A fund product determination method, electronic equipment and storage medium
By combining scoring models and KYC rules, the system accurately identifies potential fund users and sends matching products, solving the problem of finding customers accurately in the public fund market, improving prediction accuracy and reducing marketing costs.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- PING AN BANK CO LTD
- Filing Date
- 2023-03-27
- Publication Date
- 2026-06-23
AI Technical Summary
In the mutual fund market, it is becoming increasingly difficult to accurately identify fund customers, leading to increased marketing costs and the potential negative impact of malicious marketing.
The method combines a scoring model and KYC rules. The scoring model predicts user potential, and the KYC rules are used to screen potential fund users. Matching fund products are then sent to users based on their type.
It improved the accuracy of predicting potential fund users, reduced unnecessary marketing costs, and avoided the negative impact of malicious marketing.
Smart Images
Figure CN116485550B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of financial technology, and in particular to a method for determining fund products, electronic devices, and storage media. Background Technology
[0002] Currently, with the growth of the public fund market and the massive increase in data from various professional companies, accurately identifying fund customers is becoming increasingly difficult. However, accurately identifying fund customer groups can reduce marketing costs while minimizing unnecessary customer complaints and avoiding the negative impact of malicious marketing on businesses. Summary of the Invention
[0003] The main technical problem addressed by this application is to provide a method, electronic device, and storage medium for determining fund products, which can improve the accuracy of predictions for potential users.
[0004] To address the aforementioned technical problems, this application provides a method for determining fund products. This method includes: acquiring a user cluster; using a scoring model to predict the performance of each user in the user cluster to obtain a first potential user cluster, the first potential user cluster including users predicted as first potential users by the scoring model; and using KYC rules to screen each user in the user cluster to obtain a second potential user cluster, the second potential user cluster including users screened as second potential users by KYC rules; combining the first and second potential user clusters to obtain potential fund users within the user cluster; determining the user type of each potential fund user; and sending fund products matching the corresponding user type to each potential fund user.
[0005] The process involves using a scoring model to predict the performance of each user in a user cluster, resulting in a first potential user cluster. This includes: predicting the target score for each user in each user cluster using the scoring model; identifying a user as a first potential user if the target score is greater than or equal to a scoring threshold; where the scoring threshold is the smallest model score in historical statistics where the number of conversions exceeds a certain threshold, and the number of conversions is the number of users who purchased funds within a preset time period after the prediction, based on the predicted number of users corresponding to the model score; and extracting each first potential user from the user cluster to obtain the first potential user cluster.
[0006] The scoring model includes at least two components. The scoring model is used to predict the target score of the user. This includes: using at least two scoring models to predict the user based on the user data corresponding to the user, and obtaining at least two initial scores for the user; wherein each initial score corresponds to one scoring model, and the user data includes at least one of the following: basic data, behavioral data, financial data, and risk data; and obtaining the target score for the user based on the at least two initial scores for the user.
[0007] The scoring model includes two components: one is a random forest, and the other is XGBC. The basic data includes at least one of age, education level, marital status, occupation, and annual income; the behavioral data includes at least one of logging into the client, clicking on a finance-related page, clicking on a finance product, and purchasing a finance product; the financial data includes at least one of asset information, investment information, policy information, claims information, premium information, loan information, and repayment information; and the risk data includes at least one of risk level and risk tolerance level. The target score for the corresponding user is obtained based on at least two initial scores, including weighted summation of the initial scores.
[0008] Before using a scoring model to predict the target score of a user based on the user's corresponding user data, the method for determining the fund product also includes: data cleaning of the user's corresponding user data; whereby data cleaning includes at least one of the following: field classification, saturation filtering, field type conversion, filling missing values, and chi-square filtering.
[0009] The KYC rules include a KYC tag pool, which contains several KYC tags. The KYC rules are used to filter users in the user cluster to obtain a second potential user cluster. This includes: selecting users from the user cluster who match any KYC tag as second potential users; where KYC tags include non-financial management users, non-fund holding users, non-bank valid users, and users holding at least one of the following preset number of funds: The second potential user cluster is then formed by combining the second potential users.
[0010] The method includes: a scoring model predicts a target score for a user and determines whether the user is a first potential user based on whether the target score is greater than or equal to a scoring threshold; after sending fund products matching the user type to each potential fund user, the method for determining the fund product further includes: obtaining update parameters corresponding to the scoring model; wherein the update parameters include the current stability index of the scoring model, scoring data, and conversion number; wherein the scoring data is the target score predicted by the scoring model for each user, and the conversion number is the number of users who purchased the fund product within a preset time period after receiving the sent fund product; in response to any update parameter meeting the update requirements, sending a first update message that requires updating the scoring model; and / or, KYC rules include a KYC tag pool, the KYC tag pool includes several KYC tags, and KYC rules are predicted based on several tags; after sending fund products matching the user type to each potential fund user, the method for determining the fund product further includes: determining whether there are any missing tags in the KYC tag pool; in response to any missing tag in the KYC tag pool, sending a second update message that requires updating the tag pool.
[0011] Specifically, by combining the first potential user cluster and the second potential user cluster, potential fund users in the user cluster are obtained, including: obtaining an overlapping user cluster, a first non-overlapping user cluster, and a second non-overlapping user cluster based on the first and second potential user clusters, and using users in the overlapping user cluster, the first non-overlapping user cluster, and the second non-overlapping user cluster as potential fund users; wherein each user in the overlapping user cluster exists in both the first and second potential user clusters, each user in the first non-overlapping user cluster exists in the first potential user cluster, and each user in the second non-overlapping user cluster exists in the second potential user cluster; and / or, the fund is a public fund.
[0012] To solve the above-mentioned technical problems, another technical solution adopted in this application is to provide an electronic device, which includes a processor and a memory, wherein the memory stores program instructions, and the processor executes the program instructions to implement the above-mentioned method for determining fund products.
[0013] To solve the above-mentioned technical problems, another technical solution adopted in this application is to provide a computer-readable storage medium for storing program instructions that can be executed to implement the above-mentioned method for determining fund products.
[0014] The above technical solution uses a scoring model to obtain the first potential user cluster and KYC rules to obtain the second potential user cluster. Combining the first and second potential user clusters yields the fund's potential users within the user cluster. Therefore, by combining the algorithm model and KYC rules to predict fund potential users within the user cluster, compared to methods that rely solely on either the algorithm model or KYC rules, the inclusion of KYC rules supplements the prediction of users not considered by the algorithm model, thus improving the accuracy of fund potential user prediction.
[0015] In addition, the user type of each potential user will be determined so that fund products matching their type can be sent to them. Attached Figure Description
[0016] Figure 1 This is a flowchart illustrating an embodiment of the method for determining fund products provided in this application;
[0017] Figure 2 yes Figure 1 The flowchart of step S12 shown is a schematic diagram of one embodiment;
[0018] Figure 3 yes Figure 2 The flowchart of step S21 shown is a schematic diagram of one embodiment;
[0019] Figure 4 yes Figure 1 A flowchart illustrating another embodiment of step S12 is shown.
[0020] Figure 5 This is a schematic diagram of the structure of an embodiment of the electronic device provided in this application;
[0021] Figure 6 This is a schematic diagram of an embodiment of the computer-readable storage medium provided in this application. Detailed Implementation
[0022] The embodiments of this application will now be described in detail with reference to the accompanying drawings.
[0023] In the following description, specific details such as particular system architectures, interfaces, and technologies are presented for illustrative purposes rather than for limiting purposes, in order to provide a thorough understanding of this application.
[0024] In this document, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " generally indicates that the preceding and following related objects have an "or" relationship. Furthermore, "many" in this document means two or more. Moreover, the term "at least one" in this document means any combination of at least two of any one or more of a plurality of objects. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C.
[0025] Please see Figure 1 , Figure 1 This is a flowchart illustrating one embodiment of the method for determining fund products provided in this application. It should be noted that if substantially the same result is obtained, this embodiment is not necessarily identical. Figure 1 The illustrated process sequence is limited. For example... Figure 1 As shown, this embodiment includes:
[0026] Step S11: Obtain the user cluster.
[0027] The method in this embodiment is used to identify potential users who wish to purchase funds and to send them fund products that match their type. The funds mentioned herein include, but are not limited to, public funds and private funds, etc., and are not specifically limited to these categories.
[0028] In this embodiment, a user cluster is obtained to facilitate the subsequent identification of potential fund buyers from within the user cluster. In one embodiment, the user cluster can be obtained from local storage or cloud storage. Of course, in other embodiments, user data can also be directly extracted from the relevant financial platform to obtain the user cluster.
[0029] It should be noted that users in a user cluster can belong to the same financial platform or different financial platforms. These financial platforms can be individual financial platforms, tagging platforms, AlCloud, Z... + For example, if Ping An Bank needs to determine whether customers under the Ping An Group are potential customers for purchasing funds, it will obtain information from relevant financial platforms under the Ping An Group (such as personal banking platforms, tagging platforms, AlCloud, Z). + Customer data (such as customer data) is used as a user cluster.
[0030] Step S12: Use the scoring model to predict each user in the user cluster to obtain the first potential user cluster; and use KYC rules to filter each user in the user cluster to obtain the second potential user cluster.
[0031] In this embodiment, a scoring model is used to predict the behavior of each user in the user cluster, resulting in a first potential user cluster, which includes users predicted as first potential users by the scoring model. Furthermore, KYC rules are used to filter the users in the user cluster, resulting in a second potential user cluster, which includes users filtered as second potential users by the KYC rules. In other words, potential user prediction is performed using both the scoring model and KYC rules to obtain the first potential users corresponding to the scoring model and the second potential users corresponding to the KYC rules within the user cluster.
[0032] In one implementation, the scoring model can be random forest, XGBC, etc., without specific limitations.
[0033] In one implementation, a scoring model is used to predict the corresponding user based on the user data of each user in the user cluster.
[0034] Step S13: Combine the first potential user cluster and the second potential user cluster to obtain the potential fund users in the user cluster.
[0035] In this embodiment, the potential fund users in the user cluster are obtained by combining the first potential user cluster and the second potential user cluster. Since the first potential user cluster is obtained using a scoring model, and the second potential user cluster is obtained using KYC rules, the potential fund users in the user cluster obtained by combining the first and second potential user clusters can be regarded as a prediction of potential fund users in the user cluster by combining the algorithm model and KYC rules. Compared with the method of predicting potential fund users by using only the algorithm model or KYC rules, the addition of KYC rules supplements the users that the algorithm model does not consider, thereby improving the accuracy of the prediction of potential fund users.
[0036] Since both the scoring model and KYC rules are based on user clusters to predict potential fund users, the same user may exist in both the first and second potential user clusters. Therefore, in one implementation, it is necessary to identify the overlapping portions in the first and second potential user clusters to avoid duplication.
[0037] Specifically, based on the first and second potential user clusters, overlapping user clusters, a first non-overlapping user cluster, and a second non-overlapping user cluster are obtained. Each user in the overlapping user cluster exists in both the first and second potential user clusters; each user in the first non-overlapping user cluster exists in the first potential user cluster; and each user in the second non-overlapping user cluster exists in the second potential user cluster. Then, the users in the overlapping, first, and second non-overlapping user clusters are considered as potential fund users within the user clusters. For example, if the user cluster includes users A, B, C, D, E, and F; the first potential user cluster includes users A, B, and C; and the second potential user cluster includes users A, B, D, and E, then the overlapping user cluster includes users A, B, and D. Therefore, the potential fund users obtained by combining the overlapping, first, and second non-overlapping user clusters are: users A, B, C, D, and E.
[0038] Of course, in other embodiments, when there is no overlap between the first potential user cluster and the second potential user cluster, the first potential user cluster and the second potential user cluster can be directly superimposed to obtain the fund potential users in the user cluster.
[0039] Step S14: Determine the user type of potential users for each fund.
[0040] In this embodiment, the user type of each fund's potential users is determined so that fund products matching the user type can be sent to the user in the future. That is, fund products that are more suitable for the user's current stage can be sent to the user in the future.
[0041] The user types include at least one of the following: investment novices, those who have not purchased funds, growth investors, and mature investors.
[0042] Step S15: Send fund products that match the corresponding user type to each fund's potential users.
[0043] In this implementation, fund products matching the corresponding user type are sent to each potential fund user. In other words, sending fund products matching the corresponding user type to each potential fund user increases the likelihood of subsequent users purchasing fund products.
[0044] For example, let's consider user types including novice investors, those who haven't invested in funds, growth investors, and experienced investors. User A is a novice investor, and User B is someone who hasn't invested in funds. So, for users A and B, the focus is on breaking the ice, so we send user A fund products like "Wealth Snowflakes." User C is a growth investor, and User D is an experienced investor. So, for users C and D, the focus is on product recommendations, so we send user C and D fund products like "Safe Investment."
[0045] KYC rules include a KYC tag pool, which contains several KYC tags. KYC rules predict potential users based on these tags. Therefore, to improve the accuracy of subsequent potential user predictions based on KYC rules, in one embodiment, it is determined whether the tags in the KYC tag pool need to be updated at a first preset frequency. The first preset frequency is not limited and can be set according to actual usage needs; for example, the first preset frequency is one day, meaning that it is determined daily whether the tags in the KYC tag pool need to be updated.
[0046] In one specific implementation, determining whether the tags in the KYC tag pool need to be updated involves: determining whether any tags are missing in the KYC tag pool; and in response to the absence of any tag in the KYC tag pool, sending a second update message indicating that the tag pool needs to be updated. In other words, the tags in the KYC tag pool are monitored, and when a tag is missing, an exception warning is issued so that users can promptly replenish the missing tags in the KYC tag pool to update the tag pool.
[0047] To improve the accuracy of subsequent potential user predictions based on the scoring model, in one embodiment, a second preset frequency is used to determine whether the scoring model needs to be updated. The second preset frequency is not limited and can be set according to actual usage needs; for example, the second preset frequency is once a week, meaning that a weekly determination is made regarding whether the scoring model needs to be updated.
[0048] In one specific implementation, the scoring model predicts a target score for a user and determines whether the user is a first potential user based on whether the target score is greater than or equal to a scoring threshold. It then determines whether the scoring model needs to be updated. Specifically, this involves: obtaining update parameters corresponding to the scoring model, where the update parameters include the current stability index of the scoring model, scoring data, and the number of conversions. The scoring data is the target score predicted by the scoring model for each user, and the number of conversions is the number of users who purchased the fund product within a preset time period after receiving the sent fund product. In response to any update parameter meeting the update requirements, a first update message indicating that the scoring model needs to be updated is sent.
[0049] For example, taking the updated parameters as rating data, count the number of users corresponding to each target rating. If the number of users corresponding to any target rating is 0, it indicates that there is an anomaly in the rating model and the update requirements of the rating model are met. At this time, the first update message that needs to update the rating model is sent.
[0050] For example, let's take the updated parameter as the number of conversions and the current rating threshold as 5 points. Since the current rating threshold is 5 points, users with a target rating of -5 points and above are all considered as first potential users. However, if the number of conversions corresponding to a target rating of 5 points is small, that is, the number of users with a target rating of -5 points who actually purchase funds is small, the update requirements are met. Therefore, it is no longer necessary to consider users with a target rating of -5 points as first potential users. Instead, the rating threshold needs to be adjusted to 6 points. Therefore, at this time, the first update message that requires updating the rating model is sent.
[0051] In the above implementation, the first potential user cluster is obtained using a scoring model, while the second potential user cluster is obtained using KYC rules. Combining the first and second potential user clusters yields the potential fund users within the user cluster. Therefore, by combining the algorithm model and KYC rules to jointly predict potential fund users within the user cluster, compared to using either the algorithm model or KYC rules alone, the inclusion of KYC rules supplements the prediction of users not considered by the algorithm model, thus improving the accuracy of predicting potential fund users.
[0052] In addition, the user type of each potential user will be determined so that fund products matching their type can be sent to them.
[0053] Please see Figure 2 , Figure 2 yes Figure 1 The flowchart shown is a schematic diagram of one embodiment of step S12. It should be noted that if substantially the same result is achieved, this embodiment does not necessarily follow the same pattern. Figure 2 The illustrated process sequence is limited. For example... Figure 2 As shown, this embodiment includes:
[0054] Step S21: For users in each user cluster, use the rating model to predict the target rating of the user.
[0055] In this embodiment, for users in each user cluster, a rating model is used to predict the target rating of the corresponding user.
[0056] In one implementation, for users in each user cluster, a rating model is used to predict the target rating of the user based on the user's user data.
[0057] In one specific implementation, a rating model is used. When making predictions about a user based on their user data, the rating predicted by the rating model is directly used as the target rating. In other implementations, such as... Figure 3 As shown, Figure 3 yes Figure 2 The flowchart shown in step S21 is a schematic diagram of an embodiment. The scoring model includes at least two models. Based on the user's user data, the user is predicted using the at least two scoring models. Specifically, it includes the following sub-steps:
[0058] Step S31: Using at least two rating models based on the user's user data, predict the user's rating and obtain at least two initial ratings for the corresponding user.
[0059] In this embodiment, at least two rating models are used to predict the user's performance based on the user's data, resulting in at least two initial ratings for each user. Each initial rating corresponds to one rating model. The number and type of rating models are not limited and can be set according to actual needs; for example, the number of rating models may include 2, 3, 4, or 5, and the rating models may be random forests or XGBC, etc.
[0060] The user data includes at least one of the following: basic data, behavioral data, financial data, and risk data. In one embodiment, basic data includes at least one of age, education level, marital status, occupation, and annual income; behavioral data includes at least one of logging into the client, clicking on wealth management-related pages, clicking on wealth management products, and purchasing wealth management products; financial data includes at least one of asset information, wealth management information, policy information, claims information, premium information, loan information, and repayment information; and risk data includes at least one of risk level and risk tolerance level.
[0061] To improve the accuracy of the rating model predictions, in one embodiment, before using at least two rating models to predict user data and obtain at least two initial ratings for the corresponding user, the user data is cleaned. This data cleaning includes at least one of the following: field classification, saturation filtering, field type conversion, filling in missing values, and chi-square filtering. In other words, the user data is cleaned and processed before being fed into the rating model.
[0062] In one implementation, the scoring model includes two models: one is a random forest and the other is XGBC.
[0063] Step S32: Based on at least two initial ratings for the corresponding user, obtain the target rating for the corresponding user.
[0064] In this embodiment, the target score for the corresponding user is obtained based on at least two initial scores. In one embodiment, the target score is obtained by weighted summing of the at least two initial scores. Of course, in other embodiments, the target score can also be obtained by averaging the at least two initial scores.
[0065] Step S22: In response to the target score being greater than or equal to the score threshold, the user is identified as the first potential user.
[0066] In this embodiment, a user is identified as a first potential user when the target score is greater than or equal to a scoring threshold. In other words, if a user's target score is greater than the scoring threshold, it indicates a higher probability that the user will subsequently purchase the fund, thus the user is identified as a first potential user.
[0067] The scoring threshold is the minimum model score in historical statistics where the number of conversions exceeds the threshold. The number of conversions is the number of users who purchased the fund within a preset time period after the prediction, based on the predicted number of users corresponding to the model score. There is no fixed threshold; it can be set according to actual usage needs. For example, as shown in Table 1 below, with a threshold of 10,000, and historical statistics based on predicted data from January 2022, since the conversions corresponding to model score 6 (-28,133) are greater than the threshold of -10,000, model score 7 (-81,087) is greater than the threshold of -10,000, model score 8 (-28,133) is greater than the threshold of -10,000, model score 9 (-21,522) is greater than the threshold of -10,000, and model score 10 (-30,659) is greater than the threshold of -10,000, model score 6 is used as the scoring threshold. Therefore, when a user's target score is greater than or equal to 6, the user is identified as a first potential user.
[0068] Table 1
[0069]
[0070] In other implementations, the scoring threshold may also be preset; for example, the preset scoring threshold may be 6 or 7, etc.
[0071] Step S23: Extract each first potential user from the user cluster to obtain the first potential user cluster.
[0072] In this embodiment, each first potential user is extracted from the user cluster to obtain the first potential user cluster. That is, the first potential user cluster is obtained by statistically analyzing each first potential user in the user cluster.
[0073] It should be noted that each user in the user cluster needs to execute steps S21-S23 separately.
[0074] Please see Figure 4 , Figure 4 yes Figure 1 The flowchart of another embodiment of step S12 is shown. It should be noted that if substantially the same result is achieved, this embodiment does not necessarily follow the same pattern. Figure 4 The illustrated process sequence is limited. For example... Figure 4 As shown, this embodiment includes:
[0075] Step S41: From the user cluster, select users who match any KYC tag as second potential users.
[0076] The KYC rules include a KYC tag pool, which contains several KYC tags. In this implementation, users matching any KYC tag are selected from the user cluster as potential second-tier users. In other words, users can set KYC tags according to their actual needs, thereby selecting users matching those tags as potential second-tier users.
[0077] The KYC tags include non-investment management users, non-fund holders, non-bank active users, and holders of at least one of the following fund types: a predetermined number. For example, taking publicly offered funds as an example, the KYC tags include non-investment management clients, non-public fund holders, non-bank active clients, and the cumulative number of publicly offered non-loan funds held.
[0078] Step S42: Combine the second potential users to obtain the second potential user cluster.
[0079] In this embodiment, each second potential user is combined to obtain a second potential user cluster. That is, each second potential user in the user cluster is counted to obtain the second potential user cluster.
[0080] It should be noted that each user in the user cluster needs to execute steps S41-S42 separately.
[0081] The following example illustrates how to predict potential users using mutual funds, random forests and XGBC as scoring models, and KYC rules based on several tags in a tag pool. In the fintech field, combining scoring models and KYC rules can accurately identify potential customers who want to buy mutual funds from a user cluster.
[0082] Specifically, user data of the customer cluster is obtained; then, based on the customer data, the scoring model - random forest and XGBC are used to determine whether each customer in the customer cluster is a potential customer to purchase public funds, and based on the customer data, KYC rules are used to determine whether each customer in the customer cluster is a potential customer to purchase public funds; then, the potential users of the fund in the user cluster are obtained by combining the potential users obtained by the scoring model - random forest and XGBC and the potential users obtained by the KYC rules.
[0083] Please see Figure 5 , Figure 5 This is a schematic diagram of an embodiment of the electronic device provided in this application. The electronic device 50 includes a memory 51 and a processor 52 coupled to each other. The processor 52 is used to execute program instructions stored in the memory 51 to implement the steps of any of the above-described fund product determination method embodiments. In a specific implementation scenario, the electronic device 50 may include, but is not limited to, a microcomputer or a server. In addition, the electronic device 50 may also include mobile devices such as laptops and tablets, which are not limited here.
[0084] Specifically, processor 52 controls itself and memory 51 to implement the steps of any of the above-described fund product determination method embodiments. Processor 52 may also be referred to as a CPU (Central Processing Unit). Processor 52 may be an integrated circuit chip with signal processing capabilities. Processor 52 may also be a general-purpose processor, digital signal processor (DSP), application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. A general-purpose processor may be a microprocessor or any conventional processor. Furthermore, processor 52 may be implemented using integrated circuit chips.
[0085] Please see Figure 6 , Figure 6This is a schematic diagram of an embodiment of the computer-readable storage medium provided in this application. The computer-readable storage medium 60 of this application embodiment stores program instructions 61. When executed, these program instructions 61 implement the method provided by any embodiment and any non-conflicting combination of the fund product determination method of this application. The program instructions 61 can form a program file and be stored in the aforementioned computer-readable storage medium 60 in the form of a software product, so that a computer device (which may be a personal computer, server, or network device, etc.) can execute all or part of the steps of the methods of various embodiments of this application. The aforementioned computer-readable storage medium 60 includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks, or terminal devices such as computers, servers, mobile phones, and tablets.
[0086] If the technical solution of this application involves personal information, the product using this technical solution has clearly informed the user of the personal information processing rules and obtained the user's voluntary consent before processing the personal information. If the technical solution of this application involves sensitive personal information, the product using this technical solution has obtained the user's separate consent before processing the sensitive personal information, and also meets the requirement of "express consent". For example, at personal information collection devices such as cameras, clear and prominent signs are set up to inform users that they have entered the scope of personal information collection and that personal information will be collected. If an individual voluntarily enters the collection scope, it is deemed that they have agreed to the collection of their personal information; or on the personal information processing device, with clear signs / information informing users of the personal information processing rules, authorization is obtained from the individual through pop-up information or by asking the individual to upload their personal information; wherein, the personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the types of personal information processed.
[0087] The above description is merely an embodiment of this application and does not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.
Claims
1. A method for determining a fund product, characterized in that, The method includes: Obtain the user cluster; A first potential user cluster is obtained by using a scoring model to predict the users in the user cluster, and the first potential user cluster includes users who are predicted to be first potential users by the scoring model; and a second potential user cluster is obtained by using KYC rules to filter the users in the user cluster, and the second potential user cluster includes users who are filtered to be second potential users by the KYC rules. By combining the first potential user cluster and the second potential user cluster, the potential fund users in the user cluster are obtained; Determine the user type of potential users for each of the aforementioned funds; Send fund products matching the corresponding user type to each of the fund's potential users; The step of using a scoring model to predict the performance of each user in the user cluster to obtain the first potential user cluster includes: For each user in each user cluster, two rating models are used to predict the user based on the user data corresponding to the user, resulting in two initial ratings for the user; wherein, each initial rating corresponds to one rating model, the user data includes at least one of basic data, behavioral data, financial data, and risk data, and the rating models include random forest and XGBoost. The target score is obtained by weighted summation of the initial scores. In response to the target score being greater than or equal to a score threshold, the user is determined to be the first potential user; wherein, the score threshold is the smallest model score in historical statistics where the number of converted users is greater than a number threshold, and the number of converted users is the number of users who purchase the fund within a preset time period after the prediction of the predicted number of users corresponding to the model score; Extract each of the first potential users from the user cluster to obtain the first potential user cluster.
2. The method according to claim 1, characterized in that, The basic data includes at least one of age, education, marital status, occupation, and annual income; the behavioral data includes at least one of logging into the client, clicking on a wealth management-related page, clicking on a wealth management product, and purchasing a wealth management product; the financial data includes at least one of asset information, wealth management information, policy information, claims information, premium information, loan information, and repayment information; and the risk data includes at least one of risk level and risk tolerance level.
3. The method according to claim 1, characterized in that, Before using the rating model to predict the user based on the user data corresponding to the user and obtain the target rating for the corresponding user, the method further includes: Data cleaning is performed on the user data corresponding to the user; wherein, data cleaning includes at least one of field classification, saturation filtering, field type conversion, missing value imputation, and chi-square filtering.
4. The method according to claim 1, characterized in that, The KYC rules include a KYC tag pool, which contains several KYC tags; the step of using the KYC rules to filter users in the user cluster to obtain a second potential user cluster includes: From the user cluster, users matching any of the KYC tags are selected as the second potential users; wherein, the KYC tags include at least one of the following: non-investment users, non-fund holders, non-bank valid users, and holders of a preset number of funds of a certain type; The second potential users are combined to obtain the second potential user cluster.
5. The method according to claim 1, characterized in that, The scoring model predicts a target score for the user and determines whether the user is a first potential user based on whether the target score is greater than or equal to a scoring threshold. After sending fund products matching the corresponding user type to each of the fund's potential users, the method further includes: Obtain the update parameters corresponding to the scoring model; wherein, the update parameters include the current stability index of the scoring model, the scoring data, and the number of conversions; wherein, the scoring data is the target score predicted by the scoring model for each user, and the number of conversions is the number of users who purchase the fund product within a preset time period after receiving the sent fund product; In response to any of the updated parameters meeting the corresponding update requirements, a first update message requiring an update to the scoring model is sent. And / or, the KYC rules include a KYC tag pool, the KYC tag pool includes several KYC tags, and the KYC rules make predictions based on the several tags; after sending fund products matching the corresponding user type to each of the potential fund users, the method further includes: Determine whether there are any missing tags in the KYC tag pool; In response to the absence of any of the tags in the KYC tag pool, a second update message is sent requiring the tag pool to be updated.
6. The method according to claim 1, characterized in that, The step of combining the first potential user cluster and the second potential user cluster to obtain the potential fund users in the user cluster includes: Based on the first potential user cluster and the second potential user cluster, an overlapping user cluster, a first non-overlapping user cluster, and a second non-overlapping user cluster are obtained, and users in the overlapping user cluster, the first non-overlapping user cluster, and the second non-overlapping user cluster are considered as potential users of the fund; wherein, each user in the overlapping user cluster exists in both the first potential user cluster and the second potential user cluster, each user in the first non-overlapping user cluster exists in the first potential user cluster, and each user in the second non-overlapping user cluster exists in the second potential user cluster. And / or, the fund is a public fund.
7. An electronic device, characterized in that, The electronic device includes a processor and a memory, the memory storing program instructions, and the processor executing the program instructions to implement the method for determining fund products as described in any one of claims 1-6.
8. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store program instructions that can be executed to implement the method for determining fund products as described in any one of claims 1-6.