Member strategy processing method, system and device and storage medium

Abstract

The invention relates to the technical field of computers, and discloses a member strategy processing method, system and device and a storage medium. The method comprises the following steps: collecting target information of a user, wherein the target information comprises historical payment behavior data, historical use behavior data and fund change data of a platform account; based on the historical payment behavior data and fund change data of a platform account, obtaining a payment capability level of the user; and based on the historical use behavior data, obtaining the loss risk level of the user, generating a member maintenance strategy for the user based on the payment capability level and the loss risk level of the user; and based on the member maintenance strategy, executing a fee deduction operation of subscribing a member service on a platform account of the user. By adopting the method, the accurate personalized member maintenance strategy can be customized for the user, and the user experience and the member retention rate are improved.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a membership policy processing method, system, device and storage medium. Background Technology

[0002] In the era of digital services, automatic membership renewal has become a common business model for multimedia content platforms to provide continuous services. For example, automatic membership renewal services exist for everything from video websites and music websites to various software tools and knowledge payment platforms.

[0003] However, current automated membership maintenance solutions mostly use a fixed deduction model, lacking consideration for individual user differences, often resulting in membership subscription payment failures. For example, when a user's membership expires, insufficient funds in the user's platform account can cause automatic deduction failure. Consequently, this leads to a low success rate for automatic membership renewal, decreased user satisfaction, and significant member churn. Therefore, improving the success rate of automatic membership renewal and member retention within the platform is a problem that needs to be addressed. Summary of the Invention

[0004] In view of this, this application aims to provide a membership strategy processing method, system, device and storage medium to improve the success rate of automatic membership renewal and membership retention rate within the platform.

[0005] To achieve the above objectives, the technical solution of this application is as follows: The first aspect of this application provides a membership strategy processing method, the method comprising: Collect user target information, which includes: historical payment behavior data, historical usage behavior data, and platform account fund change data; the historical payment behavior data is the data generated by users purchasing goods and subscribing to membership services on the platform; Based on the historical payment behavior data and the platform account's fund change data, the user's payment ability level is obtained; and based on the historical usage behavior data, the user's churn risk level is obtained. Based on the user's payment ability level and churn risk level, generate a membership maintenance strategy for the user; Based on the aforementioned membership maintenance strategy, the platform account of the user will be charged for the subscription membership service.

[0006] According to a second aspect of the embodiments of this application, a membership strategy processing system is provided for implementing the steps of the method provided in the first aspect of the embodiments of this application, the system comprising: The data collection module is configured to collect target information of users, including: historical payment behavior data, historical usage behavior data, and fund change data of the platform account; the historical payment behavior data is the data generated by users purchasing goods and subscribing to membership services within the platform. The analysis module is configured to obtain the user's payment ability level based on the historical payment behavior data and the platform account's fund change data; and to obtain the user's churn risk level based on the historical usage behavior data. The strategy generation module is configured to generate a membership maintenance strategy for the user based on the user's payment ability level and churn risk level. The deduction execution module is configured to perform a deduction operation for the user's platform account for subscription services based on the membership maintenance strategy.

[0007] According to a third aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the method provided in the first aspect of the present application.

[0008] According to a fourth aspect of the present application, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When executed by the processor, the computer program implements the steps of the method provided in the first aspect of the present application.

[0009] The membership strategy processing method provided in this application analyzes each user's historical usage behavior, historical payment behavior, and fund change information on the platform to specifically assess each user's payment ability and churn risk. Based on this, precise personalized membership maintenance strategies are customized for users with different payment abilities and churn risks. The platform account is then charged for subscription services according to the membership maintenance strategy, thus better adapting to different users' spending power and habits, improving the success rate of automatic membership renewal and membership retention, and enhancing user experience. Attached Figure Description

[0010] 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.

[0011] Figure 1 This is a flowchart of the membership strategy processing method proposed in this application; Figure 2This is a flowchart of an embodiment of the automatic renewal of a membership platform account in this application; Figure 3 This is a flowchart of the workflow of each module in the membership strategy processing system in one embodiment of this application; Figure 4 This is a schematic diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0012] 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.

[0013] It should be understood that the phrase "one embodiment" or "an embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments.

[0014] In the various embodiments of this application, it should be understood that the sequence number of each process described below 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.

[0015] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects as detailed in this application.

[0016] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0017] Most current automatic membership renewal plans use fixed deduction times, cycles, and pricing strategies, lacking consideration for individual user differences. That is, regardless of changes in a user's recent viewing frequency or spending power, the platform account is automatically deducted from the user's account on a fixed date at a fixed amount. This often leads to some users failing to renew because their platform account does not have enough funds at the time of deduction.

[0018] This application analyzes the historical behavior of member users to assess their payment ability and churn risk, and generates personalized membership maintenance strategies for different users. The application will now be described in detail with reference to the accompanying drawings and embodiments.

[0019] Figure 1 This is a flowchart of a membership policy processing method proposed in an embodiment of this application. Figure 1 As shown, the method includes: S1: Collect user target information, including: historical payment behavior data, historical usage behavior data, and platform account fund change data; the historical payment behavior data is the data generated by the user purchasing goods and subscribing to membership services on the platform; S2: Based on the historical payment behavior data and the platform account's fund change data, obtain the user's payment ability level; and based on the historical usage behavior data, obtain the user's churn risk level; S3: Generate a membership maintenance strategy for the user based on the user's payment ability level and churn risk level; S4: Based on the membership maintenance strategy, perform a deduction operation for the user's platform account to subscribe to membership services.

[0020] In this embodiment, by collecting historical behavioral data of member users within the platform and analyzing and assessing users' payment capabilities and churn risk, personalized membership maintenance strategies are customized for member users. First, target information of the user is collected, including historical payment behavior data, historical usage behavior data, and platform account fund change data. Specifically, historical payment behavior data refers to data generated by users purchasing goods and subscribing to memberships within the platform, including: payment time, payment amount, payment channel, etc.; platform account fund change data includes: the user's platform account balance, the time of fund inflow and outflow from the platform account, and the amount of fund changes, etc. Historical usage behavior data includes: the user's login frequency, number of check-ins, usage duration, browsed content, and interactive behaviors within the platform (such as likes, comments, and shares).

[0021] When a user purchases goods or subscribes to membership services on the platform, the platform account generates corresponding fund change data. The platform account balance is generated when the user tops up the platform account, and during the top-up process, the platform account generates corresponding fund inflow information (including the time of fund inflow, the amount of fund change, etc.).

[0022] If the platform account has sufficient balance, the balance can be used to pay for orders for goods and subscription services within the platform. During the payment process using the balance, the platform account generates corresponding information on the outflow of funds (including the time of the outflow and the amount of the change).

[0023] If the platform account has no balance or insufficient funds to pay for an order, the platform account will complete the deduction operation through the user's linked third-party payment platform or financial institution. The platform account initiates a deduction request to the third-party payment platform or financial institution, which then transfers the funds corresponding to the order amount to the platform account, and the payment is completed through the platform account. During this process, the platform account generates corresponding information on fund inflows and outflows (including the time of fund inflow and outflow, and the amount of fund changes).

[0024] Based on the collected historical behavioral data and platform account fund change data, an analysis is conducted to assess users' payment ability and churn risk. First, the collected data undergoes data cleaning to remove noise and anomalies. For example, obviously erroneous payment amounts (e.g., negative payments) or illogical usage durations (e.g., usage exceeding 24 hours in a single day) are corrected or deleted. Then, the cleaned data undergoes standardization and normalization to eliminate dimensional differences and make data from different dimensions comparable. Specifically, the data for each dimension is normalized, mapping the numerical range of each dimension to the same numerical range (e.g., [0,1]). For example, payment amounts and usage durations are normalized separately, mapping their respective numerical ranges to the [0,1] range.

[0025] When analyzing historical behavioral data and platform account fund change data, a comprehensive approach is taken, considering multiple dimensions of user behavior (such as payment behavior, usage habits, and platform account fund changes). This multi-dimensional data allows for a comprehensive and accurate assessment of users' payment capabilities and churn risk. Consequently, the membership maintenance strategies generated based on the assessment results are more targeted and effective. Then, based on the assessment results of payment capability level and churn risk level, a matching personalized membership maintenance strategy is generated. The platform account of each user is then charged according to the corresponding membership maintenance strategy.

[0026] This solution considers multi-dimensional data such as user payment behavior, usage habits, and account fund changes to comprehensively analyze users' payment ability and churn risk. Compared with traditional analysis methods that rely on only a single dimension of data such as payment amount or usage frequency, it can more accurately and comprehensively assess users' payment ability and churn risk. This allows for the generation of personalized membership maintenance strategies tailored to different users, meeting their different payment habits and needs, improving the success rate of automatic membership renewal and membership retention, and enhancing the user experience.

[0027] As one embodiment of this application, step S3 above includes: S31-1: When the user's payment ability level is the first payment ability level and the churn risk level is the first churn risk level, based on the user's historical payment behavior data and the platform account's fund change data, determine the target time period when the user has sufficient funds; the first payment ability level indicates that the user's payment behavior is unstable, or the user's average payment amount is less than the amount threshold; the first churn risk level indicates that the user's usage frequency has not decreased, or the user's interest in platform activities has not decreased; S32-1: Determine a target date from the target time period and postpone the user's billing date to the target date; The above step S4 specifically includes: S41: Deduct the subscription fee for the user's platform account according to the target date.

[0028] In this embodiment, if a user has high payment ability and low churn risk, the default deduction policy can be maintained, and the user's platform account can be charged on the default deduction date. For users with low payment ability but a willingness to renew, the deduction date can be appropriately extended to ensure the success rate of membership renewal and membership retention. Specifically, if a user's payment ability is at the first payment ability level (i.e., low payment ability) and the churn risk is at the first churn risk level (i.e., low churn risk), it indicates that although the user may have insufficient funds at times, they still have the willingness to renew.

[0029] Based on this, further analysis of the user's historical payment behavior data and platform account fund change data is conducted to determine the target time period for sufficient funds in the user's account. Specifically, the target time period for sufficient funds in the user's platform account is derived by analyzing the stable inflow time of funds and the payment time of consumption orders within the platform.

[0030] Based on a preset statistical period (e.g., 6 months), the first statistical result is obtained by counting the dates on which the user recharged their platform account each month within the most recent statistical period. It is worth noting that if no recharge was made in a given month, the date in the statistics is empty. The second statistical result is obtained by counting the payment times of purchases made by the user from the platform's online store each month within the statistical period. Again, it is worth noting that if no purchases were made in a given month, the date in the statistics is empty. Based on the dates recorded in the first and second statistical results, the dates are sorted from highest to lowest frequency of occurrence, resulting in a sorting result. Based on at least two dates with the highest frequency in this sorting result, at least one candidate time period with a number of days not exceeding a first value (e.g., 5) is generated. One candidate time period is selected as the target time period for which the user has sufficient funds. Then, a target date is determined within the target time period, and the user's deduction date is postponed to that target date. It is worth noting that if the target date is before the current deduction date, the target date is postponed to the same day of the next month.

[0031] For example, the statistics show that the three dates with the highest frequency are the 16th (4 times), the 15th (2 times), and the 20th (2 times). Therefore, the candidate time period is the 16th-20th. The 16th-20th of each month is identified as the target time period for sufficient funds.

[0032] Based on the prediction results of users' payment ability and churn risk, this embodiment uses a dynamic adjustment mechanism to flexibly adjust the deduction date for users' platform accounts through a dynamic adjustment mechanism for users with low payment ability but a willingness to renew. This can better adapt to the needs and payment ability of such users, reduce the churn rate of platform members, improve the membership retention rate and renewal success rate of users with low payment ability, and meet the personalized needs of users.

[0033] As one embodiment of this application, after step S41 above, the method further includes: S42: If the payment is successfully deducted, generate new payment behavior data and update the user's historical payment behavior data; S43: In the event of a failed deduction, determine the retry deduction cycle for the user; and retry deducting fees from the user's platform account according to the retry deduction cycle.

[0034] In this embodiment, after adjusting the deduction date for users with low payment ability, the platform account of the user is charged according to the deduction date. If the deduction is successful, the deduction information is recorded and the user's membership validity period is updated, new payment behavior data is generated, and the newly generated data is added to the user's historical payment behavior data. If the deduction fails, the user's membership benefits are not retained. Furthermore, this embodiment also sets a corresponding retry deduction cycle for users with low payment ability. Based on the retry deduction cycle and the adjusted deduction date, the user's platform account is recharged, further reducing the probability of deduction failure and improving member retention rate and membership renewal success rate.

[0035] Optionally, to improve the success rate of retrying billing, the retry period for user accounts with lower payment capabilities is longer than the platform's default retry period for user accounts. For example, if the platform's default retry period for user accounts is 1 day, for users with lower payment capabilities but a willingness to renew, the retry period can be set to 2 days. If a billing attempt to these users' platform accounts fails based on the adjusted billing date, a retry attempt will be made after a 2-day interval.

[0036] Optionally, a notification message can be generated if the payment fails, and the user can be notified of the upcoming retry time via in-platform message or SMS, so that the user can proactively keep sufficient funds in the payment platform account.

[0037] By flexibly setting deduction dates and retry deduction cycles for users with different payment capabilities and churn risks, a flexible membership maintenance strategy can be formed, improving user experience and membership renewal success rate, and reducing membership churn caused by renewal factors.

[0038] As one embodiment of this application, the historical usage behavior data is data generated by users logging in, checking in, browsing, and interacting within the platform; the target information also includes basic data, which includes: registration time, age, and gender; step S3 above includes: S31-2: When the user's churn risk level is the second churn risk level, the types of content the user is interested in are determined based on the user's historical usage behavior data and basic data; the second churn risk level indicates that the user's usage frequency has decreased and the user's interest in platform activities has decreased. S32-2: Based on the content types that the user is interested in and the user's payment ability level, generate corresponding membership service discount information and push it to the user.

[0039] In one embodiment, basic user data is also collected. When a user's churn risk level is the second churn risk level (i.e., high churn risk), the types of content the user is interested in are determined based on the user's historical usage behavior data and basic data. The basic data reflects the user's own characteristics within the platform, including registration time, age, gender, etc. Historical behavior data refers to data generated by the user's login, check-in, browsing, and interactions (such as likes, comments, and shares) within the platform. Specifically, in this embodiment, the types of content the user may be interested in are determined based on pages accessed more frequently than a second threshold in the user's historical usage behavior data; or, the types of content the user may be interested in are determined based on the user's gender and age group in their basic data. For example, based on user A's age group (e.g., young adult) and gender (e.g., male), it is inferred that user A may be interested in live sports broadcasts.

[0040] Based on the analysis of user interest in specific content types, the system generates and pushes membership service offers tailored to each user's spending power and interests. For example, users with a high spending power level (second tier) and an interest in movies receive a discount on an annual premium movie membership; users with a low spending power level (first tier) and an interest in movies receive discounts on limited-time packages for popular movies and monthly discounts on regular movie memberships. By targeting high-risk users with precise marketing and personalized offers, the system aims to improve user satisfaction and membership retention.

[0041] In practical applications, corresponding discount information can be generated according to specific needs, such as discounts for feature upgrades, discounts for bundled packages, discounts for content on demand, discounts for membership subscriptions, etc. This application does not impose any restrictions on this.

[0042] Optionally, in one embodiment, in order to increase the detection frequency of usage behavior data for users with a high risk of churn, an assessment of the user's risk of churn is performed immediately when the user makes a target behavior (e.g., online complaint, service feedback, etc.) so as to retain platform members in a timely manner and improve user experience and member retention rate.

[0043] As one implementation of this application, the historical payment behavior data includes: the payment time, payment amount, and payment channel for users purchasing goods and subscribing to membership services on the platform; the platform account fund change data includes: the user's platform account balance, the time of fund inflows and outflows from the platform account, and the amount of fund changes; the step S2 above, "obtaining the user's payment ability level based on the historical payment behavior data and the platform account fund change data," specifically includes: S21-1: Based on the payment time, payment amount, and payment channel within the first time period, determine whether the user's payment behavior is stable, and determine whether the user's average payment amount reaches the amount threshold; S22-1: Determine the user's payment ability level based on whether the user's payment behavior is stable and whether the user's average payment amount reaches the amount threshold.

[0044] In one embodiment, the user's payment ability level is determined by analyzing the user's historical payment behavior and platform account fund change data within a first time period. Optionally, the length of the first time period is not less than 6 months. Specifically, the monthly payment frequency is calculated based on the payment time of user orders within the first time period, and the monthly payment frequency is compared to determine whether the monthly payment frequency is decreasing month by month. The number of payment channels in user orders is counted, and the number of payment channels is compared with a first threshold (e.g., the first threshold is 3) to obtain the comparison result. Based on the comparison result of the monthly payment frequency and the comparison result of the number of payment channels, it is determined whether the user's payment behavior is stable. Specifically, if at least one of the following conditions is met, the user's payment behavior is determined to be unstable: (1) The trend of monthly payment frequency is downward; (2) The number of payment channels is greater than the first threshold.

[0045] For example, if a user has 15 orders in the past 3 months, 12 of them are concentrated in the first month, while the user has 2 and 1 orders in the second and third months, respectively, calculations show that the user's monthly payment frequency is decreasing. Furthermore, all of the user's orders in the past 3 months include 4 payment channels, exceeding the first threshold of 3. Based on this, it can be determined that the user's payment behavior is unstable.

[0046] In this embodiment, the average payment amount within the first time period is calculated based on the order's payment amount, and this average payment amount is compared with a threshold. Combining the analysis results regarding the stability of user payment behavior and the comparison results between the average payment amount and the threshold, the user's payment ability level is determined. If the analysis determines that the user's payment behavior is unstable, or the average payment amount is less than the threshold, the user's payment ability is judged to be low, and the corresponding payment ability level is output as the first payment ability level. If the analysis determines that the user's payment behavior is stable, and the average payment amount reaches the threshold, the user's payment ability is judged to be high, and the corresponding payment ability level is output as the second payment ability level.

[0047] As one implementation of this application, the historical usage behavior data includes: the user's login frequency, usage duration, browsing content, and interaction behavior on the platform; step S2 above, "obtaining the user's churn risk level based on the historical usage behavior data," specifically includes: S21-2: Based on the login frequency, usage duration, and browsing content within the second and third time periods, determine whether the user's usage frequency has decreased; the second and third time periods do not overlap, and the time between the end of the second time period and the current time is less than the time between the end of the third time period and the current time. S22-2: Based on the interactive behavior during the second time period and the interactive behavior during the third time period, determine whether the user's interest in platform activities has decreased; S23-2: Determine the user's churn risk level based on whether the user's usage frequency has decreased and whether the user's interest in platform activities has decreased.

[0048] In one embodiment, the risk of user churn is assessed based on the user's historical behavior data within a second and a third time period. The second and third time periods are non-overlapping, and the time between the end of the second time period and the current time is less than the time between the end of the third time period and the current time. For example, the second time period is the most recent 3 months, and the third time period is the most recent 3-6 months.

[0049] In this embodiment, based on historical usage behavior data within the second and third time periods, it is determined whether the user's usage frequency has decreased and whether their interest in platform activities has decreased, thereby assessing the risk level of user churn. Specifically, the statistical results are obtained by statistically analyzing the user's average monthly login days (i.e., login frequency), average monthly usage time, the types and quantities of content viewed, and the number of interactions within the platform during the second and third time periods. User interactions within the platform include: check-in, comments, sharing, and participation in activities.

[0050] Compare the statistical results from the second and third time periods. If the comparison results meet at least two of the following conditions, it is determined that the user's usage frequency has decreased: (1) The average number of login days per month in the second time period was lower than the average number of login days per month in the third time period; (2) The average monthly usage time in the second time period was lower than the average monthly usage time in the third time period; (3) The number of types of content viewed by users in the second time period is less than the number of types of content viewed by users in the third time period.

[0051] Comparing the statistical results of the second and third time periods, if the number of interactions in the second time period is less than the number of interactions in the third time period, it is determined that the user's interest in the platform activities has decreased.

[0052] Analysis revealed that if a user's usage frequency or interest level did not decrease, the user's churn risk was low, thus classifying the user's churn risk as Level 1. Conversely, if both the user's usage frequency and interest level decreased, the user's churn risk was high, thus classifying the user's churn risk as Level 2.

[0053] As one implementation of this application, step S2 above, "obtaining the user's payment ability level based on the historical payment behavior data and the platform account's fund change data," specifically includes: S210-3: Input the user's historical payment behavior data and platform account fund change data into the first prediction model to obtain the user's payment ability level output by the first prediction model; the first prediction model is trained based on a logistic regression model. Step S2 above, "obtaining the user's churn risk level based on the historical usage behavior data," specifically includes: S210-4: Input the user's historical usage behavior data into the second prediction model to obtain the user's churn risk level output by the second prediction model; the second prediction model is trained based on the random forest model.

[0054] In this embodiment, in addition to assessing users' payment ability and membership churn risk through data statistical analysis, machine learning algorithms such as logistic regression, decision trees, and random forests can be used to build predictive models to achieve a more accurate assessment of users' payment ability and churn risk.

[0055] In one embodiment, a first predictive model for detecting a user's payment ability is constructed based on a logistic regression model using machine learning algorithms, and a second predictive model for detecting a user's membership churn risk is constructed based on a random forest model. The first predictive model analyzes the user's historical payment behavior data and platform account fund change data to assess the user's payment ability level, while the second predictive model analyzes the user's historical usage behavior data to assess the user's churn risk level.

[0056] As one embodiment of this application, before step S210-3, the method further includes: S211-3: Select sample users from all users and add a payment ability level label to each sample user; S212-3: Construct a first sample set based on the payment ability level tags, historical payment behavior data, and platform account fund change data of the sample users; S213-3: The first prediction model is obtained by training a logistic regression model using the first sample set.

[0057] In one embodiment, the steps for training the first prediction model are as follows: (1) First, sample users are drawn from all member users, and a label representing the payment capability level is added to each sample user. In the subsequent training process, the label added to the sample user is associated with the user's data (including historical payment behavior data and platform account fund change data). In practical applications, the corresponding label can be set according to the number of payment capability levels. For example, if there are two payment capability levels, the label can be set to 0 to represent the first payment capability level and 1 to represent the second payment capability level. (2) Then, a first sample set is constructed based on the historical payment behavior data of the sample users and the fund change data of the platform account; (3) Based on the logistic regression algorithm, a logistic regression model is constructed to predict the user's payment ability level. The first sample set is divided into two groups of data, a training set and a test set, according to a preset ratio, and the distribution of user features in the two datasets is kept consistent. Optionally, the ratio of the training set to the test set is 7:3. The logistic regression model is trained through the first sample set. During the training process, the output threshold of the model is adjusted according to the actual business needs, thereby improving the accuracy of the model in detecting payment ability and reducing misjudgments. For example, the model output probability is in [0,1]. The default classification threshold for payment ability is 0.5. If the output result is greater than or equal to 0.5, the user is judged to have strong payment ability. However, in actual application, the output threshold is increased to 0.6 according to the actual situation (that is, if the output result is ≥0.6, it is judged to have strong payment ability) to obtain more accurate prediction results.

[0058] The feature weights are iteratively updated by minimizing the loss function (cross-entropy loss) using the gradient descent algorithm. For example, for the feature "average payment amount of users in the past year," if the coefficient is positive and the absolute value is large, it indicates that this feature is positively correlated with payment ability, meaning that the higher the amount, the greater the probability of strong payment ability. During training, after every 100 iterations, the metrics (accuracy and F1 score) of the training set are calculated. Training stops if there is no improvement in the metrics after three consecutive iterations, resulting in the first prediction model.

[0059] Optionally, during training, the model is trained on a first sample set based on cross-validation to improve its accuracy and generalization ability. Hierarchical K-fold cross-validation (e.g., K=5) is used for both training and validation on the training set to prevent overfitting. For example, with 5-fold cross-validation, the training set is divided into 5 subsets, and 4 subsets are used for training and 1 subset for validation in turn. The average validation accuracy is calculated, and the regularization parameter is adjusted. The model performance is evaluated on the test set, and iterative training is performed until the accuracy is ≥85% and the F1 score is ≥0.8, at which point the model is considered to have met the requirements.

[0060] Optionally, during the training of the first prediction model, the first sample set is dynamically updated using newly generated user payment behavior data and platform account fund change data. The updated first sample set is then used to iteratively train and optimize the first prediction model, enabling it to dynamically adapt to changes in user behavior (such as fluctuations in payment habits during holidays), ensuring that the model always maintains high prediction accuracy, and thus ensuring long-term stable provision of high-quality membership maintenance services to member users.

[0061] As one embodiment of this application, step S212-3 above includes: S2121-3: Select first feature data from the historical payment behavior data and platform account fund change data of the sample users to construct a first initial sample set; the first feature data includes: the user's average payment amount, payment frequency, platform account balance fluctuation coefficient, fund inflow stability, number of automatic deduction failures, membership level, login frequency, usage duration, historical renewal success rate and payment channel stability during the fourth time period. S2122-3: Based on the degree of influence of each first feature data on payment ability, set corresponding feature weights for each first feature data, wherein the first feature data with a greater degree of influence corresponds to a larger feature weight; and, calculate the ratio of the number of sample users of different payment ability levels, and assign corresponding sample weights to sample users of different payment ability levels based on the ratio of the number of samples, wherein the sample weights of sample users of payment ability levels with a larger number of samples are smaller. S2123-3: Based on the sample weights and the feature weights of each first feature data, update the first initial sample set to obtain the first sample set.

[0062] In the above embodiments, the steps for constructing the first sample set are as follows: (1) From the historical payment behavior data and platform account fund change data of sample users in the fourth time period (e.g., within 1 year), the following data are selected as the first feature data: average payment amount, payment frequency, platform account balance fluctuation coefficient, fund inflow stability, number of automatic deduction failures, membership level, login frequency, usage duration, historical renewal success rate, and number of payment channels, etc. Based on all the first feature data, a first initial sample set is constructed; (2) Based on the pre-set influence coefficients of each first feature data on payment ability, assign corresponding feature weights to each first feature data. The greater the influence of the first feature data, the greater its feature weight. Specifically, the feature weights of each first feature data are calculated using the following expression: ; in, For the first feature data Feature weights; For the first feature data The degree of influence coefficient; It is the total number of all first feature data.

[0063] Because the initial sample set may have an imbalance between positive and negative samples—for example, the number of users with "strong payment ability" is much greater than the number of users with "weak payment ability"—to avoid bias in model training towards the majority class, sample weights are assigned to different payment ability levels based on the ratio of the number of users with different payment ability levels. The more users in a payment ability level have, the smaller the sample weight corresponding to that level, thus mitigating the imbalance between positive and negative samples in the sample set. Specifically, when there are two payment ability levels, the sample weights are set as follows: ; ; in, The sample weights are the samples corresponding to the first payment capability level. The sample weights corresponding to the samples of the second payment capability level; The number of samples representing the first level of payment capability; The number of samples for the second level of payment capability; (3) Update the first initial sample set based on the sample weights and the feature weights of each first feature data, add corresponding weight information to each sample data in the first initial sample set, and obtain the first sample set used to train the first prediction model.

[0064] Then, the updated first sample set was used to train a fine-tuned logistic regression model to improve the accuracy of the model's ability to predict payments.

[0065] As one embodiment of this application, before steps S210-4 above, the method further includes: S211-4: Extract sample users from all users and add a churn risk level label to each sample user; S212-4: Construct a second sample set based on the churn risk level labels and historical usage behavior data of the sample users; S213-4: Train the random forest model using the second sample set to obtain the second prediction model.

[0066] In one embodiment, the steps for training the second prediction model are as follows: (1) Sample users are drawn from all member users, and a label representing the churn risk level is added to each sample user. In the subsequent training process, the label added to the sample user is associated with the user's historical usage behavior data. In practical applications, the corresponding label can be set according to the number of churn risk levels. For example, if there are two churn risk levels, the label can be set to 0 to represent the first churn risk level and 1 to represent the second churn risk level. (2) Based on the churn risk level labels of the sample users and the historical usage behavior data of the users, a second sample set is constructed; (3) A random forest model is constructed based on the random forest algorithm to predict the churn risk level of users. The model consists of multiple decision trees, and the output results are combined through the Bagging strategy. Specifically, the second sample set is divided into a training set and a test set according to a preset ratio to ensure that the user feature distribution of the two datasets is consistent. Optionally, the ratio of the training set to the test set is 7:3. During the training process, two levels are set for the user churn risk level, including the first churn risk level, which represents a lower churn risk, and the second churn risk level, which represents a higher churn risk. Hierarchical K-fold cross-validation (e.g., K=5) is used to train and validate the model on the training set to prevent overfitting. The training error and validation error of each tree are recorded. When the validation error no longer decreases for 10 consecutive rounds, the training is stopped, and the second prediction model is obtained. During the training process, key indicators, including accuracy, recall, and F1 score, are calculated on the test set to analyze the model's performance under different probability thresholds. Through actual verification, it is determined that the accuracy of the model can reach 78.3% when the threshold for churn risk judgment is 0.6, and the false positive rate is controlled within 15%.

[0067] Optionally, during the training of the second prediction model, the second sample set is dynamically updated using newly generated user behavior data. Furthermore, based on the dynamically updated second sample set, the second prediction model is continuously iteratively trained and optimized to adapt to market changes and the dynamic evolution of user behavior, ensuring that the model maintains high prediction accuracy and thus ensuring long-term, stable provision of high-quality membership maintenance services to members.

[0068] As one embodiment of this application, step S212-4 above includes: S2121-4: Select second feature data from the historical usage behavior data of the sample users to construct a second initial sample set; the second feature data includes: the number of login days, average daily usage time, page browsing depth, number of customer service inquiries, complaint rate, number of check-in days, and number of shares during the fifth time period. S2122-4: Based on the degree of influence of each second feature data on the risk of churn, set corresponding feature weights for each second feature data, where the second feature data with a greater degree of influence has a larger feature weight. S2123-4: Update the second initial sample set based on the feature weights to obtain the second sample set.

[0069] In the above embodiments, the steps for constructing the second sample set are as follows: (1) Select the following data as the second feature data from the historical usage behavior data of the sample users within the fifth time period (e.g., within 1 year): number of login days, average daily usage time, page browsing depth, number of customer service inquiries, complaint rate, number of check-in days, and number of shares. Based on all the second feature data, construct the second initial sample set; (2) Considering that different dimensions of second feature data have different degrees of impact on user churn risk, for example, in the process of predicting churn risk on a certain e-commerce platform, it was found that the feature "login frequency in the past month" has a greater impact on user churn risk, so a larger feature weight can be set, while the feature "number of product categories viewed" has a smaller impact on churn risk, so a smaller feature weight can be set. Based on this, in this embodiment, according to the pre-set impact coefficient of each second feature data on churn risk, a corresponding feature weight is set for each second feature data. For the second feature data with a greater degree of impact, its feature weight is greater. Specifically, the feature weight of each second feature data is calculated by the following expression: ; in, For the second feature data Feature weights; For the second feature data The degree of influence coefficient; The number of second feature data; (3) Based on the feature weights of each second feature data, the second initial sample set is updated, and corresponding weight information is added to each sample data in the sample set to obtain the second sample set used to train the second prediction model.

[0070] Furthermore, the updated second sample set was used to train the calibrated random forest model to improve the accuracy of the model in predicting churn risk.

[0071] Optionally, in one embodiment, social behavior data from outside the platform is also acquired to expand the second sample set, further improving the model's generalization ability and reliability through user data from different source platforms. For social platforms, publicly available information from external social platforms (such as user interaction data on social platforms) can be accessed through user authorization operations within the platform, or user behavior data related to membership services in social scenarios can be obtained through cooperation with social platforms (such as users sharing in-site video content to social platforms).

[0072] Optionally, surveys or user feedback forms can be randomly distributed within the platform at fixed intervals to obtain data such as the types of content users are interested in, newly launched activities, promotional information, and service satisfaction, thereby expanding the second sample set and improving the model's ability to quickly adapt to changes in users' dynamic behavior.

[0073] Optionally, graph neural networks can be used to model the influence relationships between user groups and capture the interconnected interactions between individuals within a user group. Specifically, users are treated as nodes, and interactions between users (such as sharing videos or watching shows together) are considered as edges, constructing a user relationship graph. For example, if user A invited user B to subscribe to a membership, or if user A and user B frequently interact within the platform (such as sharing playlists or participating in activities together), then an edge exists between them. For two users with a relationship edge, when the churn risk of one user increases, the churn risk of the other user may also increase, or a user's higher purchasing power may affect the renewal intention of another user (for example, user A recommending paid content prompts user B to renew their membership). Based on this, user relationship data from the graph neural network is added to the first and second sample sets, providing group relationship characteristics for the first and second prediction models. This enables the prediction models to learn the impact of user group relationships on purchasing power and churn risk within the platform, expanding from individual analysis of single users to relational analysis of user groups, further improving the accuracy of the evaluation results. For example, after incorporating group relationship features, the second prediction model's assessment of the churn risk for a user with low usage frequency but in a highly active user group changed from high churn risk to low churn risk. This is because group interaction drives the user to continue using the platform services.

[0074] In one embodiment, promotional information is pushed to multiple users with group relationships in a graph neural network. This makes the marketing strategy for member retention more aligned with users' social attributes and group behavior patterns, further improving user experience and member retention rates. For example, group-buying discounts and friend-assisted discounts are pushed to users with group relationships. As one embodiment of this application, the above steps S213-4 include: S2131-4: Determine the target parameter combination through grid search, construct a random forest model based on the target parameter combination, and calibrate the random forest model through Platt scaling; S2132-4: Using the second sample set, train the calibrated random forest model to obtain the second prediction model.

[0075] In one embodiment, for a scenario of predicting user churn risk, the random forest model is fine-tuned before training to obtain a calibrated random forest model, specifically including: (1) Optimize the depth and number of model trees. The optimal parameter combination is determined by grid search. Finally, n_estimators=100 (number of decision trees), max_depth=8 (maximum depth of the tree), and min_samples_split=10 (minimum number of samples required to split internal nodes) are set to balance model complexity and generalization ability. (2) Calibrate the churn probability by scaling the original score (between 0 and 1) of the model output with Platt to make it closer to the actual churn probability. For example, if a user's original output score is 0.65, it becomes 0.72 after calibration, which can more accurately reflect the likelihood of churn.

[0076] Then, the calibrated random forest model is trained using the second sample set to obtain the second prediction model, so as to further improve the accuracy of the model in predicting the churn risk level.

[0077] Figure 2 This is a flowchart illustrating the automatic renewal process for a membership platform account in one embodiment of this application. For example... Figure 2 As shown, in one embodiment, multi-dimensional data is first collected from the user, including the user's historical usage behavior data, historical payment behavior data, and platform account fund change data. The collected data is then cleaned and standardized, and relevant features are extracted. For example, the user's average payment amount is determined based on the payment amount of all orders placed within the past year, and the user's average payment interval is determined based on the number of orders placed.

[0078] Then, the processed data is input into the prediction model. Specifically, data related to the user's payment behavior and platform account funds are input into the first prediction model to obtain the payment capability level output by the first prediction model; and data related to the user's usage behavior are input into the second prediction model to obtain the churn risk level output by the second prediction model.

[0079] Furthermore, based on the churn risk level and payment ability level, a membership maintenance strategy is generated for the user, and the subscription fee is deducted from the user's platform account according to the strategy. If the deduction is successful, the new payment behavior data is recorded, and the user's membership validity period is updated. If the deduction fails, a retry message is generated and sent to the user via push notification or SMS.

[0080] The system will retry deducting fees from the user's platform account. If the retry is successful, the system will record the new payment behavior data and update the user's membership validity period. If the retry fails, the system will generate a second notification message indicating the failure and notify the user of the reason for the failure via push notification or SMS.

[0081] Based on the same inventive concept, one embodiment of this application provides a membership strategy processing system, the system comprising: The data acquisition module 101 is configured to collect target information of users, including: historical payment behavior data, historical usage behavior data, and fund change data of the platform account; the historical payment behavior data is the data generated by users purchasing goods and subscribing to membership services on the platform. The analysis module 102 is configured to obtain the user's payment ability level based on the historical payment behavior data and the platform account's fund change data; and to obtain the user's churn risk level based on the historical usage behavior data. The strategy generation module 103 is configured to generate a membership maintenance strategy for the user based on the user's payment ability level and churn risk level. The deduction execution module 104 is configured to perform a deduction operation for the user's platform account for subscription membership services based on the membership maintenance strategy.

[0082] Figure 3 This is a flowchart illustrating the workflow of each module in a membership policy processing system according to an embodiment of this application. Figure 3As shown, the data acquisition module 101 collects users' historical data (including historical payment behavior data and historical usage behavior data). After preprocessing, this historical data is transmitted to the analysis module 102 for prediction of payment ability level and churn risk level. The strategy generation module 103 generates a membership maintenance strategy based on the prediction results output by the analysis module 102. The billing execution module 104 executes the billing operation for the user's platform account to subscribe to membership services according to the membership maintenance strategy, and feeds back the billing results (e.g., newly generated payment behavior data) to the data acquisition module for subsequent analysis and strategy adjustment.

[0083] As one embodiment of this application, the strategy generation module 103 is specifically configured to perform the following steps: When the user's payment ability level is at the first payment ability level and the churn risk level is at the first churn risk level, a target time period for when the user has sufficient funds is determined based on the user's historical payment behavior data and the platform account's fund change data; the first payment ability level indicates that the user's payment behavior is unstable, or the user's average payment amount is less than the amount threshold; the first churn risk level indicates that the user's usage frequency has not decreased, or the user's interest in platform activities has not decreased. Determine a target date from the target time period and postpone the user's billing date to the target date; Based on the membership maintenance strategy, the user's platform account is charged for the subscription membership service, specifically including: charging the user's platform account for the subscription membership service according to the target date.

[0084] As one embodiment of this application, after deducting the subscription fee for the user's platform account according to the target date, the policy generation module 103 is further configured to perform the following operations: Upon successful payment deduction, new payment behavior data is generated to update the user's historical payment behavior data; In the event of a failed deduction, a retry deduction cycle is determined for the user; the user's platform account is then retried and deducted according to the retry deduction cycle.

[0085] As one embodiment of this application, the historical usage behavior data is data generated by users logging in, checking in, browsing, and interacting within the platform; the target information also includes basic data, which includes: registration time, age, and gender; the strategy generation module 103 is further configured to perform the following operations: If the user's churn risk level is the second churn risk level, the types of content the user is interested in are determined based on the user's historical usage behavior data and basic data; the second churn risk level indicates that the user's usage frequency has decreased and the user's interest in platform activities has decreased. Based on the types of content the user is interested in and the user's payment ability level, corresponding membership service discount information is generated and pushed to the user.

[0086] As one implementation of this application, the historical payment behavior data includes: the payment time, payment amount, and payment channel for users purchasing goods and subscribing to membership services on the platform; the platform account fund change data includes: the user's platform account balance, the time of fund inflows and outflows from the platform account, and the amount of fund changes; the analysis module 102 is configured to obtain the user's payment ability level based on the historical payment behavior data and the platform account fund change data, specifically including: Based on the payment time, payment amount, and payment channel within the first time period, determine whether the user's payment behavior is stable, and determine whether the user's average payment amount reaches the amount threshold. The user's payment ability level is determined based on whether the user's payment behavior is stable and whether the user's average payment amount reaches a certain threshold.

[0087] As one embodiment of this application, the historical usage behavior data includes: the user's login frequency, usage duration, browsing content, and interaction behavior on the platform; the analysis module 102 is configured to obtain the user's churn risk level based on the historical usage behavior data, specifically including: Based on the login frequency, usage duration, and browsing content within the second and third time periods, it is determined whether the user's usage frequency has decreased; the second and third time periods do not overlap, and the time between the end of the second time period and the current time is less than the time between the end of the third time period and the current time. Based on the interactive behavior during the second time period and the interactive behavior during the third time period, it is determined whether the user's interest in platform activities has decreased; The user's churn risk level is determined based on whether the user's usage frequency decreases and whether the user's interest in platform activities decreases.

[0088] In one embodiment of this application, the analysis module 102 is further configured to perform the following steps: The user's historical payment behavior data and platform account fund change data are input into the first prediction model to obtain the user's payment ability level output by the first prediction model; the first prediction model is trained based on a logistic regression model. The user's historical usage behavior data is input into the second prediction model to obtain the user's churn risk level output by the second prediction model; the second prediction model is trained based on the random forest model.

[0089] In one embodiment of this application, the analysis module 102 further includes a first training module, configured to perform the following steps: Sample users were selected from all users, and payment ability level labels were added to each sample user; Based on the payment ability level tags, historical payment behavior data, and platform account fund change data of the sample users, a first sample set is constructed; The first prediction model is obtained by training a logistic regression model using the first sample set.

[0090] In one embodiment of this application, the first training module is configured to construct a first sample set based on the sample users' payment ability level tags, historical payment behavior data, and platform account fund change data, including: The first initial sample set is constructed by filtering the historical payment behavior data and fund change data of the sample users. The first feature data includes: the average payment amount, payment frequency, platform account balance fluctuation coefficient, fund inflow stability, number of automatic deduction failures, membership level, login frequency, usage duration, historical renewal success rate and payment channel stability of the user in the fourth time period. Based on the degree of influence of each first feature data on payment ability, a corresponding feature weight is set for each first feature data, wherein the first feature data with a greater degree of influence corresponds to a larger feature weight; and, the ratio of the number of sample users of different payment ability levels is calculated, and based on the ratio, a corresponding sample weight is assigned to sample users of different payment ability levels, wherein the sample weight of sample users of payment ability levels with a larger sample number is smaller. Based on the sample weights and the feature weights of each first feature data, the first initial sample set is updated to obtain the first sample set.

[0091] In one embodiment of this application, the analysis module 102 further includes a second training module, configured to perform the following steps: Sample users are drawn from all users, and churn risk level labels are added to each sample user; Based on the churn risk level labels and historical usage behavior data of the sample users, a second sample set is constructed. The second prediction model is obtained by training a random forest model using the second sample set.

[0092] In one embodiment of this application, the second training module is configured to construct a second sample set based on the churn risk level labels and historical usage behavior data of the sample users, including: The second feature data is selected from the historical usage behavior data of the sample users to construct a second initial sample set; the second feature data includes: the number of login days, average daily usage time, page browsing depth, number of customer service inquiries, complaint rate, number of check-in days, and number of shares during the fifth time period. Based on the degree of impact of each secondary feature data on the risk of churn, set corresponding feature weights for each secondary feature data; The second initial sample set is updated based on the feature weights to obtain the second sample set.

[0093] In one embodiment of this application, the second training module is configured to train a random forest model using the second sample set to obtain the second prediction model, including: The target parameter combination is determined by grid search, a random forest model is constructed based on the target parameter combination, and the random forest model is calibrated by Platt scaling; Using the second sample set, the calibrated random forest model is trained to obtain the second prediction model.

[0094] Based on the same inventive concept, one embodiment of this application provides a readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the membership policy processing method as described in any of the above embodiments of this application.

[0095] Based on the same inventive concept, one embodiment of this application provides an electronic device. Figure 4 This is a schematic diagram of an electronic device according to an embodiment of this application. Figure 4 As shown, the electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When executed by the processor, the computer program implements the steps in the membership policy processing method as described in any of the above embodiments of this application.

[0096] Regarding the system in the above embodiments, the specific ways in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0097] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0098] For the sake of simplicity, the method embodiments are described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and components involved are not necessarily essential to this application.

[0099] Those skilled in the art will understand that embodiments of this application can be provided as methods, apparatus, or computer program products. Therefore, embodiments of this application can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of this application can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0100] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0101] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0102] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0103] Although preferred embodiments of the embodiments of this application have been described, those skilled in the art, once they understand the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, this application is to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of this application.

[0104] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0105] The membership policy processing method, apparatus, device, and storage medium provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A membership strategy processing method, characterized in that, include: Collect user target information, which includes: historical payment behavior data, historical usage behavior data, and platform account fund change data; the historical payment behavior data is the data generated by users purchasing goods and subscribing to membership services on the platform; Based on the historical payment behavior data and the platform account's fund change data, the user's payment ability level is obtained; and based on the historical usage behavior data, the user's churn risk level is obtained. Based on the user's payment ability level and churn risk level, generate a membership maintenance strategy for the user; Based on the aforementioned membership maintenance strategy, the platform account of the user will be charged for the subscription membership service.

2. The membership strategy processing method according to claim 1, characterized in that, Based on the user's payment ability level and churn risk level, a membership maintenance strategy is generated for the user, including: When the user's payment ability level is at the first payment ability level and the churn risk level is at the first churn risk level, a target time period for when the user has sufficient funds is determined based on the user's historical payment behavior data and the platform account's fund change data; the first payment ability level indicates that the user's payment behavior is unstable, or the user's average payment amount is less than the amount threshold; the first churn risk level indicates that the user's usage frequency has not decreased, or the user's interest in platform activities has not decreased. Determine a target date from the target time period and postpone the user's billing date to the target date; Based on the membership maintenance strategy, the user's platform account is charged for the subscription membership service, specifically including: charging the user's platform account for the subscription membership service according to the target date.

3. The membership strategy processing method according to claim 2, characterized in that, After deducting the subscription fee from the user's platform account according to the target date, the method further includes: Upon successful payment deduction, new payment behavior data is generated to update the user's historical payment behavior data; In the event of a failed deduction, a retry deduction cycle is determined for the user; the user's platform account is then retried and deducted according to the retry deduction cycle.

4. The membership strategy processing method according to claim 1, characterized in that, The historical usage behavior data refers to data generated by users logging in, checking in, browsing, and interacting within the platform; the target information also includes basic data, such as registration time, age, and gender; based on the user's payment ability level and churn risk level, a membership maintenance strategy is generated for the user, including: If the user's churn risk level is the second churn risk level, the types of content the user is interested in are determined based on the user's historical usage behavior data and basic data; the second churn risk level indicates that the user's usage frequency has decreased and the user's interest in platform activities has decreased. Based on the types of content the user is interested in and the user's payment ability level, corresponding membership service discount information is generated and pushed to the user.

5. The membership strategy processing method according to any one of claims 1-4, characterized in that, The historical payment behavior data includes: the payment time, payment amount, and payment channel for users purchasing goods and subscribing to membership services on the platform; the platform account fund change data includes: the user's platform account balance, the time of fund inflows and outflows from the platform account, and the amount of fund changes; based on the historical payment behavior data and the platform account fund change data, the user's payment ability level is obtained, specifically including: Based on the payment time, payment amount, and payment channel within the first time period, determine whether the user's payment behavior is stable, and determine whether the user's average payment amount reaches the amount threshold. The user's payment ability level is determined based on whether the user's payment behavior is stable and whether the user's average payment amount reaches a certain threshold.

6. The membership strategy processing method according to any one of claims 1-4, characterized in that, The historical usage behavior data includes: the user's login frequency, usage duration, browsing content, and interaction behavior on the platform; based on the historical usage behavior data, the user's churn risk level is obtained, specifically including: Based on the login frequency, usage duration, and browsing content within the second and third time periods, it is determined whether the user's usage frequency has decreased; the second and third time periods do not overlap, and the time between the end of the second time period and the current time is less than the time between the end of the third time period and the current time. Based on the interactive behavior during the second time period and the interactive behavior during the third time period, it is determined whether the user's interest in platform activities has decreased; The user's churn risk level is determined based on whether the user's usage frequency decreases and whether the user's interest in platform activities decreases.

7. The membership strategy processing method according to any one of claims 1-4, characterized in that, Based on the historical payment behavior data and the platform account fund change data, the user's payment ability level is obtained, specifically including: inputting the user's historical payment behavior data and the platform account fund change data into a first prediction model to obtain the user's payment ability level output by the first prediction model; the first prediction model is trained based on a logistic regression model. Based on the historical usage behavior data, the user's churn risk level is obtained, specifically by: inputting the user's historical usage behavior data into a second prediction model to obtain the user's churn risk level output by the second prediction model; the second prediction model is trained based on a random forest model.

8. The membership strategy processing method according to claim 7, characterized in that, Before inputting the user's historical payment behavior data and platform account fund change data into the first prediction model, the method further includes: Sample users were selected from all users, and payment ability level labels were added to each sample user; Based on the payment ability level tags, historical payment behavior data, and platform account fund change data of the sample users, a first sample set is constructed; The first prediction model is obtained by training a logistic regression model using the first sample set.

9. The membership strategy processing method according to claim 8, characterized in that, Based on the payment ability level tags, historical payment behavior data, and platform account fund change data of the sample users, a first sample set is constructed, including: The first initial sample set is constructed by filtering the historical payment behavior data and fund change data of the sample users. The first feature data includes: the average payment amount, payment frequency, platform account balance fluctuation coefficient, fund inflow stability, number of automatic deduction failures, membership level, login frequency, usage duration, historical renewal success rate and payment channel stability of the user in the fourth time period. Based on the degree of influence of each first feature data on payment ability, a corresponding feature weight is set for each first feature data, wherein the first feature data with a greater degree of influence corresponds to a larger feature weight; and, the ratio of the number of sample users of different payment ability levels is calculated, and based on the ratio, a corresponding sample weight is assigned to sample users of different payment ability levels, wherein the sample weight of sample users of payment ability levels with a larger sample number is smaller. Based on the sample weights and the feature weights of each first feature data, the first initial sample set is updated to obtain the first sample set.

10. The membership strategy processing method according to claim 7, characterized in that, Before inputting the user's historical usage behavior data into the second prediction model, the method further includes: Sample users are drawn from all users, and churn risk level labels are added to each sample user; Based on the churn risk level labels and historical usage behavior data of the sample users, a second sample set is constructed. The second prediction model is obtained by training a random forest model using the second sample set.

11. The membership strategy processing method according to claim 10, characterized in that, Based on the churn risk level labels and historical usage behavior data of the sample users, a second sample set is constructed, including: The second feature data is selected from the historical usage behavior data of the sample users to construct a second initial sample set; the second feature data includes: the number of login days, average daily usage time, page browsing depth, number of customer service inquiries, complaint rate, number of check-in days, and number of shares during the fifth time period. Based on the degree of impact of each secondary feature data on the risk of churn, set corresponding feature weights for each secondary feature data; The second initial sample set is updated based on the feature weights to obtain the second sample set.

12. The membership strategy processing method according to claim 11, characterized in that, The second prediction model is obtained by training a random forest model using the second sample set, including: The target parameter combination is determined by grid search, a random forest model is constructed based on the target parameter combination, and the random forest model is calibrated by Platt scaling; Using the second sample set, the calibrated random forest model is trained to obtain the second prediction model.

13. A membership strategy processing system, characterized in that, For implementing the method as described in any one of claims 1-12, comprising: The data collection module is configured to collect target information of users, including: historical payment behavior data, historical usage behavior data, and fund change data of the platform account; the historical payment behavior data is the data generated by users purchasing goods and subscribing to membership services within the platform. The analysis module is configured to obtain the user's payment ability level based on the historical payment behavior data and the platform account's fund change data; and to obtain the user's churn risk level based on the historical usage behavior data. The strategy generation module is configured to generate a membership maintenance strategy for the user based on the user's payment ability level and churn risk level. The deduction execution module is configured to perform a deduction operation for the user's platform account for subscription services based on the membership maintenance strategy.

14. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps in the method as described in any one of claims 1-12.

15. An electronic 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 in the method as described in any one of claims 1-12.

Images