Service strategy determination method and device and electronic equipment

Abstract

The invention discloses a service strategy determination method and device and electronic equipment, and relates to the artificial intelligence technology, and the method comprises the steps: obtaining the credit information of a target user, the credit information at least comprises personal basic information, credit transaction information, financial institution interaction information and credit score information; respectively inputting the personal basic information, the credit transaction information, the financial institution interaction information and the credit scoring information into at least two evaluation models, and obtaining evaluation results respectively output by each evaluation model; and determining a target service strategy of the target user according to the at least two evaluation results. According to the method and the device, the loss condition of the user can be determined through the at least two evaluation results, so that the user retention rate can be improved by implementing the corresponding service strategy, and the beneficial effects of providing personalized services for different users, improving the user experience and reducing the user loss rate are achieved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular, to a method, device, and electronic device for determining a service strategy. Background Art

[0002] In the current financial market environment, a type of financial service product (such as a credit card) plays a crucial role in the overall development plan of financial institutions. Compared with the traditional credit model, this type of financial service product exhibits unique advantages. Its high flexibility and convenience greatly facilitate the fund usage scenarios of customers. Whether it is the daily consumption expenditure of individual consumers or the capital turnover and allocation in the enterprise operation process, it can quickly and effectively meet the needs. At the same time, it also opens up an important intermediate income channel for financial institutions and promotes the expansion of business scale.

[0003] However, it cannot be ignored that during the development of this type of financial service product, the problem of increasingly prominent customer loss is faced. With the continuous expansion of the number of users, due to the interactive influence of many factors, such as intensified market competition, diversified and changeable customer needs, uneven service quality, etc., the risk of user loss is increasing day by day. And the traditional user management strategy has serious defects. It is only limited to using marketing means to try to recall users hastily after customer loss has become an established fact. This passive coping method is difficult to effectively solve the core problems of poor customer experience and high churn rate. Summary of the Invention

[0004] This application provides a method, device, and electronic device for determining a service strategy, which provides services for users by formulating corresponding service strategies, improves the user experience, and reduces the user churn rate.

[0005] In a first aspect, this application provides a method for determining a service strategy, including:

[0006] Obtain the credit information of the target user, where the credit information at least includes personal basic information, credit transaction information, financial institution interaction information, and credit score information;

[0007] Input the personal basic information, the credit transaction information, the financial institution interaction information, and the credit score information into at least two evaluation models respectively, and obtain the evaluation results respectively output by each evaluation model; at least two of the evaluation models are obtained by constructing at least two sample data subsets from the same credit data set and training the at least two sample data subsets;

[0008] Determine the target service strategy for the target user according to at least two of the evaluation results.

[0009] Second aspect, the present application provides a service policy determination device, the device includes:

[0010] An information acquisition module, configured to acquire credit information of a target user, where the credit information at least includes personal basic information, credit transaction information, financial institution interaction information, and credit score information;

[0011] A result output module, configured to input the personal basic information, the credit transaction information, the financial institution interaction information, and the credit score information into at least two evaluation models respectively, and obtain evaluation results respectively output by each of the at least two evaluation models; the at least two evaluation models are obtained by constructing at least two sample data subsets from the same credit data set and training the at least two sample data subsets;

[0012] A policy determination module, configured to determine a target service policy for the target user according to the at least two evaluation results.

[0013] Third aspect, the present application further provides an electronic device, the electronic device includes:

[0014] At least one processor; and

[0015] A memory communicatively connected to the at least one processor; wherein,

[0016] The memory stores a computer program executable by the at least one processor, and when the computer program is executed by the at least one processor, the at least one processor is enabled to execute the service policy determination method according to any embodiment of the present application.

[0017] Fourth aspect, the present application further provides a computer-readable storage medium, the computer-readable storage medium stores computer instructions, and the computer instructions are used to implement the service policy determination method according to any embodiment of the present application when executed by a processor.

[0018] Fifth aspect, the present application further provides a computer program product, including a computer program, and the computer program implements the service policy determination method according to any embodiment of the present application when executed by a processor.

[0019] The service strategy determination solution provided by the embodiments of this application first obtains the credit information of the target user. The credit information in this embodiment at least includes personal basic information, credit transaction information, financial institution interaction information, and credit scoring information. By obtaining information of different dimensions of the target user, a complete data foundation is provided to improve the accuracy of the subsequent determination of the evaluation result output by the model. Then, by means of analyzing the input credit analysis through at least two evaluation models respectively, different perspective features of the credit information can be captured through different evaluation models, which helps to formulate a precise service strategy for the target user and avoid the limitations and biases brought by data analysis using a single model. Finally, the target service strategy for the target user is determined according to at least two evaluation results. The solution provided by this embodiment can determine the user churn situation through at least two evaluation results, and then improve the user retention rate by implementing corresponding service strategies, achieving the beneficial effects of providing personalized services for different users, enhancing the user experience, and reducing the user churn rate.

[0020] It should be noted that the above computer instructions can be stored in whole or in part on a computer-readable storage medium. Among them, the computer-readable storage medium can be packaged together with the processor of the service strategy determination device or separately packaged with the processor of the service strategy determination device. This application does not make any limitations in this regard.

[0021] The descriptions of the second aspect, the third aspect, the fourth aspect, and the fifth aspect in this application can refer to the detailed description of the first aspect; and the beneficial effects of the descriptions of the second aspect, the third aspect, the fourth aspect, and the fifth aspect can refer to the analysis of the beneficial effects of the first aspect, which will not be elaborated here.

[0022] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of this application, nor is it used to limit the scope of this application. Other features of this application will become easily understood through the following description.

[0023] It can be understood that before using the technical solutions disclosed in the embodiments of this application, the types, usage scopes, and usage scenarios of the personal information involved in this application should be informed to the user and the user's authorization should be obtained in an appropriate manner in accordance with relevant laws and regulations. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of this application, the drawings required to be used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other relevant drawings can be obtained based on these drawings without creative efforts.

[0025] Figure 1 It is a schematic flowchart of a service policy determination method provided by an embodiment of the present application;

[0026] Figure 2 It is another schematic flowchart of a service policy determination method provided by an embodiment of the present application;

[0027] Figure 3 It is a schematic structural diagram of a service policy determination device provided by an embodiment of the present application;

[0028] Figure 4 It is a schematic structural diagram of an electronic device provided by an embodiment of the present application. Detailed implementation manners

[0029] In order to enable those skilled in the art to better understand the solutions of the present application, the technical solutions in the present application will be clearly and completely described below in conjunction with the accompanying drawings in this embodiment. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0030] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present application described here can be implemented in an order other than those illustrated or described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those clearly listed steps or units, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0031] The present application will be further described in detail below in conjunction with the drawings and embodiments. It can be understood that the specific embodiments described here are only used to explain the present application, rather than limiting the present application. Additionally, it should be noted that for the sake of convenience of description, only parts related to the present application rather than all the structures are shown in the drawings.

[0032] Figure 1This is a flowchart of the service strategy determination method provided by the embodiments of the present application. This embodiment is applicable to the situation of evaluating the churn probability of users in financial institutions, so as to provide personalized services for users to reduce the user churn rate. This method can be executed by a service strategy determination device, which can be implemented in the form of hardware and / or software and integrated in an electronic device that executes this method. Preferably, the electronic device in the embodiments of the present application can be a server or a computer device, etc.

[0033] Refer to Figure 1 , the service strategy determination method of this embodiment includes but is not limited to the following steps:

[0034] S110. Obtain the credit information of the target user.

[0035] The target user is used to evaluate whether it is a potential churn user in the financial field. The purpose of determining whether the target user is a potential churn user is to help the financial institution provide corresponding service strategies for the target user by implementing the solution provided by this embodiment on the credit information generated by the target user. By providing personalized services for the target user, it helps to increase the stickiness between the user and the financial institution and reduce the user churn rate.

[0036] In this embodiment, the credit information at least includes personal basic information, credit transaction information, financial institution interaction information, and credit score information. The credit information in this embodiment can be the relevant data of credit users collected from various databases of financial institutions, such as credit business databases, customer relationship management systems, etc.

[0037] Among them, personal basic information refers to the basic situation data related to individual credit users, which may include basic information such as name, age, gender, occupation, education level, contact information, and home address. These information can outline the basic profile of the user and reflect some potential characteristics and behavioral tendencies of the user in credit activities from a relatively macroscopic level. For example, different age and occupation groups often have differences in borrowing choices and repayment abilities; credit transaction information refers to various transaction-related data involved by users in past credit operations, which may include loan records (such as loan amount, loan term, loan disbursement time, repayment situation, such as whether overdue, number of overdue times, overdue duration, etc.) and credit card usage records (such as credit limit, monthly consumption amount, repayment amount, consumption frequency, etc.). This information can directly reflect the key information of the user's credit behavior performance and repayment ability, and plays an important role in evaluating the user's credit status and subsequent churn probability; financial institution interaction information refers to the relevant situations of communication and interaction between users and financial institutions, such as the number of communications with customer service, communication channels (online consultation, offline network consultation, etc.), types of consultation services (such as asking questions about interest rate adjustment, loan amount increase, etc.) and the frequency of business handling, etc. This information can reflect the user's attention, demand degree and satisfaction with financial services, and helps to insight into the user's dependence on financial institutions and possible churn tendencies; credit scoring information refers to the quantitative score given by a professional credit assessment institution or within a financial institution based on a certain scoring model and algorithm, comprehensively considering various credit-related factors of the user, so as to intuitively measure the level of the user's credit. The higher the credit score, the better the user's credit status, the lower the default risk, and there will be more advantages in credit operations, and it is closely related to whether the user continues to choose this financial institution to carry out credit operations. In this embodiment, the multi-dimensional data information collected is preprocessed and integrated to ensure that each piece of data can be accurately corresponding to the corresponding target user, so as to form a complete credit data set about the target user.

[0038] It should be noted that the relevant information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for display, data for analysis, etc.) involved in this application are all information and data that have been authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data comply with the relevant laws, regulations and standards of the relevant regions; and an operation entry is provided for the user to choose to agree or reject the automated decision result; if the user chooses to reject, the expert decision-making process will be entered to avoid relevant legal risks and public opinion risks.

[0039] S120. Input the personal basic information, credit transaction information, financial institution interaction information and credit scoring information into at least two evaluation models respectively, and obtain the evaluation results output by each evaluation model respectively.

[0040] In this embodiment, at least two evaluation models are obtained by constructing at least two sample data subsets from the same credit data set and training the at least two sample data subsets. Then, the at least two evaluation models provided in this embodiment are obtained through the following steps a) to c):

[0041] a) Obtain the historical credit information of multiple credit users within a preset time period to obtain a historical credit data set.

[0042] When the credit information obtained in step S110 includes at least personal basic information, credit transaction information, financial institution interaction information, and credit score information, the historical credit information obtained during the model training phase should also include the information content corresponding to the same information dimension, so as to facilitate improving the model analysis accuracy. Among them, the preset time period in this embodiment can be the past 1 year, 2 years, or 3 years, etc., and the selection of the specific preset time period is not limited here.

[0043] Integrate the above different types of credit information collected from various channels together to obtain an original data set. In this process, it is necessary to perform data cleaning operations, remove duplicate data records, correct obviously incorrect data (such as illogical age values, abnormally large or small loan amount data, etc.), and process missing values (appropriate methods such as mean filling, median filling, most frequent value filling, or prediction filling based on other relevant data can be used, specifically depending on the data characteristics and business background), and finally obtain a relatively complete and accurate historical credit data set, providing a reliable data basis for the subsequent steps.

[0044] b) Construct at least two sample data subsets from the historical credit data set based on a preset extraction principle; each sample data subset includes multiple sample information, and each sample information corresponds to a churn label.

[0045] Construct at least two sample data subsets from the historical credit dataset according to the preset extraction principle, and use the sample data included in each sample data subset as a training sample for the model training in the subsequent steps. Among them, the above-mentioned preset extraction principle is the sampling principle with replacement, which means that each time a sample data subset is extracted, the sample data in this sample data subset will still be retained in the database, and the sample data subsets constructed subsequently may extract some identical sample data. The advantage of doing this is that it can make each training set have a certain degree of randomness and diversity, so that there are both common data and unique data combinations among different training sets, laying a foundation for each subsequent decision model to learn different feature information. When extracting, it is necessary to ensure that the scale of each training set is relatively reasonable and can fully represent the data distribution characteristics in the entire database. For example, the data ratios of users with different credit levels, the data proportions of users in different age groups, etc. should be as consistent as possible with the overall situation of the database.

[0046] At the same time, for each sample information, it is necessary to determine its corresponding churn label, and the definition of the churn label is usually closely related to the business rules. For example, set an observation window (such as the next 6 months). During this observation period, if the user no longer has any new credit business dealings with the financial institution (including behaviors such as no longer applying for new loans and no longer making repayment operations, which can be regarded as terminating the business relationship), then mark this sample information as churn, represented by "1"; otherwise, if the user still maintains credit business contact with the financial institution during the observation period, then mark it as not churn, represented by "0". Of course, according to more complex and refined business analysis requirements, multi-class churn labels can also be set to reflect different degrees of churn possibilities, such as low churn risk, medium churn risk, high churn risk, etc., which are represented by different numbers or codes respectively.

[0047] c) Use the sample features corresponding to each sample information as input and the churn label corresponding to each sample information as output to train the decision tree model. When the output results of each decision tree model reach the convergence condition respectively, at least two evaluation models are obtained.

[0048] When extracting features from each sample information, the process of obtaining basic information features, transaction information features, interaction information features, and credit score features for personal basic information, credit transaction information, financial institution interaction information, and credit score information respectively can be achieved through the following methods:

[0049] For the basic information features, they may include age features (such as the interval to which the specific age belongs, facilitating the analysis of differences in credit behaviors among users of different age groups), occupation features (such as high-risk and high-income occupations, low-risk and stable-income occupations, etc., used to reflect the potential impact of occupation on credit), geographical features, or the effectiveness and activity of analyzing user contact information (such as whether the phone number is frequently changed, the usage frequency of email, etc.). These basic information features can reflect the role of the user's basic attributes in credit and churn from different aspects; for transaction information features, some dynamic transaction indicators can be calculated as features, such as the average repayment period (obtained by dividing the cumulative repayment duration by the number of repayments, which can reflect the regularity and timeliness of the user's repayment), the overdue ratio (calculated by dividing the number of overdue times by the total number of repayments, intuitively reflecting the user's repayment credit situation), the ratio of the remaining loan principal to the original loan principal (which can be used to understand the user's current debt burden), the ratio of the loan amount to the income (assuming income data can be obtained, used to measure the user's debt repayment pressure), etc. At the same time, the loan type, loan term, etc. can also be classified and coded, and the quantity and proportion of different types of loans can be counted. Through these features, the actual performance and potential risks of users in the credit transaction process can be deeply understood; for interaction information features, the frequency features of interaction with financial institutions can be counted, such as the average number of communications with customer service per month or per quarter, the average number of business transactions, etc., to reflect the user's attention and demand for financial services; analyze the theme distribution features of consulting services. By using text mining technology to extract keywords from the user's consulting content, the consulting content can be classified into different themes (such as interest rate related, quota adjustment related, product function related, etc.), and then calculate the proportion of the number of consultations in each theme to the total number of consultations to understand the key business areas that users are concerned about; in addition, the time interval features of interaction can also be considered (such as the time interval between two adjacent communications or business transactions, whether there is frequent interaction at special time points such as near the repayment date, before the loan maturity, etc.). These features help to grasp the interaction pattern between the user and the financial institution and the change of potential needs; for credit score features, in addition to directly using the credit score itself as a feature, the change trend feature of the credit score can also be calculated. For example, calculate the change rate of the credit score in the past six months (obtained by dividing the difference between the current credit score and the credit score six months ago by the credit score six months ago, which can reflect the dynamic change of the user's credit status); analyze the proportion of scores of the credit score in different dimensions (such as repayment history, credit limit usage, number of credit accounts, etc., specifically depending on the composition dimensions of the credit assessment model), and refine the credit score into multiple sub-features to more accurately explore the impact of various aspects of the credit status on user churn. Organize the sample features extracted from all aspects in a certain order and format to ensure that the feature data corresponding to each sample information is complete and standardized, so that it can be smoothly input into the decision tree model for training.

[0050] In the process of training a decision tree model with the sample features corresponding to each sample information as input and the churn label corresponding to each sample information as output, the decision tree model starts from the root node according to the set splitting conditions, and selects the feature that can best distinguish different churn labels (i.e., churn and non-churn) among numerous sample features to perform node splitting and construct the branch structure of the tree. For example, if it is found that the feature of "overdue ratio" can maximize the classification of samples into churn and non-churn categories under the current node, the samples will be divided into different child nodes according to different values of the "overdue ratio", and then this process will be repeated on the child nodes, continuously iterating to gradually construct the complete structure of the decision tree until the stopping condition is met or the preset number of training rounds is reached (such as setting the maximum number of training rounds to 100 rounds, etc.). After each training iteration is completed, the validation set data is input into the decision tree model at the current training stage, and relevant evaluation metrics are calculated. When these performance metrics of the model on the validation set no longer improve significantly, that is, the convergence condition is reached, the training is stopped, and a trained decision tree evaluation model is obtained at this time. In the same way, a corresponding decision tree evaluation model is trained for each sample data subset, and at least two evaluation models are finally obtained.

[0051] Specifically, in this embodiment, multiple splitting conditions are set in each decision tree model according to the priority of sample features. The priority of sample features in this embodiment can be arranged in descending order as follows: credit score feature, transaction information feature, interaction information feature, and basic information feature.

[0052] Furthermore, in the process of constructing the decision tree model, priority is given to using high-priority features as splitting conditions. When constructing the root node and upper-level nodes of the tree, first select the high-priority features that have the greatest impact on distinguishing churn and non-churn users for splitting. For example, if "credit score" is the feature with the highest priority, then at the root node, the samples will be divided into different subsets according to different value ranges of the "credit score". When the "credit score" can no longer effectively distinguish the churn situation of samples in a certain subset, then consider the next important feature (such as "transaction information feature", etc.) in order of priority as the subsequent splitting condition and continue to construct the branch structure of the tree downward. By setting multiple splitting conditions in this order of priority, the decision tree model can construct the tree structure more efficiently and reasonably, learn the patterns and rules related to user churn contained in the data faster and more accurately, avoid blindly selecting features for splitting, and thus improve the performance and prediction accuracy of each evaluation model.

[0053] Through the model training method provided in this embodiment, multiple splitting conditions are set in each decision tree model according to the sample feature priority, so that the model prioritizes node splitting and rule learning according to important features during the construction process, which helps each evaluation model to more accurately grasp the core factors affecting user churn, thereby improving the accuracy of its own predictions. A combination of multiple accurate models can further improve the overall accuracy of the prediction of user churn probability; further, each evaluation model is obtained by training with different sample data subsets, and different sample data subsets have their own unique feature distributions. By capturing different angles of credit information through different evaluation models, the accuracy of the final determination of the churn probability can be improved, avoiding the limitations and deviations caused by data analysis using a single model.

[0054] For another preferred implementation, please refer to Figure 2 , Figure 2 1 is another flow chart of the service strategy determination method provided in the embodiment of the present application. Specifically, in this embodiment, the above step S120 can be implemented by the following steps S121 to S123:

[0055] S121. For each evaluation model, feature extraction is performed on personal basic information, credit transaction information, financial institution interaction information and credit score information to obtain basic information features, transaction information features, interaction information features and credit score features.

[0056] In the model application stage, feature extraction is performed on personal basic information, credit transaction information, financial institution interaction information and credit score information to obtain basic information features, transaction information features, interaction information features and credit score features. The process is the same as the feature extraction method in the model training stage introduced in the aforementioned embodiment, and will not be repeated here.

[0057] S122. Extract derivative features from the transaction information features and the interaction information features respectively to obtain transaction derivative features and interaction derivative features.

[0058] Derived features are features that cannot be intuitively obtained through data information, and need to be further analyzed or inferred based on the extracted transaction information features. In this embodiment, the purpose of determining transaction derived features and interaction derived features is to help deeply explore the user's behavior patterns in credit transactions and interactions with financial institutions, and provide a strong basis for the subsequent formulation of targeted service strategies and risk management measures.

[0059] In this embodiment, the method for extracting derivative features from transaction information features to obtain transaction derivative features can be as follows: Based on the number of overdue times and the overdue duration, calculate the overdue severity score. Considering both the number of overdue times and the length of each overdue duration, obtain an indicator reflecting the severity of the overdue situation through a weighted calculation method as the first transaction derivative feature. Also, according to the repayment situations of different types of loans, calculate the comprehensive risk coefficients of various loans, taking into account factors such as loan amount, overdue ratio, remaining loan term, etc., and determine the comprehensive contribution degree of different loan types in the overall credit risk through a specific mathematical model (such as the analytic hierarchy process combined with weight assignment, etc.) as the second transaction derivative feature. The specific feature content included in the transaction derivative features is not limited herein.

[0060] The method for extracting derivative features from interaction information features to obtain interaction derivative features can be as follows: According to the time series of a user's inquiries about different business topics, analyze the shifting trend of their key business concerns. Through statistical analysis of the chronological order of inquiries and topic changes, judge the dynamic change direction of the user's recent demand for financial services, and form the first interaction derivative feature, which helps to grasp the changing trend of the interaction mode between the user and the financial institution. It can also be to combine the interaction frequency and the type of consulted business, calculate the participation enthusiasm index of the user for specific high-value services (such as inquiries about interest rate preferential activities, inquiries about increasing large loan amounts, etc.). By counting the number of inquiries of the user for such high-value services, the number of times of participating in relevant activities, etc., and comparing and analyzing with the total interaction situation, obtain the second interaction derivative feature, which can more accurately insight into the user's interest in important financial services and potential demand changes. The specific feature content included in the interaction derivative features is not limited herein.

[0061] S123. Input the basic information features, transaction information features, interaction information features, credit score features, transaction derivative features, and interaction derivative features into at least two evaluation models to obtain the evaluation results respectively output by each evaluation model.

[0062] Organize and standardize the sorted basic information features, transaction information features, interaction information features, credit score features, transaction-derived features, and interaction-derived features according to the input formats required by each evaluation model (for example, perform unified normalization processing on various types of feature data to make their numerical ranges within appropriate intervals for convenient model calculation and comparison), ensuring that each evaluation model can accurately receive complete and standardized feature data as input information; input the processed multi-dimensional feature data into at least two evaluation models (these evaluation models are obtained through corresponding training in the foregoing embodiments). Each evaluation model will calculate and analyze the input feature data based on its internal algorithm structure, parameter settings, and learned patterns, and finally output respective evaluation results regarding aspects such as the user churn probability. For example, the result output by a certain evaluation model may be that the user churn probability is 0.3 (indicating a 30% likelihood of churn), and the result output by another model may be 0.25, etc. Due to different learned feature patterns and focuses, there will be certain differences in the evaluation results output by different models, but all provide a reference basis for subsequent comprehensive judgment of user churn situations.

[0063] The implementation methods of the above steps S121 - S123, through multi-level feature extraction of different types of information and mining of derived features, greatly enrich the feature dimensions input into the evaluation model, enabling the model to consider the user's behavior performance and potential churn tendency from more perspectives and more comprehensively, thereby improving the accuracy of evaluation results such as the user churn probability. Input the data containing multi-dimensional features into at least two evaluation models. Different models, based on their own algorithm characteristics and training and learning situations, will have different understandings and weight distributions of the features, and can analyze and predict the user churn situation from their unique perspectives.

[0064] S130. Determine the target user and determine the target service strategy according to at least two evaluation results.

[0065] In this embodiment, when there are at least two evaluation models, each model outputs an evaluation result, and the current evaluation result can be reflected in the form of the value of the user churn probability. Then, the method of determining the target user and determining the target service strategy according to at least two evaluation results can be implemented as follows: determine the churn result of the target user by calculating the mean value of at least two evaluation results; it can also be determined by dividing the probability interval according to the churn degree, determining the corresponding probability intervals according to at least two evaluation results respectively, and then obtaining the corresponding churn degree according to the corresponding probability intervals, and determining the target churn degree with the highest number of votes as the target churn degree of the target user by voting. The specific method of determining the churn probability of the target user is not limited here.

[0066] In a preferred implementation, the above step S130 can be implemented in the following manner:

[0067] Determine the churn probability of the target user based on at least two evaluation results; determine a target service strategy for the target user according to the churn probability. The advantage of doing this is that by determining the churn probability through multiple evaluation results, it can intuitively and accurately reflect the dependence of the target user on the financial institution, and then help improve the accuracy of formulating the target service strategy to personalize the service for the target user based on the target service strategy.

[0068] When the probability value is relatively high, it indicates that the target user has a low degree of dependence on the financial institution and a high churn probability; when the probability value is relatively low, it indicates that the target user has a high degree of dependence on the financial institution and a low churn probability.

[0069] Specifically, the solution for determining the churn probability of the target user according to at least two of the evaluation results provided in this embodiment is implemented in the following manner:

[0070] Obtain the reference weight corresponding to each evaluation model, where the reference weight is obtained based on the sample characteristics of the sample data subset corresponding to each evaluation model; calculate according to the reference weight and the corresponding evaluation result of each evaluation model to obtain the churn probability of the target user.

[0071] For each evaluation model, since the feature information in the sample data subset used during training has certain differences, and each sample information corresponds to a series of sample characteristics (such as personal basic characteristics, credit transaction characteristics, etc.), a feature importance evaluation tool built into the learning algorithm can be used to analyze the contribution degree of the features to the prediction target (i.e., whether the user churns). For example, use the random forest algorithm to train the sample data subset. During the training process, the random forest will automatically count the importance scores of each feature in aspects such as constructing decision trees and reducing classification errors. For an evaluation model corresponding to a certain sample data subset, those evaluation models corresponding to the subsets of the features determined to contribute more to distinguishing whether the user churns in the random forest can be given relatively higher reference weights. Then, when it is necessary to determine the churn probability of the target user, for example, there are three evaluation models. The reference weight of model A is 0.4, and the output evaluation result is 0.3; the reference weight of model B is 0.3, and the output evaluation result is 0.28; the reference weight of model C is 0.3, and the output evaluation result is 0.32. Then the churn probability of the target user = 0.3×0.4 + 0.28×0.3 + 0.32×0.3 = 0.3.

[0072] Through the method for determining the churn probability of target users provided in this embodiment, by pre-determining the reference weights of each evaluation model, when calculating the churn probability of target users by integrating the results of multiple models, the importance of different factors can be more accurately reflected, thereby improving the accuracy of the overall prediction of user churn probability.

[0073] Optionally, taking the process degree including a low churn probability interval (0 - 0.3), a medium churn probability interval (0.3 - 0.7), and a high churn probability interval (0.7 - 1) as an example for illustration, the method for determining the target service strategy for target users based on the churn probability in this embodiment can be as follows:

[0074] Determine the target probability interval from the policy mapping table according to the churn probability. In the policy mapping table, a service policy corresponds to a probability interval; determine the service policy corresponding to the target probability interval as the target service policy. By personalizing and precisely customizing different service policies for users with different churn probabilities in this embodiment, the customer retention rate and the service quality and competitiveness of financial institutions can be effectively improved. For the identified users with a high churn probability, by taking corresponding service strategies in a timely manner for early intervention, the potential credit risks brought by user churn can be effectively reduced, the safety of the credit assets of financial institutions can be guaranteed, and the stable operation of financial business can be maintained.

[0075] Among them, for users with a high churn probability (such as the churn probability is greater than 0.7): Provide extremely attractive retention offers for such users, such as significantly reducing the annual credit card fee, giving high-value points or gifts, providing low-interest installment plans, and having a dedicated customer service team for one-on-one communication to understand their needs and solve problems. At the same time, recommend some long-term preferential financial product combinations to increase user stickiness.

[0076] For users with a medium churn probability (such as the churn probability is between 0.3 - 0.7): Provide targeted preferential activities, such as increasing the cashback ratio for specific consumption categories, temporarily increasing the credit limit, inviting to participate in exclusive credit card member activities. At the same time, push personalized financial management suggestions and credit card usage skills through text messages or the APP to guide users to use credit card services more and enhance their favorability towards the bank.

[0077] For users with a low churn probability (such as the churn probability is less than 0.3): Maintain regular service care, such as birthday blessings, pushing festival preferential information, and giving small-point rewards, encourage users to continue to maintain good usage habits, and timely recommend some value-added services or new products that match the user's consumption habits, such as introducing the upgrade rights and interests of high-end credit cards, etc., to further strengthen user stickiness.

[0078] Preferably, for the service strategy determination method provided in this embodiment, the following solution can further be executed:

[0079] Obtain the behavioral data of the target user based on the service strategy feedback; evaluate the target service strategy according to the behavioral data and customer satisfaction. The purpose of doing this is to apply the formulated service marketing strategy to the actual business and regularly evaluate the effectiveness of the strategy; according to the evaluation results, adjust and optimize the service marketing strategy in a timely manner to continuously improve customer satisfaction and loyalty.

[0080] Specifically, in this embodiment, the behavioral data of the target user based on the service strategy feedback can be collected from various business systems of financial institutions. The specific behavioral data may include: obtaining the loan behavioral data of users after the implementation of the service strategy, such as whether there are new loan applications, the type of loan applied for (consumer loan, housing loan or other types), the loan amount, the loan term, etc. These data can reflect whether the service strategy stimulates the credit demand of users or changes the user's preference for credit products; obtaining the repayment behavioral data of users, including whether they repay on time, the change in the number of overdue times and overdue duration, early repayment situation, etc. If the service strategy aims to improve the repayment willingness and ability of users, then the repayment behavioral data is a key indicator for evaluating its effectiveness. For example, if the overdue situation of users significantly decreases after providing repayment discounts or personalized repayment plan adjustments, it indicates that the service strategy has played a positive role in this regard; further, the interaction behavioral data with the financial institution can also be obtained by counting the communication frequency and methods between the user and the financial institution. If the service strategy involves improving the customer service quality or increasing the communication channels with users, then by observing the number of times the user actively contacts the customer service and the communication methods (such as whether they are more inclined to use the newly opened online communication channels), the feedback of users on the service strategy in terms of interaction can be judged, and note the changes in the content of users' business consultations. For example, if the service strategy is to recommend value-added services, by analyzing whether the number of consultations related to value-added services by users increases and the depth of the consultations (such as whether they understand the details and fees of value-added services in detail), the interest and acceptance degree of users in value-added services can be understood; the marketing behavioral data can also be recorded by recording the participation of users in the marketing activities of financial institutions, such as whether they respond to the preferential activities pushed based on the service strategy. This includes whether they click on the activity link, whether they successfully participate in the activity (such as successfully applying for a preferential loan product, receiving a coupon, etc.), and the subsequent behaviors after participating in the activity (such as whether they continue to use the relevant financial services after the activity). These data can intuitively reflect the feedback of users on the marketing part of the service strategy.

[0081] Furthermore, the obtained loan behavior data, repayment behavior data, interaction behavior data with financial institutions, and marketing behavior data can be quantitatively calculated, and combined with the results of the user satisfaction survey for comprehensive calculation and evaluation to evaluate the feasibility of the target service strategy. If the comprehensive evaluation score is high and each indicator performs well, it indicates that the service strategy is effective and can be continued or appropriately optimized; if the score is low or some key indicators (such as the customer satisfaction indicator or the improvement rate of repayment quality) perform poorly, it is necessary to deeply analyze the reasons, identify the problems existing in the service strategy, such as the service content not meeting the user's needs, poor publicity, etc., and then make targeted adjustments and improvements to the service strategy to achieve the effect of increasing customer stickiness.

[0082] The service strategy determination method provided in this embodiment first obtains the credit information of the target user. The credit information in this embodiment at least includes personal basic information, credit transaction information, financial institution interaction information, and credit score information. By obtaining the information of different dimensions of the target user, a complete data basis is provided to improve the accuracy of the subsequent determination of the model output evaluation result; then, at least two evaluation models are used to analyze the input credit analysis respectively, so as to capture different angle characteristics of the credit information through different evaluation models, which helps to formulate a precise service strategy for the target user and avoid the limitations and biases brought by data analysis using a single model; finally, the target service strategy is determined according to at least two evaluation results. The solution provided in this embodiment can determine the user churn situation through at least two evaluation results, and then improve the user retention rate by implementing the corresponding service strategy, achieving the beneficial effects of providing personalized services for different users, enhancing the user experience, and reducing the user churn rate.

[0083] Figure 3 It is a structural schematic diagram of a service strategy determination device provided by an embodiment of the present application. This device is applicable to execute the service strategy determination method provided by an embodiment of the present application. As Figure 3 shown, this device may specifically include: an information acquisition module 310, a result output module 320, and a strategy determination module 330, where:

[0084] The information acquisition module 310 is used to acquire the credit information of the target user, and the credit information at least includes personal basic information, credit transaction information, financial institution interaction information, and credit score information;

[0085] A result output module 320 is configured to input the personal basic information, the credit transaction information, the financial institution interaction information, and the credit score information into at least two evaluation models respectively, and obtain evaluation results respectively output by each of the at least two evaluation models; the at least two evaluation models are obtained by constructing at least two sample data subsets from the same credit data set and training the at least two sample data subsets.

[0086] A policy determination module 330 is configured to determine a target service policy for the target user according to the at least two evaluation results.

[0087] The service policy determination device provided in this embodiment first obtains the credit information of the target user. The credit information in this embodiment at least includes personal basic information, credit transaction information, financial institution interaction information, and credit score information. By obtaining information from different dimensions of the target user, a complete data basis is provided to improve the accuracy of the evaluation results output by the subsequent determination model; then, by analyzing the input credit analysis through at least two evaluation models respectively, different-angle features of the credit information can be captured by different evaluation models, which helps to formulate a precise service policy for the target user and avoid limitations and biases brought by data analysis using a single model; finally, the target service policy for the target user is determined according to the at least two evaluation results. The solution provided in this embodiment can determine the user churn situation through at least two evaluation results, and then improve the user retention rate by implementing corresponding service policies, achieving the beneficial effects of providing personalized services for different users, enhancing the user experience, and reducing the user churn rate.

[0088] In one embodiment, the result output module 320 is specifically configured to, for each of the evaluation models, extract features from the personal basic information, the credit transaction information, the financial institution interaction information, and the credit score information respectively to obtain basic information features, transaction information features, interaction information features, and credit score features; extract derivative features from the transaction information features and the interaction information features respectively to obtain transaction derivative features and interaction derivative features; input the basic information features, transaction information features, interaction information features, credit score features, transaction derivative features, and interaction derivative features into at least two evaluation models to obtain evaluation results respectively output by each of the at least two evaluation models.

[0089] In one embodiment, the policy determination module 330 includes a probability calculation unit and a policy determination unit, where:

[0090] The probability calculation unit is configured to determine the churn probability of the target user according to the at least two evaluation results.

[0091] A policy determination unit for determining a target service policy for the target user according to the churn probability.

[0092] In one embodiment, the probability calculation unit is specifically configured to obtain the reference weight corresponding to each of the evaluation models, where the reference weight is obtained based on the sample features of the sample data subset corresponding to each of the evaluation models; calculate according to the reference weight and the corresponding evaluation result of each of the evaluation models to obtain the churn probability of the target user.

[0093] In one embodiment, the policy determination unit is specifically configured to determine a target probability interval from the policy mapping table according to the churn probability, and in the policy mapping table, a service policy corresponds to a probability interval; determine the service policy corresponding to the target probability interval as the target service policy.

[0094] In one embodiment, the device further includes: a data acquisition module and a policy evaluation module, where:

[0095] The data acquisition module is configured to acquire the behavior data fed back by the target user based on the service policy.

[0096] The policy evaluation module is configured to evaluate the target service policy according to the behavior data and customer satisfaction.

[0097] In one embodiment, the device further includes a subset construction module and a model training module, where:

[0098] The information acquisition module 310 is further configured to acquire the historical credit information of multiple credit users within a preset time period to obtain the historical credit data set.

[0099] The subset construction module is configured to construct at least two sample data subsets from the historical credit data set based on a preset extraction principle; each sample data subset includes multiple sample information, and each sample information corresponds to a churn label.

[0100] The model training module is configured to use the sample features corresponding to each sample information as inputs and the churn label corresponding to each sample information as outputs to train the decision tree model. When the output results of each decision tree model reach the convergence condition, at least two evaluation models are obtained, and multiple splitting conditions are set in each decision tree model according to the sample feature priorities.

[0101] In one embodiment, the sample feature priorities are arranged from high to low as: credit score feature, transaction information feature, interaction information feature, and basic information feature.

[0102] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above division of each functional module is used as an example. In actual applications, the above functions can be allocated to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. The specific working process of the above-described functional modules can refer to the corresponding process in the foregoing method embodiments and will not be elaborated herein.

[0103] An embodiment of the present application further provides an electronic device, where the electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the service policy determination method according to any embodiment of the present application.

[0104] An embodiment of the present application further provides a computer-readable medium, where the computer-readable storage medium stores computer instructions, and the computer instructions are used to implement the service policy determination method according to any embodiment of the present application when executed by a processor.

[0105] Next, refer to Figure 4 , Figure 4 which is a schematic structural diagram of the electronic device provided by the embodiment of the present application. It shows a schematic structural diagram of a computer system 500 of the electronic device suitable for implementing the embodiment of the present application. Figure 4 The illustrated electronic device is only an example and should not impose any limitation on the functions and usage scope of the embodiment of the present application.

[0106] As Figure 4 shown, the computer system 500 includes a central processing unit (CPU) 501, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 502 or the program loaded from the storage section 508 into the random access memory (RAM) 503. In the RAM 503, various programs and data required for the operation of the system 500 are also stored. The CPU 501, ROM 502, and RAM 503 are connected to each other through a bus 504. The input / output (I / O) interface 505 is also connected to the bus 504.

[0107] The following components are connected to the I / O interface 505: an input section 506 including a keyboard, a mouse, etc.; an output section 507 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc. as well as a speaker, etc.; a storage section 508 including a hard disk, etc.; and a communication section 509 including a network interface card such as a LAN card, a modem, etc. The communication section 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to the I / O interface 505 as needed. A removable medium 511, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 510 as needed so that a computer program read therefrom is installed into the storage section 508 as needed.

[0108] Specifically, according to the embodiments disclosed in the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the embodiments disclosed in the present application include a computer program product that includes a computer program carried on a computer-readable medium, and the computer program includes program codes for performing the methods shown in the flowcharts. In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 509, and / or installed from the removable medium 511. When the computer program is executed by a central processing unit (CPU) 501, the above-described functions defined in the system of the present application are executed.

[0109] It should be noted that the computer-readable medium shown in this application can be a computer-readable signal medium, a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples of a computer-readable storage medium can include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this application, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium can include a data signal propagated in a baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on a computer-readable medium can be transmitted using any appropriate medium, including but not limited to: wireless, wire, and optical cable, etc., or any suitable combination of the above.

[0110] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram can represent a module, a program segment, or a part of code that contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks can occur in a different order than that marked in the accompanying drawings. For example, two consecutive blocks shown can actually be executed substantially in parallel, and they can sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram or flowchart, and the combination of blocks in the block diagram or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0111] The modules and / or units involved in the embodiments of the present application can be implemented in software or in hardware. The described modules and / or units can also be provided in a processor. For example, it can be described as: a processor includes an information acquisition module, a result output module, and a policy determination module. Among them, the names of these modules do not constitute a limitation to the module itself in some cases.

[0112] As another aspect, the present application also provides a computer-readable medium. The computer-readable medium can be included in the device described in the above embodiments; it can also exist separately without being assembled into the device. The above computer-readable medium carries one or more programs. When the above one or more programs are executed by the device, the device includes: acquiring credit information of a target user, where the credit information at least includes personal basic information, credit transaction information, financial institution interaction information, and credit score information; respectively inputting the personal basic information, the credit transaction information, the financial institution interaction information, and the credit score information into at least two evaluation models to obtain evaluation results respectively output by each of the at least two evaluation models; the at least two evaluation models are obtained by constructing at least two sample data subsets from the same credit data set and training the at least two sample data subsets; determining a target service policy for the target user according to the at least two evaluation results.

[0113] According to the technical solution of this embodiment, it is possible to determine the user churn situation through at least two evaluation results, and then improve the user retention rate by implementing corresponding service policies, achieving the beneficial effects of providing personalized services for different users, enhancing the user experience, and reducing the user churn rate.

[0114] The above specific implementation manners do not constitute a limitation to the protection scope of the present application. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can occur depending on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principle of the present application shall be included within the protection scope of the present application.

Claims

1. A method for determining a service strategy, characterized in that: include: Obtaining the target user's credit information, which includes at least basic personal information, credit transaction information, financial institution interaction information, and credit score information; Inputting the basic personal information, the credit transaction information, the financial institution interaction information and the credit score information into at least two evaluation models respectively, and obtaining evaluation results output by each evaluation model respectively; at least two evaluation models are obtained by constructing at least two sample data subsets for the same credit data set, so as to train the at least two sample data subsets; The target user is determined according to at least two of the evaluation results and a target service strategy is determined.

2. The service strategy determination method according to claim 1, characterized in that: The step of inputting the basic personal information, the credit transaction information, the financial institution interaction information and the credit score information into at least two evaluation models respectively to obtain evaluation results outputted by each evaluation model respectively includes: For each of the evaluation models, feature extraction is performed on the personal basic information, the credit transaction information, the financial institution interaction information and the credit score information to obtain basic information features, transaction information features, interaction information features and credit score features; Extracting derivative features from the transaction information features and the interaction information features respectively to obtain transaction derivative features and interaction derivative features; The basic information features, transaction information features, interactive information features, credit score features, transaction derivative features and interactive derivative features are input into at least two evaluation models to obtain evaluation results output by each of the evaluation models.

3. The service strategy determination method according to claim 1, characterized in that: The step of determining the target service strategy for the target user according to at least two of the evaluation results includes: Determine the churn probability of the target user according to at least two of the evaluation results; A target service strategy is determined for the target user according to the churn probability.

4. The service strategy determination method according to claim 3, characterized in that: Determining the churn probability of the target user according to at least two of the evaluation results includes: Obtaining reference weights corresponding to each of the evaluation models, wherein the reference weights are obtained based on sample features of a sample data subset corresponding to each of the evaluation models; The churn probability of the target user is obtained by performing calculations based on the reference weights and the corresponding evaluation results corresponding to each of the evaluation models.

5. The service strategy determination method according to claim 3, characterized in that: Determining a target service strategy for the target user according to the churn probability includes: Determine a target probability interval from a strategy mapping table according to the churn probability, wherein one service strategy corresponds to one probability interval in the strategy mapping table; Determine the service strategy corresponding to the target probability interval as the target service strategy.

6. The service strategy determination method according to claim 1, characterized in that: After determining the target user and determining the target service strategy according to at least two of the evaluation results, the method further includes: Acquire behavior data of the target user based on feedback of the service strategy; The target service strategy is evaluated based on the behavioral data and customer satisfaction.

7. The service strategy determination method according to claim 1, characterized in that: At least two of the evaluation models are obtained by: Acquire historical credit information of multiple credit users within a preset time period to obtain the historical credit data set; Constructing at least two sample data subsets from the historical credit data set based on a preset extraction principle; each of the sample data subsets includes a plurality of sample information, and each of the sample information corresponds to a loss label; The sample features corresponding to each of the sample information are taken as input, and the loss labels corresponding to each of the sample information are taken as output to train the decision tree model. When the output results of each of the decision tree models reach the convergence conditions, at least two evaluation models are obtained, and multiple splitting conditions are set in each decision tree model according to the sample feature priority.

8. The service strategy determination method according to claim 7, characterized in that: The priorities of the sample features from high to low are: credit score features, transaction information features, interactive information features and basic information features.

9. A service strategy determination device, characterized in that: include: An information acquisition module is used to acquire the credit information of the target user, wherein the credit information includes at least basic personal information, credit transaction information, financial institution interaction information and credit score information; A result output module is used to input the basic personal information, the credit transaction information, the financial institution interaction information and the credit score information into at least two evaluation models respectively, and obtain the evaluation results output by each evaluation model respectively; at least two evaluation models are obtained by constructing at least two sample data subsets for the same credit data set to train the at least two sample data subsets; A policy determination module is used to determine a target service policy for the target user based on at least two of the evaluation results.

10. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the service policy determination method according to any one of claims 1 to 8.

Images