AI-based customer mining methods and related equipment

Abstract

This application relates to artificial intelligence and provides a customer mining method based on artificial intelligence. The method includes: filtering acquired user profile tag data to obtain churned user profile tag data; acquiring the churned user profile tag data after setting observation points as feature data to be filtered; extracting target data features from the feature data to be filtered; inputting the churned user profile tag data and the target data features into a customer win-back prediction model, outputting a churned user's win-back prediction score and corresponding lead tags; starting a lead batch processing task to perform batch data processing on the churned user profile tag data and the lead tags; obtaining a group stability index during the operation of the customer win-back prediction model and the lead batch processing task, and optimizing the target data features and the customer win-back prediction model based on the group stability index.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a customer mining method and related equipment based on artificial intelligence. Background Technology

[0002] In an increasingly competitive market, customers are a valuable resource for every company. Although existing technologies can maintain existing customer resources through early warning and mid-process intervention, a certain percentage of customers will still be lost. Therefore, every company needs to take certain measures to win back the customers that have already been lost.

[0003] The current approach involves companies investing a certain number of customer service personnel who rely on their personal experience to take certain measures to win back customers. This not only leads to low efficiency in winning back customers and makes it difficult to promote on a large scale, but also results in high operating costs for the company. Summary of the Invention

[0004] This application provides a customer mining method, apparatus, computer equipment, and storage medium based on artificial intelligence to solve the problems of low efficiency and high operating costs in winning back lost customers in the prior art.

[0005] A first aspect of this application provides an artificial intelligence-based customer mining method, comprising:

[0006] User profile tag data is obtained from user data based on the target information dimension, and the user profile tag data that meets the user churn judgment rules is selected as the churn user profile tag data of churned users;

[0007] Within a set first time period, an observation point is set, and the profile tag data of the churned user within a first time period after the observation point is obtained as the feature data to be filtered. The user churn time of the churned user at the observation point is within a second time period.

[0008] Extract the data features to be evaluated from the feature data to be screened, and use at least one feature evaluation dimension to screen the data features to be evaluated to obtain the target data features;

[0009] The churned user profile tag data and the target data features are input into the customer win-back prediction model, and the churned user profile tag corresponding to the churned user's win-back prediction score and the corresponding lead tag are output. The churned users are divided into predicted win-back users and predicted unwin-back users according to the win-back score threshold.

[0010] Start the lead batch processing task to perform batch data processing on the churned user profile tag data and the lead tags corresponding to the predicted cul-able users and the predicted uncul-able customers.

[0011] Obtain a first group stability index of the target data features during the operation of the customer win prediction model, obtain a second group stability index of the target data features during the operation of the lead batch task, remove the target data features that are not within the corresponding preset first stability index threshold range and preset second stability index threshold range according to the first group stability index and the second group stability index, and / or retrain the customer win prediction model.

[0012] A second aspect of this application provides an artificial intelligence-based customer mining device, comprising:

[0013] The churned user profiling module is used to obtain user profile tag data from user data based on target information dimensions, and filter the user profile tag data that meets the user churn judgment rules as churned user profile tag data.

[0014] The feature data to be filtered module is used to set observation points within a set first time period and obtain the profile label data of the churned users within a first time period after the observation point as the feature data to be filtered, wherein the user churn time of the churned user at the observation point is less than a second time threshold.

[0015] The target data feature module is used to extract the data features to be evaluated from the feature data to be screened, and to screen the data features to be evaluated using at least one feature evaluation dimension to obtain the target data features.

[0016] The customer win-back prediction model module is used to input the churned user profile tag data and the target data features into the customer win-back prediction model, output the churned user's win-back prediction score and corresponding lead tag corresponding to the churned user profile tag, and classify the churned users into predicted win-back users and predicted unwin-back users according to the win-back score threshold.

[0017] The lead batch processing module is used to start the lead batch processing task and perform batch data processing on the churned user profile tag data and the lead tags corresponding to the predicted cul-potable users and the predicted non-cul-potable customers.

[0018] The group stability monitoring module is used to obtain a first group stability index of the target data features during the operation of the customer win prediction model, obtain a second group stability index of the target data features during the operation of the lead batch task, and remove unstable target data features or retrain the customer win prediction model based on the first group stability index and the second group stability index.

[0019] A third aspect of this application provides a computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described artificial intelligence-based customer mining method.

[0020] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described artificial intelligence-based customer mining method.

[0021] The aforementioned AI-based customer acquisition method, apparatus, computer equipment, and storage medium acquire churned user profile data, construct suitable data samples using this data, and process these data samples through an AI model to obtain the probability of winning back each churned user and related leads. Furthermore, it monitors changes in each churned user and their leads for optimization. This not only improves the efficiency of churned user data processing and the execution efficiency of the customer recovery process but also increases the probability of customer recovery and reduces enterprise operating costs. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of an application environment for a customer mining method based on artificial intelligence, as described in one embodiment of this application.

[0024] Figure 2 This is a flowchart of a customer mining method based on artificial intelligence in one embodiment of this application;

[0025] Figure 3 This is a schematic diagram of the structure of an artificial intelligence-based customer mining device in one embodiment of this application;

[0026] Figure 4 This is a schematic diagram of a computer device according to one embodiment of this application. Detailed Implementation

[0027] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0028] This application's embodiments can acquire and process relevant data based on artificial intelligence (AI) technology. AI is the theory, methods, technology, and application system that uses digital computers or computers-controlled machines to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results. Basic AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operating / interactive systems, and mechatronics. AI software technologies mainly include computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0029] The AI-based customer mining method provided in this application can be applied to, for example... Figure 1 In this application environment, the computer equipment can be, but is not limited to, various personal computers and laptops. The computer equipment can also be a server, which can be a standalone server 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, content delivery networks (CDNs), and big data and artificial intelligence platforms. This is understandable. Figure 1 The number of computer devices shown is merely illustrative and can be expanded in any number according to actual needs.

[0030] In one embodiment, such as Figure 2 As shown, an artificial intelligence-based customer mining method is provided, which can be applied to... Figure 1 The following steps, S101 to S106, are used as an example of computer equipment in the example:

[0031] S101. Obtain user profile tag data from user data according to the target information dimension, and filter the user profile tag data that meets the user churn judgment rules as churned user profile tag data.

[0032] The target information dimension is an information dimension that is related to the business attributes of the target business. For example, a fintech platform might obtain user profile tag data from target information dimensions such as policyholder dimension information, insured dimension information, policy dimension, user online behavior dimension, and operational service dimension of all users who have purchased insurance products on the platform. This data is used to conduct business to win back churned customers related to insurance products. The user churn judgment rule is derived based on the business characteristics of the target business, the user profile tag data, and historical user churn data. For example, a fintech platform might classify users who have logged out of the platform app or have not opened the platform app for more than 90 days as churned users. Furthermore, considering the business characteristics of the insurance industry within fintech, users whose insurance products have expired and have not been renewed are also classified as churned users.

[0033] S102. Set observation points within a set first time period, and obtain the churned user profile label data of the churned user within a first time period after the observation point as the feature data to be filtered, wherein the user churn time of the churned user at the observation point is less than a second time threshold.

[0034] The first time period is set according to the business requirements of the target business. If the status of the customer corresponding to the target business does not change significantly within a short period, a relatively long first time period can be set, such as setting the first time period to be one month, that is, setting observation points within each month. Conversely, if the status of the customer corresponding to the target business changes significantly within a short period, a relatively short first time period can be set, such as setting the first time period to be one week, that is, setting observation points within each week.

[0035] The first time period is set according to the business requirements of the target business. If the status of customers corresponding to the target business does not change significantly within a short period of time, a relatively long first time period can be set, such as one month. That is, the profile tag data of churned users within one month after the observation point is obtained as the feature data to be filtered. Conversely, if the status of customers corresponding to the target business changes significantly within a short period of time, a relatively short first time period can be set, such as one week. That is, the profile tag data of churned users within one week after the observation point is obtained as the feature data to be filtered.

[0036] The second time threshold is set according to the business needs of the target business. If customers corresponding to the target business still have a high probability of being won back after a long period of churn, a relatively long second time threshold can be set, such as one year, meaning the churned user's churn time at the observation point is less than one year. Conversely, if customers corresponding to the target business still have a low probability of being won back after a long period of churn, a relatively short second time threshold can be set, such as one quarter, meaning the churned user's churn time at the observation point is less than one quarter.

[0037] S103. Extract the data features to be evaluated from the feature data to be screened, and use at least one feature evaluation dimension to screen the data features to be evaluated to obtain the target data features.

[0038] The feature evaluation dimensions include, but are not limited to: Information Value (IV), Information Saturation, Population Stability Index (PSI), Volatility, and Correlation.

[0039] Further, the step of using at least one feature evaluation dimension to filter the data features to be evaluated to obtain target data features includes: Firstly, calculating the information value of each data feature to be evaluated based on the data to be filtered, and removing data features whose information value is not within a preset information value threshold range. Secondly, calculating the information saturation of each data feature to be evaluated based on the data to be filtered, and removing data features whose information saturation is not within a preset information saturation threshold range. Thirdly, calculating the group stability index of each data feature to be evaluated based on the data to be filtered, and removing data features whose group stability index is not within a preset third stability index threshold range. Fourthly, inputting the data to be filtered into an adversarial verification model, outputting the time-series volatility of each data feature to be evaluated, and removing data features whose time-series volatility is not within a preset time-series volatility threshold range. Fifthly, inputting the data to be filtered into a business relevance calculation model, outputting the relevance score between each data feature to be evaluated and the target business, and removing data features whose relevance score is not within a preset relevance score threshold range.

[0040] S104. Input the churned user profile tag data and the target data features into the customer win-back prediction model, output the churned user win-back prediction score and the corresponding lead tag corresponding to the churned user profile tag, and divide the churned users into predicted win-back users and predicted unwin-back users according to the win-back score threshold.

[0041] The applicant first downsampled using K-means clustering, and then trained and evaluated multiple artificial intelligence models simultaneously. The artificial intelligence models used included, but were not limited to, logistic regression models, SVM ensemble tree models, and neural network models. Finally, the ensemble tree model was used after evaluation by model evaluation metrics such as AUC (Area Under Curve), precision, recall, and PR (Precision Recall) curve.

[0042] Furthermore, the customer winback prediction model is constructed based on an ensemble tree model. After outputting the winback prediction score and corresponding lead tags for churned users corresponding to the churned user profile tags, and classifying the churned users into predictably winbackable users and predictably unwinnable users based on the winback score threshold, the model further includes: first, inputting the lead tags and user profile tag data into a marketing script matching model, and outputting the user marketing script strategy corresponding to the user profile tag data. Then, sending the user marketing script strategy to the corresponding target business marketing personnel, and monitoring whether the users corresponding to the user profile tag data are won back within a third time period after the user marketing script strategy is delivered.

[0043] S105. Start the lead batch processing task to perform batch data processing on the churned user profile tag data and the lead tags corresponding to the predicted culprit users and the predicted non-cultivable customers.

[0044] In this context, "batch processing" typically refers to an application performing specific processing on a batch of data. For example, in the financial operations of a fintech platform, batch processing scenarios include tasks such as daily account settlement, expense accrual, batch deduction of outstanding payments, and non-performing asset handling. The leads generated from batch processing are business-related to these tasks. Because the churned user profile data tags and lead tags corresponding to the predicted churnable and unrecoverable users are sometimes covered by existing batch processing tasks, changes in this covered data after the execution of these tasks can affect the likelihood of churned users being won back. For instance, if a fintech platform configures a batch processing task during a large-scale event to issue coupon rewards to churned users, then those users who receive the coupon rewards have a chance of being won back.

[0045] S106. Obtain the first group stability index of the target data features during the operation of the customer win-back prediction model, obtain the second group stability index of the target data features during the operation of the lead batch task, remove the target data features that are not in the corresponding preset first stability index threshold range and preset second stability index threshold range according to the first group stability index and the second group stability index, and / or retrain the customer win-back prediction model.

[0046] Furthermore, after obtaining the second group stability index of the target data features during the execution of the lead batch task, the method further includes: first, calculating the third group stability index of the target data features within the second time period according to a preset second time period. Then, if the third group stability index is not within the range of a preset fourth stability index threshold, the target data feature is removed, and / or the customer win-back prediction model is retrained.

[0047] In practical applications, the first and second group stability indices are typically calculated on a daily basis, meaning they are updated daily. However, it's undeniable that changes in data have a lag effect on users. Therefore, a longer statistical period should be set to evaluate the impact of the target data characteristics after the lead batch processing task. This means setting longer second time periods, such as weeks, months, quarters, or years, to statistically analyze the group stability index of the target data characteristics within those second time periods.

[0048] Further, after removing the target data features that are outside the corresponding preset first stability index threshold range and preset second stability index threshold range based on the first group stability index and the second group stability index, and / or retraining the customer retention prediction model, the method further includes: First, monitoring the churn user conversion data associated with the lead tags at different time periods. Then, optimizing the customer retention prediction model based on the churn user conversion data. Simultaneously, obtaining the churn user conversion rate in the churn user conversion data, and removing the corresponding lead tags whose churn user conversion rate is outside the preset conversion rate threshold range. Finally, obtaining the volatility of the churn user conversion rate, adding a fourth time period setting where the volatility is outside the preset volatility change range, and / or deleting a fifth time period setting where the volatility is within the preset volatility change range.

[0049] Furthermore, after setting different time periods to monitor the churned user conversion data associated with the clue tags, the process further includes: First, obtaining the user profile tag data corresponding to the churned user conversion data as candidate tag data. Then, extracting the user underlined tag data from the candidate tag data according to the target business statistical requirements as target classification tags. Simultaneously, clustering the churned user conversion data according to the target classification tags, and obtaining the user conversion results corresponding to different values ​​of the target classification tags for the churned user conversion data in each cluster. Finally, if the user conversion result triggers a preset conversion result prompt rule, the user conversion result, the corresponding target classification tag, and the value of the corresponding target classification tag are sent to relevant operations personnel, wherein the preset conversion result includes user conversion rates greater than a preset maximum user conversion rate threshold and user conversion rates less than a preset minimum user conversion rate threshold. For example, if a fintech platform finds that the conversion rate of churned users whose channel source tag is a certain channel is less than the minimum conversion rate threshold, it can stop conversion intervention measures for churned users with that channel source tag. Conversely, if the fintech platform finds that the conversion rate of churned users whose region tag is a certain region is greater than or equal to the maximum conversion rate threshold, it can increase conversion intervention measures for churned users with that region tag.

[0050] The AI-based customer acquisition method provided in this application acquires user profile data of churned users, constructs suitable data samples using this data, and processes these data samples through an AI model to obtain the probability of winning back each churned user and related leads. Based on these leads, corresponding sales pitches are generated for each churned user, and changes in each churned user and their leads are monitored for relevant optimization. This not only improves the efficiency of churned user data processing and the execution efficiency of the customer recovery process, but also increases the probability of winning back customers and reduces enterprise operating costs.

[0051] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0052] In one embodiment, an AI-based customer mining device 100 is provided, which corresponds one-to-one with the AI-based customer mining method described in the above embodiments. For example... Figure 3 As shown, the AI-based customer mining device 100 includes a churned user profiling module 11, a target feature data module 12, a target data feature module 13, a customer winback prediction model module 14, a lead batch processing task module 15, and a group stability monitoring module 16. Detailed descriptions of each functional module are as follows:

[0053] The churned user profile module 11 is used to obtain user profile tag data from user data according to the target information dimension, and filter the user profile tag data that meets the user churn judgment rules as churned user profile tag data.

[0054] The feature data module 12 is used to set observation points within a set first time period and obtain the churned user profile label data of the churned user within a first time period after the observation point as the feature data to be filtered, wherein the user churn time of the churned user at the observation point is less than a second time threshold.

[0055] Target data feature module 13 is used to extract data features to be evaluated from the feature data to be screened, and to screen the data features to be evaluated using at least one feature evaluation dimension to obtain target data features;

[0056] The customer win-back prediction model module 14 is used to input the churned user profile tag data and the target data features into the customer win-back prediction model, output the churned user's win-back prediction score and corresponding lead tag corresponding to the churned user profile tag, and classify the churned users into predicted win-back users and predicted unwin-back users according to the win-back score threshold.

[0057] Lead batch processing task module 15 is used to start the lead batch processing task and perform batch data processing on the churned user profile tag data and the lead tags corresponding to the predicted cultivable users and the predicted uncultivable customers.

[0058] The group stability monitoring module 16 is used to obtain a first group stability index of the target data features during the operation of the customer win prediction model, obtain a second group stability index of the target data features during the operation of the lead batch task, and remove unstable target data features or retrain the customer win prediction model based on the first group stability index and the second group stability index.

[0059] Furthermore, the target data feature module 13 also includes:

[0060] The information value quantum module is used to calculate the information value of each of the data features to be evaluated based on the feature data to be screened, and to remove the corresponding data features to be evaluated whose information value is not within the preset information value threshold range.

[0061] The information saturation submodule is used to calculate the information saturation of each of the data features to be evaluated based on the feature data to be screened, and remove the corresponding data features to be evaluated whose information saturation is not within the preset information saturation threshold range;

[0062] The population stability index submodule is used to calculate the population stability index of each of the data features to be evaluated based on the feature data to be screened, and to remove the corresponding data features to be evaluated whose population stability index is not within the range of the preset third stability index threshold.

[0063] The time series volatility submodule is used to input the feature data to be screened into the adversarial verification model, output the time series volatility of each data feature to be evaluated, and remove the corresponding data features to be evaluated whose time series volatility is not within the preset time series volatility threshold range.

[0064] The correlation scoring submodule is used to input the feature data to be screened into the business correlation calculation model, output the correlation score between each data feature to be evaluated and the target business, and remove the corresponding data features to be evaluated whose correlation scores are not within the preset correlation score threshold range.

[0065] Furthermore, the customer win-back prediction model module 14 also includes:

[0066] The marketing script strategy submodule is used to input the lead tags and user profile tags into the marketing script matching model and output the user marketing script strategy corresponding to the user profile tags.

[0067] The marketing script monitoring submodule is used to send the user marketing script strategy to the corresponding target business marketing personnel and monitor whether the user corresponding to the user profile tag data is won back within the third time period after the user marketing script strategy is delivered.

[0068] Furthermore, the group stability monitoring module 16 also includes:

[0069] The third population stability submodule is used to statistically analyze the third population stability index of the target data features within the second time period according to a preset second time period.

[0070] The third group stability optimization submodule is used to remove the target data feature and / or retrain the customer win prediction model if the third group stability index is not within the range of the preset fourth stability index threshold.

[0071] Furthermore, the group stability monitoring module 16 also includes:

[0072] The churned user conversion data submodule is used to set different time periods to monitor the churned user conversion data associated with the lead tags;

[0073] The first prediction model optimization submodule is used to optimize the customer win-back prediction model based on the churned user conversion data.

[0074] The first clue tag optimization submodule is used to obtain the churn user conversion rate in the churn user conversion data and remove the corresponding clue tags whose churn user conversion rate is not within the preset conversion rate threshold range.

[0075] The time period setting optimization submodule is used to obtain the volatility of the conversion rate of the churned users, add a fourth time period setting where the volatility is not within the preset volatility change range, and / or delete a fifth time period setting where the volatility is within the preset volatility change range.

[0076] Furthermore, the churned user conversion data submodule also includes:

[0077] The candidate tag data subunit is used to obtain the user profile tag data corresponding to the churned user conversion data as candidate tag data;

[0078] The target classification label subunit is used to extract the user underlined label data from the candidate label data as the target classification label according to the target business statistical requirements.

[0079] The conversion data clustering subunit is used to cluster the churned user conversion data according to the target classification label, and obtain the user conversion result corresponding to the churned user conversion data in each cluster when the target classification label takes different values;

[0080] The conversion result monitoring subunit is used to send the user conversion result, the corresponding target category tag, and the value of the corresponding target category tag to relevant operations personnel if the user conversion result triggers a preset conversion result prompt rule. The preset conversion result includes user conversion rate greater than a preset maximum user conversion rate threshold and user conversion rate less than a preset minimum user conversion rate threshold.

[0081] The terms "first" and "second" in the above-mentioned modules / units are only used to distinguish different modules / units and are not intended to specify which module / unit has a higher priority or any other limiting meaning. 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 device that includes a series of steps or modules is not necessarily limited to those steps or modules explicitly listed, but may include other steps or modules not explicitly listed or inherent to these processes, methods, products, or devices. The module divisions appearing in this application are merely logical divisions; in actual applications, different division methods may be used.

[0082] Specific limitations regarding the AI-based customer mining device can be found in the limitations of the AI-based customer mining method described above, and will not be repeated here. Each module in the aforementioned AI-based customer mining device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0083] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores data involved in an artificial intelligence-based customer mining method. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements an artificial intelligence-based customer mining method.

[0084] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the artificial intelligence-based customer mining method described in the above embodiments, for example... Figure 2 The steps S101 to S106 shown, as well as other extensions and related steps of the method, are examples. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit of the AI-based customer mining device in the above embodiments, for example... Figure 3 The functions of modules 11 to 16 are shown. To avoid repetition, they will not be described again here.

[0085] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the computer device, connecting various parts of the computer device via various interfaces and lines.

[0086] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the computer device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, applications required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created based on the use of the mobile phone (such as audio data, video data, etc.).

[0087] The memory can be integrated into the processor or it can be set up separately from the processor.

[0088] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When executed by a processor, the computer program implements the steps of the artificial intelligence-based customer mining method described in the above embodiments, for example... Figure 2 The steps S101 to S106 shown, as well as other extensions and related steps of the method, are examples. Alternatively, when the computer program is executed by a processor, it implements the functions of each module / unit of the AI-based customer mining device in the above embodiments, for example... Figure 3 The functions of modules 11 to 16 are shown. To avoid repetition, they will not be described again here.

[0089] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0090] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0091] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A customer mining method based on artificial intelligence, characterized in that, The method is applied to insurance customer acquisition on fintech platforms, including: User profile tag data is obtained from user data based on the target information dimension, and the user profile tag data that meets the user churn judgment rules is selected as the churn user profile tag data of churned users; Within a set first time period, an observation point is set, and the profile tag data of the churned user within a first time period after the observation point is obtained as the feature data to be filtered, wherein the user churn time of the churned user at the observation point is less than a second time threshold. Extract the data features to be evaluated from the feature data to be screened, and use at least one feature evaluation dimension to screen the data features to be evaluated to obtain the target data features; The churned user profile tag data and the target data features are input into the customer win-back prediction model, and the churned user profile tag corresponding to the churned user's win-back prediction score and the corresponding lead tag are output. The churned users are divided into predicted win-back users and predicted unwin-back users according to the win-back score threshold. Start the lead batch processing task to perform batch data processing on the churned user profile tag data and the lead tags corresponding to the predicted cul-able users and the predicted uncul-able customers. Obtain a first group stability index of the target data features during the operation of the customer win prediction model, obtain a second group stability index of the target data features during the operation of the lead batch task, remove the target data features that are not within the corresponding preset first stability index threshold range and preset second stability index threshold range according to the first group stability index and the second group stability index, and / or retrain the customer win prediction model.

2. The customer mining method based on artificial intelligence according to claim 1, characterized in that, The step of using at least one feature evaluation dimension to filter the features of the data to be evaluated to obtain the target data features includes: Calculate the information value of each data feature to be evaluated based on the data to be screened, and remove the corresponding data features to be evaluated whose information value is not within the preset information value threshold range; Calculate the information saturation of each of the data features to be evaluated based on the data to be screened, and remove the corresponding data features to be evaluated whose information saturation is not within the preset information saturation threshold range; Calculate the population stability index of each data feature to be evaluated based on the data to be screened, and remove the corresponding data features to be evaluated whose population stability index is not within the preset third stability index threshold range; The data to be screened is input into the adversarial verification model, and the time-series volatility of each data feature to be evaluated is output. Data features to be evaluated whose time-series volatility is not within the preset time-series volatility threshold range are removed. The data to be screened is input into the business relevance calculation model, and the relevance score between each data feature to be evaluated and the target business is output. Data features to be evaluated that do not have a relevance score within the preset relevance score threshold range are removed.

3. The customer mining method based on artificial intelligence according to claim 1, characterized in that, The customer winback prediction model is constructed based on an ensemble tree model. After outputting the winback prediction score and corresponding lead tags for the churned user profile tags, and classifying the churned users into predictably winable and unwinnable users based on the winback score threshold, the model further includes: Input the lead tags and user profile tags into the marketing script matching model, and output the user marketing script strategy corresponding to the user profile tag data; Send the user marketing script strategy to the corresponding target business marketing personnel, and monitor whether the user corresponding to the user profile tag data is won back within the third time period after the user marketing script strategy is delivered.

4. The customer mining method based on artificial intelligence according to claim 1, characterized in that, After obtaining the second population stability index of the target data features during the execution of the batch task for obtaining the clues, the method further includes: The third group stability index of the target data characteristics is calculated according to a preset second time period. If the stability index of the third group is not within the preset threshold range of the fourth stability index, then the target data feature is removed, and / or the customer win prediction model is retrained.

5. The customer mining method based on artificial intelligence according to claim 1, characterized in that, After removing the target data features that are outside the corresponding preset first stability index threshold range and preset second stability index threshold range based on the first group stability index and the second group stability index, and / or retraining the customer win prediction model, the method further includes: Set up different time periods to monitor the conversion data of churned users associated with the aforementioned lead tags; Optimize the customer win-back prediction model based on the churned user conversion data; Obtain the churn user conversion rate from the churn user conversion data, and remove the corresponding lead tags whose churn user conversion rates are not within the preset conversion rate threshold range; Obtain the volatility of the churned user conversion rate, add a fourth time period setting where the volatility is outside the preset volatility change range, and / or delete a fifth time period setting where the volatility is within the preset volatility change range.

6. The customer mining method based on artificial intelligence according to claim 5, characterized in that, After setting different time periods to monitor the conversion data of churned users associated with the lead tags, the method also includes: Obtain the user profile tag data corresponding to the churned user conversion data as candidate tag data; The user underlined tag data is extracted from the candidate tag data as the target classification tag according to the target business statistical requirements; The churned user conversion data is clustered according to the target classification label, and the user conversion results corresponding to the churned user conversion data in each cluster when the target classification label takes different values ​​are obtained. If the user conversion result triggers a preset conversion result prompt rule, the user conversion result, the corresponding target category tag, and the value of the corresponding target category tag will be sent to the relevant operations personnel. The preset conversion result includes user conversion rate greater than a preset maximum user conversion rate threshold and user conversion rate less than a preset minimum user conversion rate threshold.

7. A customer mining device based on artificial intelligence, characterized in that, include: The churned user profiling module is used to obtain user profile tag data from user data based on target information dimensions, and filter the user profile tag data that meets the user churn judgment rules as churned user profile tag data. The feature data to be filtered module is used to set observation points within a set first time period and obtain the profile label data of the churned users within a first time period after the observation point as the feature data to be filtered, wherein the user churn time of the churned user at the observation point is less than a second time threshold. The target data feature module is used to extract the data features to be evaluated from the feature data to be screened, and to screen the data features to be evaluated using at least one feature evaluation dimension to obtain the target data features. The customer win-back prediction model module is used to input the churned user profile tag data and the target data features into the customer win-back prediction model, output the churned user's win-back prediction score and corresponding lead tag corresponding to the churned user profile tag, and classify the churned users into predicted win-back users and predicted unwin-back users according to the win-back score threshold. The lead batch processing module is used to start the lead batch processing task and perform batch data processing on the churned user profile tag data and the lead tags corresponding to the predicted cul-potable users and the predicted non-cul-potable customers. The group stability monitoring module is used to obtain a first group stability index of the target data features during the operation of the customer win prediction model, obtain a second group stability index of the target data features during the operation of the lead batch task, and remove unstable target data features or retrain the customer win prediction model based on the first group stability index and the second group stability index.

8. The customer mining device based on artificial intelligence according to claim 7, characterized in that, The target data feature module also includes: The information value quantum module is used to calculate the information value of each of the data features to be evaluated based on the feature data to be screened, and to remove the corresponding data features to be evaluated whose information value is not within the preset information value threshold range. The information saturation submodule is used to calculate the information saturation of each of the data features to be evaluated based on the feature data to be screened, and remove the corresponding data features to be evaluated whose information saturation is not within the preset information saturation threshold range; The population stability index submodule is used to calculate the population stability index of each of the data features to be evaluated based on the feature data to be screened, and to remove the corresponding data features to be evaluated whose population stability index is not within the range of the preset third stability index threshold. The time series volatility submodule is used to input the feature data to be screened into the adversarial verification model, output the time series volatility of each data feature to be evaluated, and remove the corresponding data features to be evaluated whose time series volatility is not within the preset time series volatility threshold range. The correlation scoring submodule is used to input the feature data to be screened into the business correlation calculation model, output the correlation score between each data feature to be evaluated and the target business, and remove the corresponding data features to be evaluated whose correlation scores are not within the preset correlation score threshold range.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the artificial intelligence-based customer mining method as described in any one of claims 1 to 6.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the artificial intelligence-based customer mining method as described in any one of claims 1 to 6.

Images