Risk identification factor mining method and device, computer device, and storage medium

Abstract

The application belongs to the field of artificial intelligence and finance, and relates to a risk identification factor mining method, which comprises the following steps: obtaining different data types of data, setting different types of operators according to an insurance risk identification target, constructing a genetic programming model according to the different types of operators, initializing a formula population, iteratively calculating a risk explanation degree, continuously optimizing a formula tree through a genetic algorithm, obtaining a final formula tree population, selecting a formula tree with the highest risk explanation degree from the final formula tree population as a risk identification factor, and testing and optimizing the risk identification factor to obtain an optimal risk identification model. The application also provides a risk identification factor mining device, a computer device and a storage medium. In addition, the application also relates to blockchain technology, and text data and numerical data can be stored in the blockchain. The application can better realize automatic identification of insurance risks and improve the accuracy of risk identification.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence and financial technology, and particularly relates to a risk identification factor mining method and device, computer equipment and a storage medium. BACKGROUND

[0002] Insurance fraud, as an illegal act, has long plagued the insurance industry. Fraudsters try to cheat excessive insurance money by fabricating insurance targets, intentionally creating insurance accidents, fabricating the causes of accidents or exaggerating the extent of losses, which not only leads to the insurance company paying more insurance money than it should, but also seriously damages the fairness and stability of the insurance market. The traditional risk identification method mainly relies on manual review by insurance salesmen. Although this method can identify some frauds to some extent, it is limited by manpower, time and experience, and its efficiency and accuracy are insufficient.

[0003] In recent years, with the rapid development of machine learning and artificial intelligence technology, insurance companies have begun to try to introduce these advanced technologies to strengthen the management of risk fraud. By building a risk rule library, insurance companies can use big data and algorithm models to analyze insurance cases more comprehensively and deeply, so as to more effectively identify potential frauds. However, although this method has achieved certain results, it still faces a series of problems that need to be solved in the process of actual application.

[0004] Firstly, the diversity of data types and the high dimensionality are a big challenge in current risk fraud management. In order to more accurately depict the case information and the user portrait of the insured person, the insurance company needs to collect and process a large amount of multi-dimensional data. These data include not only numerical amount variables and time variables, but also long text information in classification variables, such as accident history, diagnosis results, etc. When extracting the feature variables of these data, due to the diversity and complexity of the data types, it is difficult to extract. Especially for long text information in classification variables, although key words can be extracted to reduce dimensionality, it often leads to the problem of dimensionality explosion, further increasing the difficulty of data processing.

[0005] Secondly, the requirement of high factor combination for specific scenarios is also one of the problems that need to be solved urgently. In the special scenario of fraud detection using policy data, numerical variables such as time and amount are often not effective if they are set as factors alone. For example, the report time and the policy start time have limited ability to identify fraudulent behavior if used alone. However, if the two variables are combined, such as subtraction, a new identification factor such as the difference between the report time and the policy start time can be generated, which can more effectively identify fraudulent behavior. In addition, the ratio of the actual premium amount to the claim amount is also an important indicator commonly used in fraud detection. However, the construction and mining of these factors require high professional knowledge and experience, and often rely on the subjective judgment of business experts, with low automation.

[0006] Finally, the low automation of risk identification factor construction and mining is also a major problem currently faced. The current risk identification factors are often synthesized and explained by business experts based on their rich experience. These factors are relatively simple in structure, and usually have logic before factors. Although this method can explain the rationality of risk factors to some extent, it is difficult to mine deep-level risk combinations that are more explanatory of fraudulent factors. SUMMARY

[0007] The embodiments of the present application aim to provide a risk identification factor mining method, device, computer equipment and storage medium, to solve the technical problems of related technologies, such as the increase in data processing difficulty due to the variety and complexity of data, the high requirement of risk identification factor combination but simple structure, the difficulty in mining deep-level risk factor combinations, and the low risk identification accuracy.

[0008] To solve the above technical problems, the embodiments of the present application provide a risk identification factor mining method, which adopts the following technical solutions:

[0009] Obtain unstructured text data and numerical data in an insurance business database, perform word segmentation processing on the text data to obtain a word segmentation result, extract keywords from the word segmentation result to obtain keyword data;

[0010] Obtain a first operation formula based on a logical operator set based on the keyword data, obtain a second operation formula based on a basic operator set based on the numerical data, and construct a genetic programming model according to an insurance risk identification target through the first operation formula and the second operation formula;

[0011] Generate a formula tree population of a preset number of formula trees through random initialization as an initial formula tree population corresponding to the genetic programming model;

[0012] Obtain the risk label data already labeled in the insurance business database, divide it into training set and test set according to a preset ratio, and calculate the risk explanatory power of each formula tree in the initial formula tree population through the training set;

[0013] Based on the risk interpretability, the initial formula tree population is iteratively optimized using a genetic algorithm until a preset termination condition is met, resulting in the final formula tree population.

[0014] From the final formula tree population, select a first preset number of formula trees with the highest risk explanatory power as risk identification factors;

[0015] Using the test set, calculate the risk identification accuracy of the risk identification factor. When the risk identification accuracy is greater than or equal to a preset threshold, output the genetic programming model corresponding to the risk identification factor as the risk identification model.

[0016] Obtain the insurance business data to be identified, input the insurance business data to be identified into the risk identification model, and obtain the risk identification result.

[0017] To address the aforementioned technical problems, this application also provides a risk identification factor mining device, which employs the following technical solution:

[0018] The acquisition module is used to acquire unstructured text data and numerical data from the insurance business database, perform word segmentation on the text data to obtain word segmentation results, and extract keywords from the word segmentation results to obtain keyword data.

[0019] A construction module is used to obtain a first operational formula based on the logical operators set based on the keyword data, and to obtain a second operational formula based on the basic operators set based on the numerical data. According to the insurance risk identification target, a genetic programming model is constructed through the first operational formula and the second operational formula.

[0020] An initialization module is used to generate a population of formula trees with a preset number of formula trees through random initialization, which serves as the initial population of formula trees corresponding to the genetic programming model.

[0021] The calculation module is used to obtain the risk label data already labeled in the insurance business database, divide the training set and test set according to a preset ratio, and calculate the risk interpretability of each formula tree in the initial formula tree population through the training set;

[0022] An iterative module is used to iteratively optimize the initial formula tree population using a genetic algorithm based on the risk interpretability until a preset termination condition is reached, thereby obtaining the final formula tree population.

[0023] A selecting module is configured to select a first preset number of formula trees with the highest risk explanation degree from the final formula tree population as risk identification factors;

[0024] A testing module is configured to calculate a risk identification accuracy of the risk identification factors using the test set, and output a genetic programming model corresponding to the risk identification factors as a risk identification model when the risk identification accuracy is greater than or equal to a preset threshold.

[0025] An identifying module is configured to obtain to-be-identified insurance business data, input the to-be-identified insurance business data into the risk identification model, and obtain a risk identification result.

[0026] To solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the technical scheme as follows:

[0027] The computer device includes a memory and a processor, the memory stores computer readable instructions, and the processor implements the steps of the mining method of the risk identification factors when executing the computer readable instructions.

[0028] To solve the above technical problems, the embodiment of the present application further provides a computer readable storage medium, which adopts the technical scheme as follows:

[0029] The computer readable storage medium stores computer readable instructions, and the computer readable instructions implement the steps of the mining method of the risk identification factors when executed by the processor.

[0030] Compared with the prior art, the present application has the following beneficial effects:

[0031] The present application provides a mining method of risk identification factors, which obtains different data types of data, sets different types of operators according to the insurance risk identification target, obtains operation formulas of different data according to different types of operators, constructs a genetic programming model based on the operation formulas, initializes a formula population, iteratively calculates a risk explanation degree, and continuously optimizes formula trees through a genetic algorithm to obtain a final formula tree population, selects a formula tree with the highest risk explanation degree from the final formula tree population as a risk identification factor, and tests and optimizes the risk identification factor to obtain an optimal risk identification model. The present application applies a genetic programming algorithm to the field of insurance risk identification, can automatically discover complex nonlinear risk identification factors, improves the accuracy and efficiency of risk identification, at the same time, models different data types respectively, fully utilizes the multi-dimensional characteristics of insurance data, better realizes the automatic identification of insurance risk, and further improves the accuracy and interpretability of risk identification. BRIEF DESCRIPTION OF DRAWINGS

[0032] In order to more clearly illustrate the solutions in the present application, the drawings needed to be used in the embodiments of the present application will be briefly introduced as follows. Obviously, the drawings in the following description only represent some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort.

[0033] Figure 1 is an exemplary system architecture diagram to which the present application can be applied;

[0034] Figure 2 is a flow chart of an embodiment of a mining method of risk identification factors according to the present application;

[0035] Figure 3 is a structural schematic diagram of a specific embodiment of formula tree connection provided according to the present application;

[0036] Figure 4 is a structural schematic diagram of another specific embodiment of formula tree connection provided according to the present application;

[0037] Figure 5 is a structural schematic diagram of a specific embodiment of cross operation according to the present application;

[0038] Figure 6 is a structural schematic diagram of an embodiment of a mining device of risk identification factors according to the present application;

[0039] Figure 7 is a structural schematic diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0040] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs; the terminology used in the description of the application herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the present application; the description and claims of the present application as well as the above drawings, illustrate and set forth rather than limit the present application; the terms "comprise", "comprising", "include", "including" and "has", "having" and variants thereof in the present description and in the claims are intended to cover both the case where only the stated feature is included and the case where additional features are also included. The terms "first", "second" and the like in the description and in the claims of the present application are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the terms so used are interchangeable under appropriate circumstances such that the embodiments of the application described herein are, for example, capable of use in either orientation.

[0041] Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are the various embodiments mutually exclusive or alternative. It is expressly understood that the embodiments described herein are merely examples from a whole class of embodiments of which the application is a part. It is further expressly understood that the application is intended to encompass all structures and their equivalents that do not depart from the spirit of the application.

[0042] In order to make persons skilled in the art better understand the scheme of the present application, the technical scheme in the embodiments of the present application will be clearly and completely described below with reference to the drawings.

[0043] As shown in Figure 1 , the system architecture 100 can include a terminal device 101, a network 102 and a server 103. The terminal device 101 can be a notebook computer 1011, a tablet computer 1012 or a mobile phone 1013. The network 102 is a medium for providing a communication link between the terminal device 101 and the server 103. The network 102 can include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0044] A user can use the terminal device 101 to interact with the server 103 through the network 102 to receive or send messages, etc. Various communication client applications can be installed on the terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0045] The terminal device 101 can be various electronic devices with a display screen and supporting web browsing. In addition to the notebook computer 1011, the tablet computer 1012 or the mobile phone 1013, the terminal device 101 can also be an electronic book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 player (Moving Picture Experts Group Audio Layer IV), a laptop computer and a desktop computer, etc.

[0046] The server 103 can be a server providing various services, such as a background server supporting the page displayed on the terminal device 101.

[0047] It should be noted that the risk identification factor mining method provided by the embodiments of the present application is generally executed by a server / terminal device, and accordingly, the risk identification factor mining device is generally provided in the server / terminal device.

[0048] It should be understood that Figure 1 the number of terminal devices, networks and servers in is only illustrative. According to the needs of implementation, there can be any number of terminal devices, networks and servers.

[0049] With reference to Figure 2 , a flowchart of one embodiment of the risk identification factor mining method according to the present application is shown, including the following steps:

[0050] Step S201: Obtain unstructured text data and numerical data from the insurance business database, perform word segmentation on the text data to obtain the word segmentation results, and extract keywords from the word segmentation results to obtain keyword data.

[0051] An insurance business database is a database specifically used to store, manage, and process data related to the business activities of insurance companies. It contains a large amount of insurance data, including but not limited to customer information, product information, order information, financial data, supply chain data, and inventory data. It involves different data types, such as numerical and text. Numerical data includes the insured amount and the insurance period, while text data includes text information such as the details of the accident and the diagnosis results.

[0052] In this embodiment, the risk identification factor mining method runs on electronic devices (e.g., Figure 1 The server / terminal device shown can obtain unstructured text and numerical data from the insurance business database via wired or wireless connections. It should be noted that the aforementioned wireless connection methods may include, but are not limited to, 3G / 4G / 5G connections, WiFi connections, Bluetooth connections, WiMAX connections, Zigbee connections, UWB (ultra-wideband) connections, and other currently known or future-developed wireless connection methods.

[0053] After obtaining insurance data from the insurance business database, unstructured text data and numerical data are extracted from the insurance data. The text data is then processed by word segmentation, which can be done using word segmentation tools, BERT word segmenter, etc., to obtain the word segmentation results, and keywords are extracted from the word segmentation results.

[0054] In this embodiment, the text data can be preprocessed before word segmentation, including removing irrelevant spaces, special characters, punctuation marks, and stop words.

[0055] It should be emphasized that, to further ensure the privacy and security of text and numerical data, the aforementioned text and numerical data can also be stored in a blockchain node.

[0056] The blockchain referred to in this application is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Essentially, a blockchain is a decentralized database, a chain of data blocks linked together using cryptographic methods. Each data block contains information about a batch of network transactions, used to verify the validity of the information (anti-counterfeiting) and generate the next block. A blockchain can include an underlying blockchain platform, a platform product service layer, and an application service layer.

[0057] In some optional implementations of the embodiment, the step of extracting the keywords from the word segmentation result to obtain the keyword data comprises:

[0058] According to the word segmentation result, the number of occurrences of each word segmentation in the text data is calculated by a word frequency statistical method to obtain the word frequency of each word segmentation.

[0059] The distribution of each word segmentation in the text data is calculated by using an inverse document frequency algorithm to obtain the inverse document frequency.

[0060] The weight value of each word segmentation is calculated by multiplying the word frequency and the inverse document frequency to obtain a word weight list.

[0061] The preset word frequency threshold and the document frequency threshold are obtained, and the word segmentation in the word weight list that simultaneously satisfies the word frequency greater than or equal to the word frequency threshold and the document frequency less than or equal to the document frequency threshold is screened out as a candidate keyword.

[0062] The candidate keywords are sorted in descending order of the weight value, and the candidate keywords with a preset number of words in the front are selected as the keyword data.

[0063] In the embodiment, the greater the weight value, the more important the keyword. The final keyword data is selected by the word frequency, the inverse document frequency and the weight value, which improves the keyword extraction efficiency, eliminates non-keywords with low weight, and makes the extracted keywords have strong explanatory power in risk identification, thereby improving the accuracy of risk identification.

[0064] Step S202, a first operation formula is obtained based on the logical operator set by the keyword data, a second operation formula is obtained based on the basic operator set by the numerical data, and a genetic programming model is constructed by the first operation formula and the second operation formula according to the insurance risk identification target.

[0065] The genetic programming is a heuristic supervised learning algorithm of the brute force iteration type inspired by Darwin's natural evolution theory. The elements in the genetic programming are represented in a tree structure. The tree is a formula spliced by functions or operators acting on variables. The internal nodes are functions or operators, and the leaf nodes are filtered variables.

[0066] In the embodiment, different operators are set for different data types of data, and an operation formula can be obtained based on the operators, and the operation formula is a component of the genetic programming formula tree. Specifically, a logical operator is set for keyword data, and the logical operator includes or, and, and not; a basic operator is set for numerical data, and the basic operator includes addition, subtraction, multiplication, and division. A formula containing a logical operator is used as a first operation formula, and a formula containing an addition, subtraction, multiplication, and division basic operator is used as a second operation formula, and a genetic programming model is constructed through the first operation formula and the second operation formula according to an insurance risk identification target.

[0067] In some optional implementations, different genetic programming models are constructed for different data types. For example, three data types of amount, time, and keyword in insurance data are obtained, for the amount data, the amount data is numerical data, and a formula containing an addition, subtraction, multiplication, and division basic operator is used as an operation formula for constructing a first genetic programming model; for the time data, the time data is also numerical data, and a formula containing an addition, subtraction, multiplication, and division basic operator is used as an operation formula for constructing a second genetic programming model; for the keyword data, a formula containing a logical operator is used as an operation formula for constructing a third genetic programming model; and the first genetic programming model, the second genetic programming model, and the third genetic programming model are combined to obtain a total genetic programming model.

[0068] The genetic programming formula table is shown in Table 1 and Table 2, Table 1 is an operation formula of numerical data, and Table 2 is an operation formula of a keyword:

[0069] Table 1

[0070]

[0071]

[0072] Table 2

[0073] Formula name Number of parameters Explanation ∨ n Or operator

[0074] For example, a time / amount formula tree is shown in Figure 3 , and the formula tree is:

[0075]

[0076] The formula is explained as follows: the insured person has a high hospitalization amount per day and pays a high premium, and it is very likely that the insurance fraud case is of the type of treating illness by taking medicine.

[0077] For example, a keyword formula tree is shown in Figure 4 , and the formula tree is:

[0078] tree_level=doubling ∨ falling ∨ accident ∨ slipping,

[0079] The formula indicates that there is a fraud risk in the occurrence of the keywords of falling or falling or accident or slipping in the claim data, and the risk factor is explained as some risk cases that are relatively easy to fake by the insured.

[0080] In step S203, a preset number of formula tree populations are generated by random initialization as the initial formula tree populations corresponding to the genetic programming model.

[0081] In this embodiment, a preset number of formula tree populations are generated by random initialization as the initial formula tree populations corresponding to the genetic programming model, and each formula tree in the formula tree population is formed by the operation of the operator on the variable splicing, wherein the variable is a numerical data such as an amount variable, a time variable, and a keyword data such as a claim history, a diagnosis result, and a text classification variable. Specifically, according to the complexity of the business and the limitation of the computing resources, the number of tree structures to be generated is determined, i.e., the size of the initial population is determined, and then a random number generator is used to generate the nodes and connection relationships of each formula tree structure, including operators and variables.

[0082] In step S204, the risk label data labeled in the insurance business database is obtained, the training set and the test set are divided according to the preset proportion, and the risk explanation degree of each formula tree in the initial formula tree population is calculated through the training set.

[0083] In this embodiment, the risk label data includes but is not limited to claim risk, credit risk, fraud risk, fund risk, and misinsurance risk. The risk label data is randomly divided into a training set and a test set according to a preset proportion, such as a ratio of 8:2. The training set obtained by division is trained by using the genetic programming model, and the risk identification accuracy of the variable logistic regression model is used as the risk explanation degree of the genetic programming model. The risk explanation degree represents the explanation ability of the formula to the risk, and the higher the risk explanation degree, the better the fitting degree of the risk factor and the risk label, and the more accurate the risk identification.

[0084] In this embodiment, each formula tree in the initial formula tree population represents a potential risk identification factor, and the variable logistic regression model is used to fit the risk factor represented by each formula tree with the risk label, and the ratio of the regression mean square to the error mean square is calculated to obtain the risk explanation degree of each formula tree.

[0085] Logistic regression is a simple machine learning method for solving binary classification problems. It outputs the probability that the data belongs to a certain category according to the given independent variable data. Since the value of this probability ranges from 0 to 1, the value of this probability can be mapped to the entire real number range by using the logit function:

[0086]

[0087] In the above formula, pi represents the probability of event occurrence, Logit() represents the logit function, Beta_0…Beta_k represents the regression coefficient, and X_1…X_k represents the value of the independent variable. The coefficients of the above formula are usually estimated by maximum likelihood estimation (MLE).

[0088] In this embodiment, the F value of the univariate logistic regression can be selected as the risk interpretation degree of the single risk identification factor.

[0089]

[0090] In the above formula, MSR represents the mean square of regression, MSE represents the mean square of error, SSR represents the regression sum of squares, SSE represents the residual sum of squares, and n represents the number of samples. The total variance SST sum of squares can be decomposed into the regression square SSR and the residual square sum SSE by variance decomposition. Dividing by the respective degrees of freedom, the regression mean square MER and the error mean square MSE are obtained. The ratio of the two measures the fitting degree of the logistic function to the dependent variable. The larger the F value, the better the fitting degree to the dependent variable, that is, the higher the risk interpretation degree of the risk identification factor, and the more likely it becomes the parent of the new population.

[0091] The risk interpretation degree of the formula tree is calculated by the variable logistic regression model, which can improve the efficiency and accuracy of the calculation, and better and faster obtain the optimal formula tree population.

[0092] In step S205, the initial formula tree population is iteratively optimized by a genetic algorithm according to the risk interpretation degree until a preset termination condition is reached, and the final formula tree population is obtained.

[0093] Specifically, according to the calculated risk interpretation degree, a population selection algorithm is used to select a second preset number of formula trees with the highest risk interpretation degree as parent formula trees; the parent formula trees are copied into a new population; through a preset crossover probability, a part of the parent formula trees are randomly selected for pairing, and the sub-trees of the parent formula trees are randomly selected for exchange and recombination to generate new crossover child formula trees; according to a preset mutation probability, the parent formula trees are subjected to mutation operation to generate new mutation child formula trees; the crossover child formula trees and the mutation child formula trees are added to the new population, and the iteration and evolution are continuously performed until a preset termination condition is reached. The preset termination condition is to reach a preset number of iterations or population convergence, and the population convergence refers to that the risk interpretation degree of the optimal formula changes by less than a preset threshold value for consecutive generations.

[0094] In the embodiment, the population selection algorithm is roulette selection, tournament selection or gene selection, and the roulette selection, the tournament selection or the gene selection is used to select the formula tree with the risk explanation degree higher than the preset explanation degree threshold from the current formula tree population as the parent formula tree. The principle of the roulette selection is to select the candidate formula tree according to the probability distribution, and the probability of being selected by the candidate is determined by the performance of the candidate relative to the entire formula population. This operation is similar to the survival of the fittest in nature, and the candidate with better performance has a higher probability of being selected to generate offspring. The principle of the tournament selection is to select two or more candidate formulas from the population at a time, and then select the candidate with better performance. Compared with the global method of roulette, this method gives the candidate with local potential the opportunity to enter the next round of generating offspring. The principle of the gene selection is to sort all formula trees according to the risk explanation degree, and then the formula tree with a higher risk explanation degree replaces the formula tree with poor performance. For example, the initial formula tree population is sorted according to the risk explanation degree, and the formula tree with a risk explanation degree higher than the preset explanation degree threshold is selected as the parent formula tree.

[0095] After the parent formula tree is selected, the copying, crossing and mutation operations are performed in sequence to optimize the combination of the risk identification factor and improve the population diversity and the risk explanation degree.

[0096] In some optional implementations of the embodiment, the step of copying the parent formula tree into the new population includes:

[0097] The tree structure data of the parent formula tree is obtained;

[0098] The depth-first search algorithm is used to traverse all nodes of the parent formula tree, the parent formula tree is converted into a string form through serialization operation, and the parent formula tree string is obtained;

[0099] The parent formula tree string is copied into a new buffer, and the same memory space as the parent formula tree is applied in the memory of the new population;

[0100] The parent formula tree string is deserialized into a tree structure, and the parent formula tree is reconstructed;

[0101] The reconstructed parent formula tree is inserted into the memory address corresponding to the new population;

[0102] It is judged whether all selected parent formula trees have been copied. If yes, the copying operation is completed, otherwise the next parent formula tree is selected for copying, and the copying operation is continued until all parent formula trees are copied.

[0103] The formula trees with high risk explanation degree are copied from the parent population to the new population by the copying operation, the population diversity is maintained, the evolution process is accelerated, and the excellent genetic information is reserved to provide a basis for subsequent crossover and mutation operations.

[0104] In some optional implementations, the step of generating new crossover offspring formula trees by randomly selecting part of the parent formula trees for pairing and randomly selecting the sub-trees of the parent formula trees for exchange recombination according to the preset crossover probability includes:

[0105] According to the preset crossover probability, it is judged whether to perform a crossover operation on the copied offspring formula trees in the new population.

[0106] If yes, two formula trees are randomly selected from the new population as a first parent and a second parent.

[0107] A first sub-tree and a second sub-tree are randomly selected in the first parent and the second parent, respectively.

[0108] It is judged whether the root node types of the first sub-tree and the second sub-tree match.

[0109] If the match, the positions of the first sub-tree and the second sub-tree in their respective parents are exchanged to obtain a first crossover offspring formula tree and a crossover offspring formula tree.

[0110] The crossover operation is specifically that the formula trees copied in the new population are randomly paired two by two, the sub-trees are randomly selected for recombination according to the preset crossover probability, and the new formula trees obtained enter the new offspring population.

[0111] The basic flow of the crossover step is: first, the formula tree with the highest risk explanation degree is selected as the parent of the population, and the sub-tree is randomly selected from the formula tree to replace the random sub-tree in the formula tree with the second highest explanation degree to generate a new offspring. For example, assuming that factors 1, 2, 3, and 4 represent real premium, claim amount, final case amount, and hospitalization days, respectively, the crossover evolution process of the formula tree is as shown in Figure 5 .

[0112] The root node types of the first sub-tree and the second sub-tree match, specifically, the two root node types match, that is, both are function nodes or both are terminal nodes.

[0113] If the root node types of the first sub-tree and the second sub-tree do not match, the sub-trees are reselected until the root node type matching condition is met.

[0114] In this embodiment, the above-mentioned crossover operation is repeated until the preset number of crossover operations is reached.

[0115] By cross operation, the excellent genetic information is reserved while introducing new risk identification factor combinations, thereby accelerating the evolution speed.

[0116] In some optional implementations, the step of generating new mutated offspring formula trees from the parent formula trees according to the preset mutation probability comprises:

[0117] If the parent formula tree is subjected to the subtree mutation operation, a subtree of the parent formula tree is randomly selected, and a new subtree is randomly generated to replace the subtree, so as to obtain a first formula tree after subtree mutation;

[0118] If the parent formula tree is not subjected to the subtree mutation operation, it is determined whether the parent formula tree is subjected to the point mutation operation, if yes, a node of the parent formula tree is randomly selected, and the node is replaced by a new node with the same number of variables, so as to obtain a second formula tree after point mutation;

[0119] If the parent formula tree is not subjected to the point mutation operation, it is determined whether the parent formula tree is subjected to the Hoist mutation operation, if yes, an initial subtree of the parent formula tree is randomly selected, a replacement subtree is randomly selected from the initial subtree, and the replacement subtree is replaced by the initial subtree, so as to obtain a third formula tree after Hoist mutation;

[0120] The first formula tree, the second formula tree and the third formula tree are used as the mutated offspring formula trees.

[0121] The mutation: through the set mutation probability, the partial structure of the formula tree is mutated, and the new formula tree is obtained and enters a new offspring population. The mutation in the genetic planning can be roughly divided into three categories: subtree mutation, Hoist mutation and subtree mutation. The subtree mutation is replaced by a randomly generated subtree, the formula tree with the highest risk explanation degree is selected as the parent, the subtree structure is randomly selected, and the new tree structure offspring is generated by replacing the randomly generated subtree; the point mutation guarantees the formula diversity in the population, so that the formula eliminated in the selection step has the opportunity to return to the new population. Specifically, the formula tree with the highest risk explanation degree is selected as the parent, the node and the leaf are randomly selected, and other formula trees with the same number of variables are randomly selected from the old population to replace the node and the leaf, so as to form the new offspring; the Hoist mutation simplifies the formula tree by pruning the generated formula tree structure, and reduces the overfitting. Specifically, the Hoist mutation randomly selects the subtree structure from the parent formula tree with the high risk explanation degree, and covers and lifts the random subtree in the subtree to the position of the original subtree, so as to generate the new offspring.

[0122] The mutation operation on the parent formula tree can increase the population diversity, improve the local search ability of the algorithm, and help to deeply mine the risk identification factors.

[0123] In the embodiment, the formula tree population is evolved by genetic operations such as selection, copying, crossing, mutation, etc., to obtain a final formula tree population, which can automatically mine deep risk identification factors that cannot be obtained by experience, enhance risk prediction capability, and improve risk assessment efficiency.

[0124] Step S206, a first preset number of formula trees with the highest risk explanation degree are selected from the final formula tree population as risk identification factors.

[0125] From the high-quality final formula tree population obtained by evolution, a number of formula trees with the highest risk explanation degree are extracted as candidate risk identification factors. The candidate risk identification factors are analyzed for correlation, and redundant factors with a correlation greater than a preset correlation threshold are removed, thereby excluding redundant variables and obtaining a final risk identification factor set. The risk identification factor set is used as a characteristic variable to establish a risk identification model such as a decision tree or a logistic regression, thereby realizing automated and accurate insurance risk identification.

[0126] Step S207, the risk identification accuracy of the risk identification factors is calculated using a test set, and when the risk identification accuracy is greater than or equal to a preset threshold, the genetic programming model corresponding to the risk identification factors is output as a risk identification model.

[0127] In the embodiment, the test set is used to calculate the value of the risk identification factors of each insurance data in the test set, and when the formula calculation result is True, it is predicted that the insurance data has risks, otherwise it is predicted that there are no risks. Accordingly, a risk identification label is generated, which is compared with the existing risk label in the test set to calculate the risk identification accuracy of the optimal risk explanation formula on the test set. When the risk identification accuracy is greater than or equal to a preset threshold, it indicates that the formula has strong risk explanation capability and can be used to guide the risk management practice of insurance business, and the genetic programming model corresponding to the risk identification factors is output as a risk identification model.

[0128] If the risk identification accuracy is less than the preset threshold, it indicates that the model performance does not meet the expectation, and then the parameter settings of the genetic programming model, such as population size, crossing probability, mutation probability, etc., are adjusted for retraining and optimization.

[0129] Step S208, obtaining the insurance business data to be identified, inputting the insurance business data to be identified into the risk identification model to obtain a risk identification result.

[0130] The new insurance business data is acquired, preprocessed, including removing noise data, filling in missing values, and standardizing data formats, to obtain new to-be-recognized insurance business data, numerical features are extracted, including premium, premium, age, and keyword features are extracted, including occupation, medical history, family history, etc. The extracted numerical features and keyword features are input into a risk identification model to obtain a risk identification result. The risk identification result, including the risk score and risk level, is written into the insurance business database and stored in association with the original new insurance business data. According to the risk identification result in the insurance business database, business decisions are made, high-risk businesses are audited, and normal-risk businesses are quickly approved, improving business processing efficiency.

[0131] The present application applies the genetic programming algorithm to the field of insurance risk identification, can automatically discover complex nonlinear risk identification factors, improves the accuracy and efficiency of risk identification, at the same time, by modeling different data types respectively, fully utilizes the multi-dimensional features of insurance data, can better realize the automatic identification of insurance risk, further improves the accuracy and interpretability of risk identification.

[0132] The embodiments of the present application can acquire and process related data based on artificial intelligence technology. Among them, artificial intelligence (Artificial Intelligence, AI) is to use digital computers or digital computer controlled machines to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. Theory, method, technology and application system.

[0133] The basic technology of artificial intelligence generally includes technologies such as sensors, special artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction system, mechatronics, etc. Artificial intelligence software technology mainly includes computer vision technology, robot technology, biometric technology, speech processing technology, natural language processing technology, and machine learning / deep learning, etc. A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by a computer readable instruction instructing related hardware, which can be stored in a computer readable storage medium. When the program is executed, it can include the processes of the above-mentioned embodiments. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory (Read-Only Memory, ROM) and other non-volatile storage media, or a random access memory (Random Access Memory, RAM) and the like.

[0134] It should be understood that although each step in the flowchart of the accompanying drawings is shown in sequence according to the direction of the arrow, these steps are not necessarily executed in sequence according to the direction of the arrow. Unless explicitly stated herein, the execution of these steps is not strictly limited in sequence, and they can be executed in other sequences. Moreover, at least part of the steps in the flowchart of the accompanying drawings can include multiple sub-steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence is not necessarily sequential, but can be alternately executed with at least part of other steps or sub-steps or stages of other steps.

[0135] Further referring to Figure 6 , as an implementation of the method shown in the above Figure 2 , the present application provides an embodiment of a risk identification factor mining device, which corresponds to the method embodiment shown in Figure 2 , and the device can be specifically applied to various electronic devices.

[0136] As shown in Figure 6 , the risk identification factor mining device 600 described in the embodiment includes an acquisition module 601, a construction module 602, an initialization module 603, a calculation module 604, an iteration module 605, a selection module 606, a test module 607, and an identification module 608. Among them:

[0137] The acquisition module 601 is configured to acquire unstructured text data and numerical data in an insurance business database, perform word segmentation processing on the text data to obtain a word segmentation result, extract keywords from the word segmentation result to obtain keyword data, and obtain the keyword data.

[0138] The construction module 602 is configured to obtain a first operation formula based on a logical operator set for the keyword data, obtain a second operation formula based on a basic operator set for the numerical data, and construct a genetic programming model according to an insurance risk identification target through the first operation formula and the second operation formula.

[0139] The initialization module 603 is configured to generate a formula tree population of a preset number of formula trees through a random initialization manner as an initial formula tree population corresponding to the genetic programming model.

[0140] The calculation module 604 is configured to acquire labeled risk label data in the insurance business database, divide a training set and a test set according to a preset ratio, and calculate a risk explanation degree of each formula tree in the initial formula tree population through the training set.

[0141] The iteration module 605 is configured to iteratively optimize the initial formula tree population according to the risk interpretation degree by using a genetic algorithm until a preset termination condition is reached, to obtain a final formula tree population;

[0142] The selection module 606 is configured to select a first preset number of formula trees with the highest risk interpretation degree from the final formula tree population as risk identification factors;

[0143] The test module 607 is configured to calculate a risk identification accuracy of the risk identification factors using the test set, and output a genetic programming model corresponding to the risk identification factors as a risk identification model when the risk identification accuracy is greater than or equal to a preset threshold.

[0144] The identification module 608 is configured to obtain to-be-identified insurance business data, input the to-be-identified insurance business data into the risk identification model, and obtain a risk identification result.

[0145] It should be emphasized that, in order to further ensure the privacy and security of the text data and the numerical data, the text data and the numerical data can also be stored in a node of a block chain.

[0146] Based on the above-mentioned risk identification factor mining device 600, by applying a genetic programming algorithm to the field of insurance risk identification, complex non-linear risk identification factors can be automatically discovered, the accuracy and efficiency of risk identification are improved, and at the same time, by modeling different data types respectively, the multi-dimensional features of insurance data are fully utilized, which can better realize the automatic identification of insurance risks and further improve the accuracy and interpretability of risk identification.

[0147] In the embodiment, the acquisition module 601 includes a word segmentation sub-module configured to:

[0148] According to the word segmentation result, the number of occurrences of each word segmentation in the text data is calculated by a word frequency statistical method, to obtain the word frequency of each word segmentation;

[0149] The distribution of each word segmentation in the text data is calculated by using an inverse document frequency algorithm, to obtain the inverse document frequency;

[0150] The weight value of each word segmentation is calculated by multiplying the word frequency and the inverse document frequency, to obtain a word weight list;

[0151] A preset word frequency threshold and a document frequency threshold are obtained, and word segmentations in the word weight list that simultaneously satisfy the conditions of the word frequency being greater than or equal to the word frequency threshold and the document frequency being less than or equal to the document frequency threshold are screened out as candidate keywords;

[0152] The candidate keywords are sorted in descending order of the weight values, and a preset number of candidate keywords in the front of the sorting are selected as keyword data.

[0153] The term frequency, inverse document frequency and weight value are used to select the final keyword data, improve the keyword extraction efficiency, eliminate non-keywords with low weight, make the extracted keywords have strong explanatory power in risk identification, and further improve the accuracy of risk identification.

[0154] In some optional implementations of the embodiment, the calculation module 604 is further configured to:

[0155] The variable logistic regression model is used to fit each risk factor represented by the formula tree and the risk label, and the risk explanation degree of each formula tree is obtained by calculating the ratio of the regression mean square to the error mean square.

[0156] The risk explanation degree of the formula tree is calculated by the variable logistic regression model, which can improve the efficiency and accuracy of the calculation, and better and faster obtain the optimal formula tree population.

[0157] In some optional implementations, the iteration module 605 includes:

[0158] The selection submodule is configured to select the second preset number of formula trees with the highest risk explanation degree as parent formula trees according to the calculated risk explanation degree, using a population selection algorithm;

[0159] The replication submodule is configured to perform a replication operation on the parent formula trees and replicate them into a new population;

[0160] The crossover submodule is configured to randomly select part of the parent formula trees for pairing through a preset crossover probability, and randomly select the sub-trees of the parent formula trees for exchange and recombination to generate new crossover child formula trees;

[0161] The mutation submodule is configured to perform a mutation operation on the parent formula trees according to a preset mutation probability to generate new mutation child formula trees;

[0162] The iteration submodule is configured to add the crossover child formula trees and the mutation child formula trees to the new population, and continuously iterate and evolve until the preset termination condition is reached.

[0163] Through genetic operations such as selection, replication, crossover and mutation, the formula tree population is evolved to obtain the final formula tree population, which can automatically mine deep risk identification factors that cannot be obtained by experience, enhance the risk prediction ability, and improve the risk assessment efficiency.

[0164] In some optional implementations of the embodiment, the replication submodule is further configured to:

[0165] Obtain the tree structure data of the parent formula trees;

[0166] Traverse all nodes of the parent formula tree by using a depth-first search algorithm, convert the parent formula tree into a string form by a serialization operation, and obtain a parent formula tree string;

[0167] Copy the parent formula tree string into a new buffer, and apply a memory space with the same size as the parent formula tree in the memory of the new population;

[0168] Deserialize the parent formula tree string into a tree structure, and reconstruct the parent formula tree;

[0169] Insert the reconstructed parent formula tree into the corresponding memory address of the new population;

[0170] Determine whether all selected parent formula trees have been copied, if yes, the copying operation is completed, otherwise, return to select the next parent formula tree for copying, until all parent formula trees are copied.

[0171] The copying operation realizes copying of formula trees with high risk explanation degree from the parent population to the new population, maintains population diversity, accelerates the evolution process, and preserves excellent genetic information, thereby providing a basis for subsequent crossover and mutation operations.

[0172] In some optional implementation manners of the embodiment, the crossover submodule is further used to:

[0173] According to a preset crossover probability, determine whether to perform a crossover operation on the copied offspring formula trees in the new population;

[0174] If yes, randomly select two formula trees from the new population as a first parent and a second parent;

[0175] Randomly select a first subtree and a second subtree from the first parent and the second parent, respectively;

[0176] Determine whether the root node types of the first subtree and the second subtree match;

[0177] If the root node types match, exchange the positions of the first subtree and the second subtree in their respective parents, and obtain a first crossover offspring formula tree and a crossover offspring formula tree.

[0178] Through the crossover operation, new risk identification factor combinations are introduced while excellent genetic information is preserved, thereby accelerating the evolution speed.

[0179] In some optional implementation manners of the embodiment, the mutation submodule is further used to:

[0180] Determine whether to perform a subtree mutation operation on the parent formula tree. If so, randomly select a subtree of the parent formula tree and replace it with a randomly generated new subtree to obtain the first formula tree after subtree mutation.

[0181] If no subtree mutation operation is performed on the parent formula tree, then it is determined whether to perform a point mutation operation on the parent formula tree. If so, a node of the parent formula tree is randomly selected and replaced with a new node with the same number of variables to obtain the second formula tree after point mutation.

[0182] If no point mutation operation is performed on the parent formula tree, then it is determined whether to perform a Hoist mutation operation on the parent tree. If so, an initial subtree of the parent formula tree is randomly selected, and a replacement subtree is randomly selected from the initial subtree. The replacement subtree replaces the initial subtree to obtain the third formula tree after Hoist mutation.

[0183] The first formula tree, the second formula tree, and the third formula tree are used as mutated offspring formula trees.

[0184] By performing mutation operations on the parent formula tree, we can increase population diversity, improve the local search capability of the algorithm, and help to deeply explore risk identification factors.

[0185] To address the aforementioned technical problems, embodiments of this application also provide a computer device. Please refer to [link / reference needed]. Figure 7 , Figure 7 This is a basic structural block diagram of the computer device in this embodiment.

[0186] The computer device 7 includes a memory 71, a processor 72, and a network interface 73 that are interconnected via a system bus. It should be noted that only a computer device 7 with a memory 71, a processor 72, and a network interface 73 is shown in the figure; however, it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented alternatively. Those skilled in the art will understand that the computer device described here is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0187] The computer device can be a desktop computer, a notebook computer, a palm computer, a cloud server, or the like. The computer device can interact with a user through a keyboard, a mouse, a remote controller, a touchpad, a voice control device, or the like.

[0188] The memory 71 includes at least one type of readable storage medium, such as a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., an SD or DX memory, or the like), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, or the like. In some embodiments, the memory 71 can be an internal storage unit of the computer device 7, such as a hard disk or a memory of the computer device 7. In other embodiments, the memory 71 can also be an external storage device of the computer device 7, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, or the like. Of course, the memory 71 can include both an internal storage unit and an external storage device of the computer device 7. In this embodiment, the memory 71 is generally used to store an operating system and various application software installed in the computer device 7, such as computer readable instructions of the mining method of risk identification factors, or the like. In addition, the memory 71 can also be used to temporarily store various data that have been output or will be output.

[0189] The processor 72 can be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip in some embodiments. The processor 72 is generally used to control the overall operation of the computer device 7. In this embodiment, the processor 72 is used to run computer readable instructions or process data stored in the memory 71, such as computer readable instructions of the mining method of risk identification factors.

[0190] The network interface 73 can include a wireless network interface or a wired network interface, and is generally used to establish a communication connection between the computer device 7 and other electronic devices.

[0191] By applying the genetic programming algorithm to the insurance risk identification field, complex nonlinear risk identification factors can be automatically discovered, the accuracy and efficiency of risk identification are improved, and meanwhile, by modeling different data types respectively, the multi-dimensional characteristics of insurance data are fully utilized, the automatic identification of insurance risks is better achieved, and the accuracy and interpretability of risk identification are further improved.

[0192] The application also provides another implementation, namely providing a computer readable storage medium, the computer readable storage medium stores computer readable instructions, the computer readable instructions can be executed by at least one processor, so that the at least one processor executes the steps of the risk identification factor mining method as described above.

[0193] By applying the genetic programming algorithm to the insurance risk identification field, complex nonlinear risk identification factors can be automatically discovered, the accuracy and efficiency of risk identification are improved, and meanwhile, by modeling different data types respectively, the multi-dimensional characteristics of insurance data are fully utilized, the automatic identification of insurance risks is better achieved, and the accuracy and interpretability of risk identification are further improved.

[0194] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment method can be realized by means of software and the necessary general hardware platform, of course, it can also be realized by hardware, but in many cases, the former is a better implementation. Based on such understanding, the technical solutions of the application can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a plurality of instructions for making a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) execute the method described in each embodiment of the application.

[0195] Obviously, the above-described embodiments are only some of the embodiments of the application, not all the embodiments, and the preferred embodiments of the application are given in the drawings, but do not limit the patent scope of the application. The application can be realized in many different forms, and conversely, the purpose of providing these embodiments is to make the disclosure of the application more thorough and comprehensive. Although the application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions recorded in the foregoing specific embodiments, or make equivalent replacements to some technical features. Any equivalent structure made by using the contents of the specification and drawings, directly or indirectly applied to other related technical fields, is also within the scope of the patent protection of the application.

Claims

1. A method of mining risk identification factors, characterized by, The method comprises the following steps: obtaining unstructured text data and numerical data in an insurance business database, performing word segmentation on the text data to obtain a word segmentation result, extracting keywords from the word segmentation result to obtain keyword data; obtaining a first operation formula based on a logical operator set according to the keyword data, obtaining a second operation formula based on a basic operator set according to the numerical data, and constructing a genetic programming model according to an insurance risk identification target through the first operation formula and the second operation formula; generating a formula tree population of a preset number of formula trees through random initialization as an initial formula tree population corresponding to the genetic programming model; obtaining labeled risk label data in the insurance business database, dividing the labeled risk label data into a training set and a test set according to a preset ratio, and calculating a risk explanation degree of each formula tree in the initial formula tree population through the training set; iteratively optimizing the initial formula tree population through a genetic algorithm according to the risk explanation degree until a preset termination condition is reached to obtain a final formula tree population; selecting a first preset number of formula trees with the highest risk explanation degree from the final formula tree population as risk identification factors; calculating a risk identification accuracy of the risk identification factors using the test set, and outputting a genetic programming model corresponding to the risk identification factors as a risk identification model when the risk identification accuracy is greater than or equal to a preset threshold; obtaining to-be-identified insurance business data, inputting the to-be-identified insurance business data into the risk identification model, and obtaining a risk identification result.

2. The method of claim 1, wherein the risk identification factor is a risk of a patient. The step of extracting keywords from the word segmentation result to obtain keyword data comprises: calculating the frequency of each word segmentation in the text data according to the word segmentation result through a word frequency statistical method to obtain the word frequency of each word segmentation; calculating the distribution of each word segmentation in the text data using an inverse document frequency algorithm to obtain the inverse document frequency; multiplying the word frequency and the inverse document frequency to calculate the weight value of each word segmentation to obtain a word weight list; obtaining a preset word frequency threshold and a document frequency threshold, and screening word segmentations in the word weight list that simultaneously satisfy the word frequency being greater than or equal to the word frequency threshold and the document frequency being less than or equal to the document frequency threshold as candidate keywords; sorting the candidate keywords from high to low according to the weight values, and selecting a preset number of candidate keywords in the front of the sorting as keyword data.

3. The method of claim 1, wherein the risk identification factor is a risk of a patient. The step of calculating the risk explanation degree of each formula tree in the initial formula tree population through the training set comprises: fitting a risk factor represented by each formula tree and a risk label using a variable logistic regression model, and calculating the risk explanation degree of each formula tree by calculating the ratio of the regression mean square to the error mean square.

4. The method of claim 1, wherein the risk identification factor is a risk of a patient. The step of iteratively optimizing the initial formula tree population through a genetic algorithm according to the risk explanation degree until a preset termination condition is reached comprises: selecting a second preset number of formula trees with the highest risk explanation degree as parent formula trees using a population selection algorithm according to the calculated risk explanation degree; performing a copy operation on the parent formula trees and copying them into a new population; The parent formula trees are randomly selected for pairing and randomly selected sub-trees of the parent formula trees are randomly selected for recombination by exchange according to a preset crossover probability, to generate new crossover offspring formula trees; The parent formula trees are subjected to mutation operation according to a preset mutation probability, to generate new mutation offspring formula trees; The crossover offspring formula trees and the mutation offspring formula trees are added to the new population, and iterative evolution is continuously performed until a preset termination condition is reached.

5. The method of claim 4, wherein the risk identification factor is determined by the steps of: The step of copying the parent formula trees into the new population includes: ​ Obtaining tree structure data of the parent formula trees; All nodes of the parent formula trees are traversed by using a depth-first search algorithm, and the parent formula trees are converted into string form by serialization operation, to obtain parent formula tree strings; The parent formula tree strings are copied into a new buffer, and a memory space with the same size as the parent formula trees is applied in the memory of the new population; The parent formula tree strings are deserialized into tree structure, to reconstruct the parent formula trees; The reconstructed parent formula trees are inserted into the corresponding memory addresses of the new population; It is determined whether all selected parent formula trees have been copied, if yes, the copying operation is completed, otherwise the next parent formula tree is selected for copying, until all parent formula trees are copied.

6. The method of claim 5, wherein the risk identification factor is determined by the steps of: The step of randomly selecting part of the parent formula trees for pairing and randomly selecting sub-trees of the parent formula trees for recombination by exchange according to a preset crossover probability, to generate new crossover offspring formula trees includes: ​ It is determined whether to perform crossover operation on the copied offspring formula trees in the new population according to a preset crossover probability; If yes, two formula trees are randomly selected from the new population as a first parent and a second parent; A first sub-tree and a second sub-tree are randomly selected in the first parent and the second parent, respectively; It is determined whether the root node types of the first sub-tree and the second sub-tree match; If yes, the positions of the first sub-tree and the second sub-tree in their respective parents are exchanged, to obtain a first crossover offspring formula tree and a crossover offspring formula tree.

7. The method of claim 4, wherein the risk identification factor is a risk of a patient. The step of performing mutation operation on the parent formula trees according to a preset mutation probability, to generate new mutation offspring formula trees includes: It is determined whether to perform sub-tree mutation operation on the parent formula trees, if yes, a sub-tree of the parent formula trees is randomly selected, and the sub-tree is replaced by a newly generated sub-tree, to obtain a first formula tree after sub-tree mutation; If no sub-tree mutation operation is performed on the parent formula trees, it is determined whether to perform point mutation operation on the parent formula trees, if yes, a node of the parent formula trees is randomly selected, and the node is replaced by a new node with the same number of variables, to obtain a second formula tree after point mutation; If the point mutation operation is not performed on the parent formula tree, it is determined whether a Hoist mutation operation is performed on the parent tree, if yes, an initial subtree of the parent formula tree is randomly selected, an alternative subtree is randomly selected from the initial subtree, the initial subtree is replaced by the alternative subtree, and a third formula tree after Hoist mutation is obtained; The first formula tree, the second formula tree and the third formula tree are used as mutated offspring formula trees.

8. A risk identification factor mining device characterized by comprising: Comprise: The acquisition module is configured to acquire unstructured text data and numerical data in an insurance business database, perform word segmentation processing on the text data to obtain a word segmentation result, and extract keywords from the word segmentation result to obtain keyword data; The construction module is configured to obtain a first operation formula based on a logical operator set according to the keyword data, obtain a second operation formula based on a basic operator set according to the numerical data, and construct a genetic programming model according to an insurance risk identification target and the first operation formula and the second operation formula; The initialization module is configured to generate a formula tree population of a preset number of formula trees by random initialization as an initial formula tree population corresponding to the genetic programming model; The calculation module is configured to acquire annotated risk label data in the insurance business database, divide a training set and a test set according to a preset proportion, and calculate a risk explanation degree of each formula tree in the initial formula tree population by using the training set; The iteration module is configured to iteratively optimize the initial formula tree population by using a genetic algorithm according to the risk explanation degree until a preset termination condition is reached, and obtain a final formula tree population; The selection module is configured to select a first preset number of formula trees with the highest risk explanation degrees from the final formula tree population as risk identification factors; The test module is configured to calculate a risk identification accuracy of the risk identification factors by using the test set, and output a genetic programming model corresponding to the risk identification factors as a risk identification model when the risk identification accuracy is greater than or equal to a preset threshold. The identification module is configured to acquire to-be-identified insurance business data, input the to-be-identified insurance business data into the risk identification model, and obtain a risk identification result.

9. A computer device, comprising: The device comprises a memory and a processor, the memory stores computer readable instructions, and the processor implements the steps of the risk identification factor mining method according to any one of claims 1 to 7 when executing the computer readable instructions.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer readable instructions, and the computer readable instructions are executed by the processor to implement the steps of the risk identification factor mining method according to any one of claims 1 to 7.

Images