behavior, determine the first time step at which the preset behavior occurs and the first pricing corresponding to the first time step, obtain the compensation amount and the first profit margin corresponding to the preset behavior at the first time step, and obtain the first reward value through the compensation amount, the first profit margin and the first pricing; When the target behavior data completes state detection and no preset behavior is detected in the target behavior data, a second time step at the last moment within the preset time period and a second profit margin corresponding to the second time step are determined, and a second price corresponding to the second time step are obtained, and a second reward value is obtained through the second profit margin and the second price; When the target behavior data has not completed the state detection, the preset value is used as the third reward value.

8. An insurance pricing device based on policy learning, characterized in that: The device comprises: A data acquisition module, used to acquire user behavior data and acquire target behavior data of the behavior data within a preset time period, wherein the behavior data includes data of multiple time steps; A pricing determination module, used to determine an initial premium pricing according to the data category of the target behavior data; A feature extraction module is used to extract features from the target behavior data through a preset multi-view learning model to obtain a feature vector of the current time step; A pricing prediction module, used to perform pricing prediction on the initial premium pricing and the feature vector through a preset strategy network to obtain a pricing ratio and a premium variance; A pricing estimation module, configured to determine an estimated current pricing based on the pricing ratio, the premium variance, and the initial premium pricing; A value assessment module, used for inputting the estimated current pricing, the initial premium pricing and the historical feature vector of the previous time step of the current time step into a preset value network for pricing assessment, and outputting the assessment value corresponding to the estimated current pricing; The pricing adjustment module is used to adjust the estimated current pricing corresponding to the data category according to the assessed value.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the insurance pricing method based on policy learning as described in any one of claims 1 to 7 when executing the computer program.

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

Images