User behavior prediction method and device, electronic device, and storage medium

Abstract

Embodiments of the present application provide a user behavior prediction method and device, an electronic device and a storage medium, and belong to the technical field of artificial intelligence. The method comprises: randomly generating a plurality of user features according to target user data; filtering out candidate features from the plurality of user features by using a null importance algorithm; processing the candidate features according to an evolutionary method to obtain offspring features; iteratively processing the offspring features until a preset condition is met to stop iteration, and obtaining a final feature combination for predicting user behavior. Based on this, the present application can automatically complete the combination analysis of features for user insurance premium payment behavior prediction.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a user behavior prediction method and apparatus, electronic device and storage medium. Background Technology

[0002] In insurance premium forecasting, the first step should be to determine the business objectives the forecast should support. Businesses most commonly use premium forecasting for one of the following objectives: calculating potential premium payment dates for users, predicting possible payment dates such as D-5, D-10, and D-20. This involves calculating the probability of a user making the payment on time: what is the probability that the user will make the payment on time, and outputting a probability value for each user.

[0003] Currently, predicting user insurance premiums requires establishing a predictive or ranking model to clearly define business objectives. However, the prediction of these objectives is often limited by insufficient or sparse target features. Therefore, it is frequently necessary to combine features as new inputs to train the model. In this case, relying solely on manual feature selection and combination is insufficient to meet the requirements of rapid model training and deployment. Therefore, how to automatically perform feature combination analysis for predicting user insurance premium behavior has become an urgent technical problem to be solved. Summary of the Invention

[0004] The main objective of this application is to propose a user behavior prediction method, device, electronic device, and storage medium that can automatically perform combined feature analysis for predicting user insurance payment behavior.

[0005] To achieve the above objectives, a first aspect of this application proposes a user behavior prediction method, the method comprising:

[0006] Multiple user characteristics are randomly generated based on the target user data;

[0007] The null importance algorithm is used to filter candidate features from multiple user features;

[0008] The candidate features are processed using an evolutionary method to obtain offspring features;

[0009] The child features are iteratively processed until a preset condition is met, at which point the iteration stops, resulting in the final feature combination used to predict user behavior.

[0010] In some embodiments, the null importance algorithm is used to filter candidate features from multiple user features, including:

[0011] The fitness of each user feature is calculated using the null importance algorithm, wherein the fitness is used to characterize the importance and stability of the user feature;

[0012] User features with fitness values ​​higher than a preset threshold are selected as candidate features.

[0013] In some embodiments, calculating the fitness of each user feature using the null importance algorithm includes:

[0014] Each user feature and its corresponding data label are repeatedly input into a preset tree model to obtain the first importance value of each user feature.

[0015] The data labels are shuffled and then input into the tree model repeatedly to obtain a second importance value for each user feature;

[0016] The stability value of each user feature is determined based on the first importance value and the second importance value obtained each time;

[0017] The fitness of each user feature is determined based on the stability value of each user feature.

[0018] In some embodiments, the step of filtering user features whose fitness is higher than a preset threshold as candidate features includes:

[0019] By repeatedly comparing the first importance value and the second importance value of each user feature, a stability value for each user feature is obtained.

[0020] User features whose stability value is higher than a preset threshold are selected as candidate features.

[0021] In some embodiments, the process of repeatedly comparing the first importance value and the second importance value of each user feature to obtain a stability value for each user feature includes:

[0022] The difference between the first importance value and the second importance value of the user feature obtained each time is calculated to obtain multiple importance difference values;

[0023] The stability value of each user feature is obtained by averaging the multiple importance differences.

[0024] In some embodiments, the iterative processing of the child features until a preset condition is met, resulting in a final feature combination for predicting user behavior, includes:

[0025] The child features are iteratively processed until the number of iterations reaches a preset threshold, at which point the iteration stops, resulting in the final feature combination used to predict user behavior.

[0026] In some embodiments, the evolutionary method includes at least one of heredity, variation, natural selection, and hybridization.

[0027] To achieve the above objectives, a second aspect of this application provides a user behavior prediction device, the device comprising:

[0028] The generation module is used to randomly generate multiple user features based on the target user data.

[0029] A filtering module is used to filter candidate features from multiple user features using a null importance algorithm;

[0030] An evolution module is used to process the candidate features according to an evolution method to obtain offspring features;

[0031] An iterative module is used to iteratively process the child features until a preset condition is met, at which point the iteration stops, resulting in the final feature combination used to predict user behavior.

[0032] To achieve the above objectives, a third aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method described in the first aspect.

[0033] To achieve the above objectives, a fourth aspect of the present application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in the first aspect.

[0034] This application proposes a user behavior prediction method, apparatus, electronic device, and storage medium. It randomly generates multiple user features based on target user data; employs a null importance algorithm to filter candidate features from these user features; processes the candidate features using an evolutionary method to obtain offspring features; it iteratively processes the offspring features until a preset condition is met, at which point the iteration stops, resulting in a final feature combination for predicting user behavior. Based on this, this application introduces a null importance algorithm to remove noisy user features and automatically extracts truly important user features as candidate features, thereby improving the success rate of identifying key features. Through multiple evolutionary iterations, it finds the final feature combination for predicting user insurance payment behavior. Therefore, compared to existing technologies that rely on manual analysis and feature combination searching, this application can automatically perform feature combination analysis for predicting user insurance payment behavior, greatly improving efficiency. This application can be applied to automatic feature evaluation, automatic model training, evaluation, and deployment, significantly shortening the overall model deployment time to adapt to increasingly complex user behavior prediction. Attached Figure Description

[0035] Figure 1 This is a flowchart of the user behavior prediction method provided in the embodiments of this application;

[0036] Figure 2 yes Figure 1 The flowchart of step S102 in the document;

[0037] Figure 3 yes Figure 1 The flowchart of step S201 in the text;

[0038] Figure 4 yes Figure 3 The flowchart of step S202 in the text;

[0039] Figure 5 yes Figure 3 The flowchart of step S401 in the text;

[0040] Figure 6 yes Figure 5 The flowchart of step S104 in the process;

[0041] Figure 7 This is a schematic diagram of the user behavior prediction device provided in the embodiments of this application;

[0042] Figure 8 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0044] It should be noted that although functional modules are divided in the device schematic diagram and a 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 aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0045] 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 this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0046] First, let's analyze some of the terms used in this application:

[0047] Artificial intelligence (AI) is a new branch of computer science that studies, develops, and applies theories, methods, technologies, and systems to simulate, extend, and expand human intelligence. It aims to understand the essence of intelligence and produce intelligent machines that can react in a way similar to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing, and expert systems. AI can simulate the information processes of human consciousness and thought. Furthermore, AI utilizes digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceiving the environment, acquiring knowledge, and using that knowledge to achieve optimal results.

[0048] Insurance refers to a commercial insurance activity in which the policyholder pays premiums to the insurer according to the contract, and the insurer assumes the responsibility of compensating for property losses caused by accidents that may occur as stipulated in the contract, or assumes the responsibility of paying insurance benefits when the insured dies, becomes disabled, falls ill, or reaches the age or term stipulated in the contract.

[0049] Null importance is a heuristic algorithm that modifies the importance measure of biased features. In general, we assume an algorithm that evaluates the relevance of a set of features relative to a response vector. The PIMP (Permutation Importance) algorithm performs multiple permutations on the response vector. For each permutation of the response vector, the relevance of all predictor variables is evaluated. The importance measure for each variable is obtained, called null importance.

[0050] Genetic algorithms are search algorithms used in computational mathematics to solve optimization problems; they are a type of evolutionary algorithm. Evolutionary algorithms were initially developed by drawing inspiration from phenomena in evolutionary biology, including heredity, mutation, natural selection, and hybridization. Genetic algorithms are typically implemented through computer simulation. For an optimization problem, a population of abstract representations (called chromosomes) of a certain number of candidate solutions (individuals) evolves towards better solutions. Traditionally, solutions are represented in binary (i.e., strings of 0s and 1s), but other representations can also be used. Evolution begins with a population of completely random individuals and proceeds generation by generation. In each generation, the fitness of the entire population is evaluated, and several individuals are randomly selected from the current population (based on their fitness) to generate a new population of life through natural selection and mutation. This new population becomes the current population in the next iteration of the algorithm.

[0051] Crossover: A subtree from the winning formula tree is randomly selected and replaced by a random subtree from another formula tree. This other formula tree is usually the one with the highest fitness among the remaining formula trees. In the algorithm model, the probability of subtree crossover is controlled by the p_cross_over parameter.

[0052] Subtree Mutation: This is a more aggressive mutation strategy: one of the winner's subtrees is replaced by a completely random new subtree. In the algorithm model, the probability of subtree mutation is controlled by the p_subtree_mutation parameter.

[0053] Hoist mutation is a method to combat formula tree bloating (i.e., excessive complexity): A subtree A is randomly selected from the winning formula tree, and then a subtree B is randomly selected from A. B is then promoted to the original position of A, replacing A. "Hoist" means "to elevate" or "to lift." In the algorithm model, the probability of subtree mutation is controlled by the parameter `p_hoist_mutation`.

[0054] Point mutation occurs when a random node is altered; for example, addition can be replaced with division, or the variable X0 can be replaced with the constant -2.5. Point mutation can reintroduce previously discarded functions and variables, thus promoting formula diversity. In the algorithm model, the probability of point mutation is controlled by the p_point_replace parameter.

[0055] Fitness refers to the relative ability of an individual with a known genotype to pass on its genes to its offspring's gene pool under certain environmental conditions. It is a measure of an individual's survival and reproductive chances. Although survival is a prerequisite for reproduction, survival without reproduction is meaningless for evolution; therefore, reproduction (gene continuation) is more essential. In revising Darwin's theory and reinterpreting the principle of natural selection, modern synthesis replaced "survival" with "reproduction" to measure fitness. "Survival of the fittest" became "reproduction of the fittest" (Mayr, 1977), and the new concept of fitness was used to quantitatively represent the degree of adaptation.

[0056] Iteration: An activity involving repeated feedback processes, typically aimed at approximating a desired goal or result. Each repetition of the process is called an "iteration," and the result of each iteration serves as the initial value for the next. It involves repeatedly executing a series of computational steps, deriving subsequent quantities from preceding ones. Each result in this process is obtained by performing the same computational steps on the previous result. For example, iterative methods can be used to solve a mathematical problem. In computer programs, repeating a subroutine (a set of instructions) that needs to be executed repeatedly, i.e., repeatedly executing a loop in the program until a certain condition is met, is also called iteration.

[0057] Tree models are a type of model with a tree-like branching structure based on feature space partitioning. Tree models are characterized by high variance, no requirement for units, consist of multiple rules, can handle both numerical and categorical data, and have high interpretability. Tree models require relatively small amounts of data, and their algorithm time complexity is the logarithm of the data points used to train the decision tree. They can handle both numerical and categorical data and, compared to neural networks, offer stronger interpretability.

[0058] Based on this, embodiments of this application provide a user behavior prediction method, apparatus, electronic device, and storage medium. The method randomly generates multiple user features based on target user data; uses a null importance algorithm to filter candidate features from these user features; processes the candidate features using an evolutionary method to obtain offspring features; iterates through the offspring features until a preset condition is met, at which point the iteration stops, resulting in a final feature combination for predicting user behavior. This application introduces a null importance algorithm to remove noisy user features and automatically extracts truly important user features as candidate features, thereby improving the success rate of identifying key features. Through multiple evolutionary iterations, a final feature combination is found for predicting user insurance payment behavior. Therefore, compared to existing technologies that rely on manual analysis and feature combination searching, this application can automatically perform feature combination analysis for predicting user insurance payment behavior, greatly improving efficiency. This application can be applied to automatic feature evaluation, automatic model training and evaluation, and deployment, significantly shortening the overall model deployment time to adapt to increasingly complex user behavior prediction.

[0059] The user behavior prediction method, apparatus, electronic device, and storage medium provided in this application are specifically described through the following embodiments. First, the user behavior prediction method in this application is described.

[0060] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0061] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0062] The user behavior prediction method provided in this application relates to the field of artificial intelligence technology. This method can be applied to a terminal, a server, or software running on either a terminal or a server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, etc.; the server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application implementing the user behavior prediction method, but is not limited to the above forms.

[0063] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: 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, and distributed computing environments including any of the above systems or devices. This 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 types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0064] It should be noted that in all specific embodiments of this application, when processing data related to user identity or characteristics, such as user information, user behavior data, user historical data, and user location information, user permission or consent is obtained first. Furthermore, the collection, use, and processing of this data comply with relevant laws, regulations, and standards. In addition, when embodiments of this application require access to sensitive personal information of users, separate permission or consent from the user is obtained through pop-ups or redirection to confirmation pages. Only after obtaining the user's separate permission or consent is the necessary user-related data required for the proper functioning of these embodiments acquired.

[0065] Figure 1 This is an optional flowchart of the user behavior prediction method provided in the embodiments of this application. Figure 1 The method may include, but is not limited to, steps S101 to S106.

[0066] Step S101: Randomly generate multiple user features based on the target user data;

[0067] Step S102: The null importance algorithm is used to select candidate features from multiple user features;

[0068] Step S103: Process the candidate features according to the evolutionary method to obtain offspring features;

[0069] Step S104: Iterate through the child features until a preset condition is met, then stop iterating to obtain the final feature combination used to predict user behavior.

[0070] In step S101 of some embodiments, multiple user features are randomly generated based on the target user data. These user features may include, but are not limited to, features that can be used to predict user insurance payment behavior, such as the insurer's city, income, insurance type, age, and occupation.

[0071] In step S102 of some embodiments, a null importance algorithm is used to filter candidate features from multiple user features. The null importance algorithm removes features that are considered noise and extracts truly important features. It should be noted that the null importance of each feature is measured using training data, in which the target variable is randomly shuffled. If a feature is very important, its importance in the correct label is high, but its importance in the shuffled label model is low. Comparing these two factors allows for the selection of stable features. Based on this principle, the null importance algorithm can be used to filter truly important user features from multiple user features and use these truly important user features as candidate features, where candidate features can also be called parent features.

[0072] In step S103 of some embodiments, candidate features are processed according to an evolutionary method to obtain offspring features. This evolutionary method includes, but is not limited to, calculations using heredity, mutation, natural selection, and crossover. The corresponding evolutionary calculation methods here are genetic algorithms such as crossover, subtree mutation, point mutation, and Hoist mutation to obtain offspring features. By achieving the evolution of the expression of individuals within each generation of the population through the above methods, the "survival of the fittest" forest law is realized, ultimately obtaining the optimal factor combination expression. This simultaneously solves the problems of factor selection and factor combination methods, automatically completing factor combination analysis and evaluating feature importance. By combining features through a genetic programming scheme, these features are automatically combined in insurance premium prediction, and the combined features are selected. This application can be applied to automatic feature evaluation, automatic model training, evaluation, and deployment, significantly shortening the overall model deployment time to adapt to increasingly diverse user behaviors.

[0073] In step S104 of some embodiments, the child features are iteratively processed until a preset condition is met, at which point the iteration stops, resulting in a final feature combination used to predict user behavior. After obtaining the child features, the next round of iterative training begins, stopping when a preset condition is met, for example, when a preset number of iterations has been reached. The final feature combination is then output and used to predict user behavior. In an insurance premium prediction scenario, the user's possible premium payment time and probability can be predicted based on the final feature combination.

[0074] Steps S101 to S104 of this application embodiment involve: randomly generating multiple user features based on target user data; using a null importance algorithm to select candidate features from the multiple user features; processing the candidate features according to an evolutionary method to obtain offspring features; iteratively processing the offspring features until a preset condition is met, at which point the iteration stops, resulting in the final feature combination used to predict user behavior. Based on this, this application introduces the null importance algorithm to remove user features that become noise and automatically extracts truly important user features as candidate features, thereby improving the success rate of identifying key features. Through multiple evolutionary iterations, the final feature combination is found for predicting user insurance payment behavior. Therefore, compared to the prior art's manual analysis and feature combination search, this application can automatically complete the feature combination analysis for predicting user insurance payment behavior, greatly improving efficiency. This application can be applied to automatic feature evaluation, automatic model training, evaluation, and deployment, significantly shortening the overall model deployment time to adapt to increasingly changing user behavior prediction.

[0075] Please see Figure 2In some embodiments, step S102 may include, but is not limited to, steps S201 to S202:

[0076] Step S201: The null importance algorithm is used to calculate the fitness of each user feature, where fitness is used to characterize the importance and stability of the user feature;

[0077] Step S202: Select user features with fitness values ​​higher than a preset threshold as candidate features.

[0078] In step S201 of some embodiments, the null importance algorithm is used to calculate the fitness of each user feature, where fitness is used to characterize the importance and stability of the user feature. That is, the higher the fitness of a user feature, the higher its importance and stability; conversely, the lower the fitness of a user feature, the lower its importance and stability.

[0079] In step S202 of some embodiments, user features with fitness values ​​higher than a preset threshold are selected as candidate features. Since there are multiple user features, to ensure that the selected user features are truly important, an appropriate threshold can be preset. Only user features with fitness values ​​higher than the preset threshold are selected as candidate features. Based on this, this application can remove user features that become noise and automatically extract truly important user features as candidate features, thereby improving the success rate of identifying key features.

[0080] Please see Figure 3 In some embodiments, step S201 may include, but is not limited to, steps S301 to S304:

[0081] Step S301: Repeatedly input each user feature and the corresponding data label of the user feature into the preset tree model to obtain the first importance value of each user feature;

[0082] Step S302: Repeat the process of shuffling the data labels and inputting them into the tree model multiple times to obtain the second importance value of each user feature;

[0083] Step S303: Determine the stability value of each user feature based on the first importance value and the second importance value obtained each time;

[0084] Step S304: Determine the fitness of each user feature based on the stability value of each user feature.

[0085] In step S301 of some embodiments, each user feature and its corresponding data label are input into a preset tree model to obtain a first importance value for each user feature. In this case, each user feature and its corresponding correct data label are input into the tree model, thus obtaining a first importance value for each user feature without shuffling the labels. It should be noted that if a feature is very important, its importance in the correct label will be high.

[0086] In step S302 of some embodiments, the data labels are shuffled before being input into the tree model. In this case, each user feature and the shuffled data labels are input into the tree model, thus obtaining a second importance value for each user feature under the shuffled label condition. It should be noted that if a feature is very important, its importance in the shuffled label model will be very low.

[0087] In step S303 of some embodiments, after repeatedly executing steps S301 and 302, the stability value of each user feature can be determined based on the first importance value and the second importance value obtained each time. This is because if a feature is very important, its importance in the correct label is high, while its importance in the shuffled label model is low. Comparing these two values ​​allows for the selection of stable features.

[0088] In step S304 of some embodiments, the fitness of each user feature is determined based on the stability value of each user feature. The magnitude of the stability value can be used to reflect the stability of each user feature, and the fitness of each user feature is determined based on the stability value, so that user features with fitness higher than a preset threshold can be selected as candidate features, thereby improving the success rate of finding key features, and finding the final feature combination through multiple evolutionary iterations for use in predicting user insurance payment behavior.

[0089] It should be noted that, in addition to using a tree model, the embodiments of this application can also be implemented using a simple neural network model.

[0090] Please see Figure 4 In some embodiments, step S202 may include, but is not limited to, steps S401 to S402:

[0091] Step S401: Compare the first importance value and the second importance value of each user feature multiple times to obtain the stability value of each user feature;

[0092] Step S402: Select user features with stability values ​​higher than a preset threshold as candidate features.

[0093] In step S401 of some embodiments, the first importance value and the second importance value of each user feature are compared multiple times. For example, the comparison is repeated n times, and the average value of the n feature importance values ​​is taken to obtain the stability value of each user feature.

[0094] In step S402 of some embodiments, user features with stability values ​​higher than a preset threshold are selected as candidate features. For example, the stability value of each user feature can be measured by comparing the differences between a first importance value and a second importance value multiple times and averaging the differences obtained from the multiple comparisons. An appropriate threshold can then be set to select user features with values ​​higher than the preset threshold as candidate features. Based on this, the present application can remove user features that become noise and automatically extract truly important user features as candidate features, thereby improving the success rate of identifying key features.

[0095] Please see Figure 5 In some embodiments, step S401 may include, but is not limited to, steps S501 to S502:

[0096] Step S501: Calculate the difference between the first importance value and the second importance value of the user feature obtained each time to obtain multiple importance difference values;

[0097] Step S502: Calculate the average of multiple importance differences to obtain the stability value of each user feature.

[0098] In step S501 of some embodiments, multiple importance difference values ​​are obtained by subtracting the first importance value and the second importance value obtained each time.

[0099] In step S502 of some embodiments, the average of multiple importance differences is calculated to measure the stability value of each user feature. That is, the fitness of each user feature can be determined based on the stability value, so that user features with fitness higher than a preset threshold can be selected as candidate features, thereby improving the success rate of identifying key features. Through multiple evolutionary iterations, the final feature combination is found for use in predicting user insurance payment behavior.

[0100] Please see Figure 6 In some embodiments, step S104 may include, but is not limited to, step S601:

[0101] Step S601: Iterate through the child features until the number of iterations reaches a preset threshold, then stop iterating to obtain the final feature combination used to predict user behavior.

[0102] In step S601 of some embodiments, after obtaining the offspring features, the next round of iterative training begins until a preset condition is met, at which point the iteration stops. For example, iteration stops when a preset number of iterations has been reached, and the final feature combination is output. This final feature combination is then used to predict user behavior. In the insurance premium prediction scenario, the user's possible premium payment time and payment probability can be predicted based on the final feature combination.

[0103] In some embodiments of this application, the evolutionary methods include at least one of heredity, mutation, natural selection, and hybridization. That is, the evolutionary methods include, but are not limited to, using heredity, mutation, natural selection, and hybridization for calculation. The corresponding evolutionary calculation methods here are genetic algorithms such as crossover, subtree mutation, point mutation, and Hoist mutation to obtain offspring characteristics. By achieving the evolution of the expression of individuals within each generation of the population through the above methods, the "survival of the fittest" forest law is realized, ultimately obtaining the optimal factor combination expression. This simultaneously solves the problems of factor selection and factor combination methods, automatically completing factor combination analysis and evaluating the importance of features. By combining features through genetic programming schemes, these features are automatically combined in insurance premium prediction, and the combined features are selected. This application can be applied to automatic feature evaluation, automatic model training, evaluation, and deployment, which can significantly shorten the overall model deployment time to adapt to increasingly diverse user behaviors.

[0104] The user behavior prediction method in this application embodiment can randomly generate multiple user features based on target user data. These user features may include, but are not limited to, features such as the insurer's city, income, insurance type, age, and occupation, which can be used to predict user insurance payment behavior. A null importance algorithm is used to filter candidate features from these multiple user features. The null importance algorithm removes features that are considered noise and extracts truly important features. It should be noted that the null importance of each feature is measured using training data, in which the target variable is randomly shuffled. If a feature is very important, its importance in the correct label is high, while its importance in the shuffled label model is low. Comparing these two factors allows for the selection of stable features. Based on this principle, the null importance algorithm can be used to filter truly important user features from multiple user features, and these truly important user features are used as candidate features, which can also be called parent features. The candidate features are then processed using an evolutionary method to obtain offspring features. The evolutionary methods include, but are not limited to, using genetics, mutation, natural selection, and hybridization for computation. The corresponding evolutionary computation methods here are crossover, subtree mutation, point mutation, and Hoist mutation to obtain offspring features. The offspring features are iteratively processed until a preset condition is met, at which point iteration stops, resulting in the final feature combination used to predict user behavior. After obtaining the offspring features, the next round of iterative training begins, stopping again when a preset condition is met, such as reaching a preset number of iterations. The final feature combination is then output and used to predict user behavior. In the insurance premium prediction scenario, the user's possible premium payment time and probability can be predicted based on the final feature combination.

[0105] Based on this, the user behavior prediction method provided in this application randomly generates multiple user features based on target user data; a null importance algorithm is used to filter candidate features from the multiple user features; the candidate features are processed according to an evolutionary method to obtain offspring features; the offspring features are iteratively processed until a preset condition is met, at which point the iteration stops, resulting in a final feature combination for predicting user behavior. Based on this, this application introduces the null importance algorithm to remove user features that become noise and automatically extracts truly important user features as candidate features, thereby improving the success rate of identifying key features. Through multiple evolutionary iterations, a final feature combination is found for predicting user insurance payment behavior. Therefore, compared to the prior art that relies on manual analysis and feature combination searching, this application can automatically complete the feature combination analysis for predicting user insurance payment behavior, greatly improving efficiency. This application can be applied to automatic feature evaluation, automatic model training, evaluation, and deployment, significantly shortening the overall model deployment time to adapt to increasingly complex user behavior prediction.

[0106] Please see Figure 7 This application also provides a user behavior prediction device that can implement the above-described user behavior prediction method. The device includes:

[0107] The generation module 710 is used to randomly generate multiple user features based on the target user data;

[0108] The filtering module 720 is used to filter candidate features from multiple user features using the null importance algorithm;

[0109] Evolution module 730 is used to process candidate features according to the evolution method to obtain offspring features;

[0110] The iteration module 740 is used to iteratively process the child features until the preset conditions are met, and then stop the iteration to obtain the final feature combination used to predict user behavior.

[0111] In some embodiments of this application, the generation module 710 randomly generates multiple user features based on the target user data, the filtering module 720 uses the null importance algorithm to filter out candidate features from the multiple user features, the evolution module 730 processes the candidate features according to the evolution method to obtain offspring features, and the iteration module 740 iteratively processes the offspring features until the preset conditions are met and then stops the iteration to obtain the final feature combination used to predict user behavior.

[0112] In some embodiments of this application, the generation module 710 can randomly generate multiple user features based on target user data. These user features may include, but are not limited to, features that can be used to predict user insurance payment behavior, such as the insurer's city, income, insurance type, age, and occupation.

[0113] In some embodiments of this application, the screening module 720 can use a null importance algorithm to select candidate features from multiple user features. The null importance algorithm can remove features that are considered noise and extract truly important features. It should be noted that the null importance of each feature is measured using training data, in which the target variable is randomly shuffled. If a feature is very important, its importance in the correct label is high, while its importance in the shuffled label model is low. Comparing these two aspects allows for the selection of stable features. Based on this principle, the null importance algorithm can be used to select truly important user features from multiple user features and use these truly important user features as candidate features, where candidate features can also be called parent features.

[0114] In some embodiments of this application, the evolution module 730 can process candidate features according to evolutionary methods to obtain offspring features. These evolutionary methods include, but are not limited to, calculations using heredity, mutation, natural selection, and crossover. The corresponding evolutionary calculation methods here are genetic algorithms such as crossover, subtree mutation, point mutation, and Hoist mutation to obtain offspring features. By achieving the evolution of the expression of individuals within each generation of the population through the above methods, the "survival of the fittest" forest law is realized, ultimately obtaining the optimal factor combination expression. This simultaneously solves the problems of factor selection and factor combination methods, automatically completing factor combination analysis and evaluating feature importance. In insurance premium prediction, features are automatically combined using genetic programming schemes, and the combined features are then selected. This application can be applied to automatic feature evaluation, automatic model training, evaluation, and deployment, significantly shortening the overall model deployment time to adapt to increasingly diverse user behaviors.

[0115] In some embodiments of this application, the iteration module 740 can iteratively process the child features until a preset condition is met, at which point the iteration stops, resulting in a final feature combination for predicting user behavior. After obtaining the child features, the next round of iterative training begins, continuing until a preset condition is met, such as stopping iteration after a preset number of iterations has been reached. The final feature combination is then output and used to predict user behavior. In an insurance premium prediction scenario, the user's possible premium payment time and probability can be predicted based on the final feature combination.

[0116] Based on this, the user behavior prediction device of this application embodiment includes a generation module 710 that randomly generates multiple user features based on target user data, a filtering module 720 that uses a null importance algorithm to filter candidate features from the multiple user features, an evolution module 730 that processes the candidate features according to an evolutionary method to obtain offspring features, and an iteration module 740 that iteratively processes the offspring features until a preset condition is met, at which point the iteration stops, resulting in the final feature combination used to predict user behavior. Therefore, compared to the prior art that relies on manual analysis and feature combination searching, this application can automatically complete the feature combination analysis for predicting user insurance payment behavior, greatly improving efficiency. This application can be applied to automatic feature evaluation and automatic model training, evaluation, and deployment, significantly shortening the overall model deployment time to adapt to increasingly diverse user behavior predictions.

[0117] The specific implementation of this user behavior prediction device is basically the same as the specific embodiment of the user behavior prediction method described above, and will not be repeated here.

[0118] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the aforementioned user behavior prediction method. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0119] Please see Figure 8 , Figure 8 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes:

[0120] The processor 801 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, 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 this application.

[0121] The memory 802 can be implemented as a read-only memory (ROM), static storage device, dynamic storage device, or random access memory (RAM). The memory 802 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 802 and called by the processor 801 to execute the user behavior prediction method of the embodiments of this application. Specifically, the method involves: randomly generating multiple user features based on target user data; using a null importance algorithm to select candidate features from the multiple user features; processing the candidate features according to an evolutionary method to obtain offspring features; iteratively processing the offspring features until a preset condition is met, at which point the iteration stops, resulting in the final feature combination used to predict user behavior. Based on this, this application introduces a null importance algorithm to remove user features that become noise and automatically extracts truly important user features as candidate features, thereby improving the success rate of identifying key features. Through multiple evolutionary iterations, the final feature combination is found for predicting user insurance payment behavior. Therefore, compared to existing technologies that rely on manual analysis and feature combination searching, this application can automatically complete feature combination analysis for predicting user insurance payment behavior, greatly improving efficiency. This application can be applied to automatic feature evaluation, automatic model training, evaluation, and deployment, significantly shortening the overall model deployment time to adapt to increasingly complex user behavior predictions.

[0122] The input / output interface 803 is used to implement information input and output.

[0123] The communication interface 804 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0124] The bus transmits information between various components of the device, such as processor 801, memory 802, input / output interface 803, and communication interface 804.

[0125] The processor 801, memory 802, input / output interface 803, and communication interface 804 communicate with each other within the device via a bus.

[0126] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described user behavior prediction method.

[0127] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0128] The user behavior prediction method, device, electronic device, and storage medium provided in this application embodiment randomly generate multiple user features based on target user data; a null importance algorithm is used to filter candidate features from the multiple user features; the candidate features are processed according to an evolutionary method to obtain offspring features; the offspring features are iteratively processed until a preset condition is met, at which point the iteration stops, resulting in a final feature combination for predicting user behavior. Based on this, this application introduces a null importance algorithm to remove user features that become noise and automatically extracts truly important user features as candidate features, thereby improving the success rate of identifying key features. Through multiple evolutionary iterations, a final feature combination is found for predicting user insurance payment behavior. Therefore, compared to the prior art that relies on manual analysis and feature combination searching, this application can automatically complete the feature combination analysis for predicting user insurance payment behavior, greatly improving efficiency. This application can be applied to automatic feature evaluation, automatic model training, evaluation, and deployment, significantly shortening the overall model deployment time to adapt to increasingly complex user behavior prediction.

[0129] Those skilled in the art will understand that all or some of the steps and systems in the methods disclosed above can be implemented as software, firmware, hardware, and suitable combinations thereof. Some or all of the physical components can be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable programs, data structures, program modules, or other data). Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, as is known to those skilled in the art, communication media typically contain computer-readable programs, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0130] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0131] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0132] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0133] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0134] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0135] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0136] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0137] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0138] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0139] If the integrated unit is implemented as 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 this application, in essence, 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. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0140] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A user behavior prediction method, characterized in that, The method includes: Multiple user characteristics are randomly generated based on the target user data; The fitness of each user feature is calculated using the null importance algorithm, including: repeatedly inputting each user feature and its corresponding data label into a preset tree model to obtain a first importance value for each user feature; repeatedly shuffling the data labels and inputting them into the tree model to obtain a second importance value for each user feature; determining a stability value for each user feature based on the first and second importance values ​​obtained each time; and determining the fitness of each user feature based on its stability value, wherein the fitness is used to characterize the importance and stability of the user feature. The user features with fitness values ​​higher than a preset threshold are selected as candidate features; The candidate features are processed using an evolutionary method to obtain offspring features; The child features are iteratively processed until a preset condition is met, at which point the iteration stops, resulting in the final feature combination used to predict user behavior.

2. The method according to claim 1, characterized in that, The step of selecting user features with fitness values ​​higher than a preset threshold as candidate features includes: By repeatedly comparing the first importance value and the second importance value of each user feature, a stability value for each user feature is obtained. User features whose stability value is higher than a preset threshold are selected as candidate features.

3. The method according to claim 2, characterized in that, The process of repeatedly comparing the first importance value and the second importance value of each user feature to obtain a stability value for each user feature includes: The difference between the first importance value and the second importance value of the user feature obtained each time is calculated to obtain multiple importance difference values; The stability value of each user feature is obtained by averaging the multiple importance differences.

4. The method according to claim 1, characterized in that, The iterative processing of the child features, stopping the iteration when a preset condition is met, to obtain the final feature combination used to predict user behavior includes: The child features are iteratively processed until the number of iterations reaches a preset threshold, at which point the iteration stops, resulting in the final feature combination used to predict user behavior.

5. The method according to any one of claims 1 to 4, characterized in that, The evolutionary methods include at least one of heredity, variation, natural selection, and hybridization.

6. A user behavior prediction device, characterized in that, The apparatus is used to implement the user behavior prediction method according to any one of claims 1 to 5, and the apparatus comprises: The generation module is used to randomly generate multiple user features based on the target user data; A filtering module is used to filter candidate features from multiple user features using a null importance algorithm; An evolution module is used to process the candidate features according to an evolution method to obtain offspring features; An iterative module is used to iteratively process the child features until a preset condition is met, at which point the iteration stops, resulting in the final feature combination used to predict user behavior.

7. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the user behavior prediction method according to any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the user behavior prediction method according to any one of claims 1 to 5.

Images