Financial product recommendation method and device, silent customer activation method and device and electronic equipment

Abstract

The invention provides a financial product recommendation method, and relates to the field of artificial intelligence. The method comprises the following steps: inputting user characteristics of a user into a pre-trained asset prediction model; obtaining N types of predicted underlying assets output by the asset prediction model and proportion information of each type of predicted underlying assets; on the basis of the risk assessment level of the user, querying one or more candidate financial products matched with the risk assessment level from a database to form a candidate product set; based on the N types of predicted underlying assets and the proportion information of each type of predicted underlying assets, matching M target financial products from a candidate product set, the candidate product set comprising at least one candidate product and the proportion information of each type of underlying assets of each candidate product; and recommending the M target financial products to the user. The invention further provides a silent customer activation method.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence, and more specifically, to a financial product recommendation method, a silent customer activation method, a device, a device, a medium, and a program product. Background Art

[0002] Financial product recommendation usually refers to providing investors with suitable financial products or investment portfolios based on their financial status, risk preferences, investment goals, etc. In related technologies, financial products that match the investor's information are generally recommended based on the investor's information.

[0003] In the process of realizing the invention concept of the present disclosure, the inventors found that when the investor's information is incomplete, there is no suitable recommendation plan, resulting in a lack of effective activation methods for silent customers of financial institutions. In addition, the existing recommendation methods usually screen out recommended products from all available financial products based on the characteristics of investors, and in this process, the investor characteristics are matched and calculated with the characteristics of each product. In fact, matching and calculating the investor characteristics with the product characteristics that obviously do not meet the recommendation requirements is an ineffective calculation that wastes resources and takes a lot of time when the amount of product data is large. Summary of the invention

[0004] In view of the above problems, the present disclosure provides a financial product recommendation method, a silent customer activation method, an apparatus, a device, a medium and a program product.

[0005] According to a first aspect of the present disclosure, a financial product recommendation method is provided, characterized in that it includes: inputting user characteristics of a user into a pre-trained asset prediction model, wherein the asset prediction model is trained based on financial product samples, and the financial product samples include information on the proportion of each type of underlying assets; obtaining N types of predicted underlying assets and proportion information of each type of predicted underlying assets output by the asset prediction model; based on the risk assessment level of the user, querying one or more candidate financial products matching the risk assessment level from a database to form a candidate product set, wherein candidate financial product information corresponding to different risk assessment levels is pre-stored in the database; based on the N types of predicted underlying assets and proportion information of each type of predicted underlying assets, matching M target financial products from the candidate product set, the candidate product set including at least one candidate product and proportion information of each type of underlying assets of each candidate product, N and M are both integers greater than or equal to 1; and recommending the M target financial products to the user.

[0006] According to an embodiment of the present disclosure, based on the N types of predicted underlying assets and the proportion information of each type of predicted underlying assets, matching M target financial products from the candidate product set includes: mapping the N types of predicted underlying assets with the underlying asset categories in each candidate product; for candidate products with a preset number of categories in the N types of predicted underlying assets, comparing the proportion information of each type of underlying assets in the candidate product with the proportion information of each of the N types of predicted underlying assets to obtain a proportion information comparison result; and determining the target financial product when the proportion information comparison result is characterized as similar or identical. .

[0007] According to an embodiment of the present disclosure, the asset prediction model includes K asset prediction sub-models corresponding one-to-one to K risk levels, and inputting the user characteristics of the user into the pre-trained asset prediction model includes: inputting the user characteristics of the user into the asset prediction sub-model corresponding to the risk assessment level; wherein different asset prediction sub-models are trained using financial product samples that meet different risk levels.

[0008] According to an embodiment of the present disclosure, the risk level includes level R1, and the K asset prediction sub-models include an R1-level sub-model. When the risk assessment level is level R1, obtaining the N types of predicted underlying assets output by the asset prediction model and the proportion information of each type of predicted underlying assets includes: obtaining the respective proportion information of the money market underlying assets and the bond underlying assets output by the R1-level sub-model.

[0009] According to an embodiment of the present disclosure, the risk level includes level R2, and the K asset prediction sub-models include an R2-level sub-model. When the risk assessment level is level R2, obtaining the N types of predicted underlying assets output by the asset prediction model and the proportion information of each type of predicted underlying assets includes: obtaining the respective proportion information of the money market underlying assets, bond underlying assets and equity underlying assets output by the R2-level sub-model.

[0010] According to an embodiment of the present disclosure, the risk level includes R3 and above, the K asset prediction sub-models include an R3 sub-model, and when the risk assessment level is R3 or above, obtaining the N types of predicted underlying assets output by the asset prediction model and the proportion information of each type of predicted underlying assets includes: obtaining the respective proportion information of money market underlying assets, bond underlying assets, equity underlying assets and derivative underlying assets output by the R3 sub-model.

[0011] According to an embodiment of the present disclosure, each of the asset prediction sub-models includes a meta-learner and multiple base learners, and inputting the user characteristics of the user into the asset prediction sub-model corresponding to the risk assessment level includes: calling the multiple base learners of the asset prediction sub-model, so that the multiple base learners process the user characteristics of the user respectively, and output multiple first prediction results one by one; performing weighted calculation based on the multiple intermediate prediction results and their respective weight coefficients to obtain multiple second prediction results; inputting the multiple second prediction results into the meta-learner, so that the meta-learner processes the multiple second prediction results, wherein the meta-learner is used to output the N types of predicted underlying assets and the proportion information of each type of predicted underlying assets.

[0012] Another aspect of an embodiment of the present disclosure provides a silent customer activation method, characterized in that it includes: determining a silent customer set; determining at least one silent customer level to which each of the users belongs based on transaction information of each user in the silent customer set; for different silent customer levels, activating the users in the level in turn in different orders and using different activation strategies; wherein the silent customer level includes a financial product silent level, and for users in the financial product silent level, executing the financial product recommendation method as described in any of the above items to construct an activation strategy.

[0013] Another aspect of the disclosed embodiment provides a financial product recommendation device, characterized in that it includes: an input module, used to input the user characteristics of a user into a pre-trained asset prediction model, wherein the asset prediction model is obtained by training based on financial product samples, and the financial product samples include information on the proportion of each type of underlying assets; a prediction module, used to obtain N types of predicted underlying assets and the proportion information of each type of predicted underlying assets output by the asset prediction model; a pre-screening module, used to query one or more candidate financial products matching the risk assessment level from a database based on the risk assessment level of the user to form a candidate product set, wherein candidate financial product information corresponding to different risk assessment levels is pre-stored in the database; a matching module, used to match M target financial products from the candidate product set based on the N types of predicted underlying assets and the proportion information of each type of predicted underlying assets, the candidate product set including at least one candidate product and the proportion information of each type of underlying assets of each candidate product, and N and M are both integers greater than or equal to 1; a recommendation module, used to recommend the M target financial products to the user.

[0014] Another aspect of an embodiment of the present disclosure provides a silent customer activation device, characterized in that it includes: a first determination module, used to determine a silent customer set; a second determination module, used to determine at least one silent customer level to which each user belongs based on transaction information of each user in the silent customer set; a silent customer activation module, used to activate users in different silent customer levels in turn according to different orders and using different activation strategies; wherein the silent customer level includes a financial product silent level, and for users in the financial product silent level, the financial product recommendation method as described in any of the above items is executed to construct an activation strategy.

[0015] Another aspect of an embodiment of the present disclosure provides an electronic device, including: one or more processors; and a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors execute the method described above.

[0016] Another aspect of an embodiment of the present disclosure provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the processor executes the method as described above.

[0017] Another aspect of an embodiment of the present disclosure provides a computer program product, including a computer program, which implements the above method when executed by a processor.

[0018] One or more of the above embodiments have the following beneficial effects:

[0019] By predicting the proportion of underlying assets, personalized investment advice can be provided to users to meet the financial goals and risk preferences of different users. In addition, a set of candidate products is first determined according to the user's risk assessment level to divide the data range of the product to be recommended, and then matching is performed within the data range, that is, M target financial products are matched from the candidate product set. In this way, through a two-stage matching process, the range of candidate products can be narrowed down first, and then matching calculations can be performed within a determined range. The computer resources required for matching the target financial product stage can be saved, the overall calculation speed of financial product recommendations can be improved, and invalid matching calculation processes that do not meet the risk level can be omitted. In addition, for example, when the user's risk level is determined, only the candidate financial product information corresponding to the risk level can be pre-stored, which reduces the resources consumed by the database to pre-store data and saves storage costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The above contents and other objects, features and advantages of the present disclosure will become more apparent through the following description of the embodiments of the present disclosure with reference to the accompanying drawings, in which:

[0021] Figure 1 A diagram schematically showing an application scenario of a financial product recommendation method or a silent customer activation method according to an embodiment of the present disclosure;

[0022] Figure 2 A flowchart of a financial product recommendation method according to an embodiment of the present disclosure is schematically shown;

[0023] Figure 3 A flowchart for determining a candidate product set according to an embodiment of the present disclosure is schematically shown;

[0024] Figure 4 A control block diagram of financial product recommendation according to an embodiment of the present disclosure is schematically shown;

[0025] Figure 5 A control block diagram of R1-level financial product recommendation according to an embodiment of the present disclosure is schematically shown;

[0026] Figure 6 A control block diagram of R2-level financial product recommendation according to an embodiment of the present disclosure is schematically shown;

[0027] Figure 7 A control block diagram for recommending financial products at level R3 and above according to an embodiment of the present disclosure is schematically shown;

[0028] Figure 8 Schematically shows a flowchart of asset prediction sub-model data processing according to an embodiment of the present disclosure;

[0029] Fig. 9 Schematically shows a flow chart of a silent client activation method according to an embodiment of the present disclosure;

[0030] Fig.10 The structure block diagram of the financial product recommendation device according to the embodiment of the present disclosure is schematically shown;

[0031] Fig.11 The structure block diagram of the financial product recommendation device according to the embodiment of the present disclosure is schematically shown;

[0032] Fig.12 A block diagram of an electronic device suitable for implementing a financial product recommendation method or a silent customer activation method according to an embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0033] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0034] In the technical solution of the present invention, the user information (including but not limited to user personal information, user image information, user device information, such as location information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved are all information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data comply with relevant laws, regulations and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0035] In the scenario of using personal information for automated decision-making, the methods, devices, and systems provided by the embodiments of the present invention provide users with corresponding operation portals for users to choose to agree or reject the automated decision-making results; if the user chooses to reject, the expert decision-making process will be entered. The expression "automated decision-making" here refers to the activity of automatically analyzing and evaluating an individual's behavioral habits, interests and hobbies, or economic, health, credit status, etc. through computer programs, and making decisions. The expression "expert decision-making" here refers to the activity of making decisions by people who specialize in a certain field, have specialized experience, knowledge and skills, and have reached a certain level of professionalism.

[0036] Figure 1 The application scenario diagram of the financial product recommendation method or silent customer activation method according to the embodiment of the present disclosure is schematically shown. It should be noted that: Figure 1 What is shown are merely examples to which the embodiments of the present disclosure can be applied, so as to help those skilled in the art understand the technical content of the present disclosure, but it does not mean that the embodiments of the present disclosure cannot be used in other devices, systems, environments or scenarios.

[0037] like Figure 1 As shown, the application scenario 100 according to this embodiment may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for a communication link between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0038] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only examples).

[0039] The terminal devices 101 , 102 , and 103 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, and desktop computers.

[0040] The server 105 may be a server that provides various services, such as a background management server that provides support for websites browsed by users using the terminal devices 101, 102, and 103 (for example only). The background management server may analyze and process the received data such as user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal device. For example, the server 105 may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud computing, network services, and middleware services.

[0041] In applications such as client applications and web applications (abbreviated as APP in English), the client (i.e., front-end) and the server (i.e., back-end) can communicate data through network messages. For example, in the APP client, a financial product recommendation instruction or a silent customer activation instruction is received, and the relevant instructions or user information are sent to the server by calling the network request method.

[0042] It should be noted that the financial product recommendation method or silent customer activation method provided in the embodiments of the present disclosure can generally be executed by at least one of a terminal device or a server. Accordingly, the financial product recommendation device or silent customer activation device provided in the embodiments of the present disclosure can generally be set in at least one of a terminal device or a server.

[0043] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0044] The following will be based on Figure 1 The scene described by Figure 2~Figure 9 The financial product recommendation method and silent customer activation method of the embodiments of the present disclosure are described in detail.

[0045] Figure 2The flowchart of the financial product recommendation method according to the embodiment of the present disclosure is schematically shown.

[0046] like Figure 2 As shown, this embodiment includes:

[0047] In operation S210, the user characteristics of the user are input into a pre-trained asset prediction model, wherein the asset prediction model is trained based on financial product samples, and the financial product samples include information on the proportion of various underlying assets;

[0048] The user of this embodiment may be a customer of a financial institution, and the user characteristics may be extracted based on one or more information such as basic information, product information held, asset status, off-bank funds and financial management status, on-bank funds and financial management status, online browsing behavior, etc. The asset prediction model may be a model trained based on a machine learning algorithm, such as a neural network algorithm, a genetic algorithm, a decision tree, or a random forest. Financial product samples may be determined based on the historical sales data of a financial institution, or may be automatically generated financial products. The underlying assets are the most basic assets that constitute a financial product, such as stocks, bonds, real estate, etc.

[0049] In operation S220, the N types of predicted underlying assets output by the asset prediction model and the proportion information of each type of predicted underlying assets are obtained; the underlying asset configuration suitable for the user is predicted, for example, the asset prediction model outputs that bond assets account for 20% and stock assets account for 80%.

[0050] In operation S230, based on the risk assessment level of the user, one or more candidate financial products matching the risk assessment level are searched from a database to form a candidate product set, wherein information of candidate financial products corresponding to different risk assessment levels is pre-stored in the database;

[0051] For example, a corresponding database table is created in the database for each risk level. Each database table stores information of one or more candidate financial products. Alternatively, each candidate financial product on sale in the database is marked with a suitable risk level identifier.

[0052] In operation S240, based on N types of predicted underlying assets and the proportion information of each type of predicted underlying assets, M target financial products are matched from the candidate product set, where the candidate product set includes at least one candidate product and the proportion information of each type of underlying assets of each candidate product, where N and M are both integers greater than or equal to 1;

[0053] The candidate product set includes one or more financial products that can be publicly sold by financial institutions now or in the future, such as fund products and financial products.

[0054] In some embodiments, operation S240 includes mapping the distribution of N types of predicted underlying assets with the underlying asset categories in each candidate product; for candidate products having a preset number of categories in the N types of predicted underlying assets, comparing the proportion information of each type of underlying asset in the candidate product with the proportion information of each type of N types of predicted underlying assets to obtain a proportion information comparison result; when the proportion information comparison result is characterized as similar or identical, it is determined as a target financial product.

[0055] Specifically, in the matching process, for each candidate product in the candidate product set, the N types of predicted underlying assets are first mapped to the underlying asset categories in the candidate product to determine whether they contain or approximately contain the N types of asset categories predicted by the prediction model. If part (filtering conditions can be set in advance) or all of the N types of assets are included, then for the proportion information of each type of underlying asset, find products in the candidate products whose asset allocation matches it. For example, if the model predicts that stock assets should account for 60% of the investment portfolio, then look for products whose stock allocation is close to this proportion. Finally, M target financial products are obtained. In some embodiments, M best matches can also be selected based on expected returns, product fees, and historical performance.

[0056] In operation S250 , M target financial products are recommended to the user.

[0057] The product name, issuing institution, product type, risk level, performance benchmark, expected rate of return, expected rate of return, fee structure, fee structure, investment period and underlying assets of each target financial product will be displayed to users, and users can purchase it according to their own circumstances.

[0058] According to the embodiments of the present disclosure, by predicting the proportion of underlying assets, personalized investment advice can be provided to users to meet the financial goals and risk preferences of different users. In addition, a set of candidate products is first determined according to the user's risk assessment level to divide the data range of the product to be recommended, and then matching is performed within the data range, that is, M target financial products are matched from the candidate product set. In this way, through a two-stage matching process, the range of candidate products can be narrowed down first, and then matching calculations can be performed within a determined range. The computer resources required for matching the target financial product stage can be saved, the overall calculation speed of financial product recommendations can be improved, and useless matching calculation processes that do not conform to the risk level can be omitted. Moreover, for example, when the user's risk level is determined, only the candidate financial product information corresponding to the risk level can be pre-stored, which reduces the resources consumed by the database to pre-store data and saves storage costs.

[0059] In addition, the predicted underlying asset ratio helps users achieve optimal asset allocation, reduce investment risks and increase expected returns. Therefore, it can improve users' acceptance and satisfaction with the recommended target financial products.

[0060] Figure 3 The flowchart of determining a candidate product set according to an embodiment of the present disclosure is schematically shown.

[0061] Before matching M target financial products from the candidate product set based on N types of predicted underlying assets and the proportion of each type of predicted underlying assets, such as Figure 3 As shown, this embodiment is one of the embodiments of operation S230, which specifically includes:

[0062] In operation S310, the risk assessment level of the user is obtained; the risk preference of the user can be understood, providing a basis for personalized recommendation.

[0063] In operation S320, a set of candidate products that meet the risk assessment level is determined, for example, by using a query statement to obtain the set of candidate products from a database.

[0064] The risk assessment level is the result of the classification of the user's risk tolerance, which is usually obtained based on questionnaires or data analysis. For example, investors evaluate their tolerance to market fluctuations through online questionnaires, and the system gives risk assessment levels such as "conservative (R1 level)", "robust (R2 level)" or "balanced or aggressive (R3 level and above)" based on the answers.

[0065] For example, after the user completes the risk assessment questionnaire, the system automatically gives a risk assessment level based on the answer. Based on the risk assessment level, the system uses database query to select products that meet this level from all financial products. For example, for "conservative" investors, low-volatility, fixed-income products will be included in the candidate product set.

[0066] According to the embodiments of the present disclosure, users will not be recommended products that exceed their risk tolerance, which reduces the user's investment risk, improves the personalization of financial product recommendations, and better meets the needs of different users.

[0067] Figure 4 A control block diagram of financial product recommendation according to an embodiment of the present disclosure is schematically shown.

[0068] In some embodiments, Figure 4 As shown, the asset prediction model includes K asset prediction sub-models corresponding to K risk levels, K is greater than or equal to 1, and inputting the user characteristics of the user into the pre-trained asset prediction model includes: inputting the user characteristics of the user into the asset prediction sub-model corresponding to the risk assessment level; wherein different asset prediction sub-models are trained using financial product samples that meet different risk levels.

[0069] Exemplarily, each asset prediction sub-model is independent of each other. Each asset prediction sub-model can be obtained by using a machine learning algorithm through a sample of financial products corresponding to a risk level. K risk levels are all possible evaluation results, and the user's risk evaluation level is the result determined by the system from the K risk levels based on the questionnaire survey results.

[0070] According to the embodiments of the present disclosure, by using financial product samples of a specific risk level to train a sub-model, and using the asset prediction sub-model corresponding to the risk assessment level for prediction, the accuracy of the prediction of the underlying assets of a specific risk level and the relevance of the recommendation are improved.

[0071] Figure 5 A control block diagram of R1-level financial product recommendation according to an embodiment of the present disclosure is schematically shown.

[0072] In some embodiments, Figure 5 As shown, the risk level includes R1 level, and the K asset prediction sub-models include the R1 level sub-model. When the risk assessment level is R1 level, the N types of predicted underlying assets output by the asset prediction model and the proportion information of each type of predicted underlying assets include: obtaining the respective proportion information of the money market underlying assets and the bond underlying assets output by the R1 level sub-model. Figure 5 Rectangles of different lengths are used to represent the respective proportions of money market underlying assets and bond underlying assets.

[0073] In the financial field, "R1-level financial products" are financial products with the lowest risk level. R1-level financial products prefer the safety of funds rather than high returns. Therefore, due to the lower risk, the expected return of such products is usually lower than that of medium or high-risk products, but they provide more stable returns. In terms of investment objects, R1-level products may mainly invest in low-risk assets such as money market instruments, short-term bonds, bank deposits, etc., which usually have lower credit risk and market risk.

[0074] Therefore, in the prediction and reasoning stages, the R1-level sub-model is constrained to be within the scope of money market underlying assets and bond underlying assets, and predict the proportion of the two types of underlying assets.

[0075] According to the embodiments of the present disclosure, it is possible to match the risk preference of level R1, so that the model prediction results are concentrated on low-risk assets, ensuring the accuracy of asset allocation, meeting the needs of investors with risk preferences of level R1, and being able to more accurately match the risk tolerance of investors and provide more personalized asset predictions.

[0076] Figure 6 A control block diagram of R2-level financial product recommendation according to an embodiment of the present disclosure is schematically shown.

[0077] In some embodiments, Figure 6 As shown, the risk level includes R2 level, and the K asset prediction sub-models include the R2 level sub-model. When the risk assessment level is R2 level, the N types of predicted underlying assets output by the asset prediction model and the proportion information of each type of predicted underlying assets include: obtaining the respective proportion information of the money market underlying assets, bond underlying assets and equity underlying assets output by the R2 level sub-model. Figure 6 Rectangles of different lengths are used to represent the respective proportions of money market underlying assets, bond underlying assets and equity underlying assets.

[0078] R2 is between conservative (R1) and balanced (R3 and above) products, matching investors who can accept a certain degree of volatility. R2 products may invest in a variety of asset classes, including bonds, money market instruments, stocks and other financial instruments to achieve risk diversification.

[0079] During the prediction and inference stages, the R2-level sub-model is constrained to predict the proportion of the three types of underlying assets within the scope of money market underlying assets, bond underlying assets and equity underlying assets.

[0080] According to the embodiments of the present disclosure, it is possible to match the risk preference of level R2 so that the model prediction results are concentrated on medium-risk assets, meeting the needs of level R2 investors.

[0081] Figure 7 A control block diagram for recommending financial products at level R3 and above according to an embodiment of the present disclosure is schematically shown.

[0082] In some embodiments, Figure 7 As shown, the risk level includes R3 and above, the K asset prediction sub-models include the R3 sub-model, and when the risk assessment level is R3 or above, the N types of predicted underlying assets output by the asset prediction model and the proportion information of each type of predicted underlying assets include: the proportion information of money market underlying assets, bond underlying assets, equity underlying assets and derivative underlying assets output by the R3 sub-model. Figure 7 Rectangles of different lengths are used to represent the respective proportions of money market underlying assets, bond underlying assets, equity underlying assets and derivative underlying assets.

[0083] From R1 to R5, the larger the number, the higher the risk level. R3 or above involves some medium-to-high-risk investments, such as partial equity investments.

[0084] The money market underlying assets involved in this disclosure include short-term, highly liquid financial instruments with low risk, such as short-term Treasury bonds and commercial paper that money market funds may invest in. Bond underlying assets refer to fixed-income assets, including bonds issued by governments or companies, such as corporate bonds or government bonds, which generally provide fixed interest payments and repay the principal at maturity. Equity underlying assets represent part of the ownership of a company, mainly stocks, and their value will change with market fluctuations, reflecting the ownership and growth potential of listed companies. Derivative underlying assets are financial instruments derived from other underlying assets, including options, futures and swaps, such as stock options, which give the holder the right to buy or sell stocks at a specific price within a specific time.

[0085] Corresponding to the higher risk, financial products rated R3 or above may invest in a variety of asset classes, including stocks, bonds, derivatives, etc., in pursuit of higher returns, thereby providing higher income potential, but with greater volatility and higher risk.

[0086] In the prediction and reasoning stage, the R3 sub-model predicts a larger range of underlying asset categories and obtains the proportion of each underlying asset category. In the proportions involved in this disclosure, the proportion of each underlying asset category is between 0% and 100%.

[0087] According to the embodiments of the present disclosure, it is possible to match risk preferences at level R3 and above, so that the model prediction results are concentrated on medium and high risk assets, meeting the needs of investors at level R3 and above.

[0088] For example, refer to Figure 4~Figure 7 If the user has not completed the risk assessment, the user will be guided to complete the risk assessment first. If the user has completed the risk assessment, the risk rating of the user who has not held funds or wealth management products for more than one month will be processed.

[0089] For R1 level users, the R1 level sub-model predicts the ratio of underlying assets in the money market and underlying assets in the debt category. Finally, according to the model prediction ratio, the fund and wealth management products of the R1 risk level to which the user belongs are matched with the product with the closest underlying asset ratio. Based on the purchase of all users in the past month, the top five products in user marketing are given certain fixed management fees, sales handling fees, custody fees or floating management fees.

[0090] For users at R2 level, the R2 sub-model predicts the proportions of three types of assets: money market underlying assets, debt underlying assets, and equity underlying assets. Finally, according to the model prediction ratio, the user's R2 risk level and below funds and wealth management products are matched with products with the closest underlying asset ratios. Based on the purchases of all users in the past month, the top five products in user marketing are given certain fixed management fees, sales fees, custody fees, or floating management fee discounts.

[0091] For users at R3 level and above, the R3 sub-model predicts the proportions of four types of assets: money market underlying assets, debt underlying assets, equity underlying assets, and derivative underlying assets. Finally, based on the model prediction ratio, according to the user's risk level and below, the fund and wealth management products are matched with the product with the closest underlying asset ratio. Based on the purchase of all users in the past month, the top five products in user marketing are given certain fixed management fees, sales fees, custody fees, or floating management fee discounts.

[0092] The independent variables of the R1-level sub-model, R2-level sub-model or R3-level sub-model include: basic information of users (gender, age, occupation, education level, nature of unit, marital status, etc.), information on products held (activation and balance of current deposits, time deposits, insurance, third-party custody, precious metals, treasury bonds, bonds, consumer loans, housing loans, operating loans, etc.), asset status (average monthly and daily assets in the past 1-6 months, etc.), off-bank fund and wealth management status (purchase of off-bank wealth management and off-bank funds with debit and credit cards in the past 1-6 months, etc.), in-bank fund and wealth management status (whether wealth management business is activated, whether fund business is activated, whether wealth management products have been purchased, whether fund products have been purchased, the proportion of four types of assets purchased for fund products, the proportion of four types of assets purchased for wealth management products, etc.), online channels (the number of visits to the wealth management page of the in-bank online platform in the past 1-6 months, the length of visits, the proportion of four types of assets visited for wealth management products, the number of visits to the fund page of the in-bank online platform in the past 1-6 months, the length of visits, the proportion of four types of assets visited for fund products), and other characteristics.

[0093] For example, customer A's risk assessment level in a commercial bank is R3 balanced. He has not purchased the bank's funds and wealth management products since he redeemed his wealth management product three months ago. The relevant characteristics of the customer are input into the trained R3-level sub-model, including basic customer information, product holding information, asset status, off-bank funds and wealth management status, in-bank funds and wealth management status, online channels and other characteristics. The model predicts that customer A's underlying assets are 35% of money market underlying assets, 25% of debt underlying assets, 20% of equity underlying assets, and 20% of derivative underlying assets. At this time, among the commercial bank's R3 and below funds and wealth management products, the model predicts that the model predicts that the underlying assets of 20 products are within the range of 20 products. According to the sales amount of the 20 products in the past month, the top five products are selected for marketing to customer A, and a sales commission discount of 0.05% is given.

[0094] Figure 8 The flowchart of asset prediction sub-model data processing according to an embodiment of the present disclosure is schematically shown.

[0095] Each asset prediction sub-model includes a meta-learner and multiple base learners, which input the user characteristics of the user into the asset prediction sub-model corresponding to the risk assessment level. Figure 8 As shown, inputting the user characteristics of the user into the asset prediction sub-model corresponding to the risk assessment level includes:

[0096] In operation S810, multiple base learners of each asset prediction sub-model are called, so that the multiple base learners process the user features of the user respectively, and output multiple first prediction results in a one-to-one correspondence; the first prediction results may include various types of underlying assets and their respective proportion information.

[0097] In operation S820, a weighted calculation is performed based on the plurality of intermediate prediction results and respective weight coefficients to obtain a plurality of second prediction results;

[0098] In operation S830, the plurality of second prediction results are input into a meta-learner, so that the meta-learner processes the plurality of second prediction results, wherein the meta-learner is configured to output N types of predicted underlying assets and proportion information of each type of predicted underlying assets.

[0099] For example, the base learner can be a single machine learning model or a weighted fusion model of multiple machine learning models, such as random forest, decision tree, GBDT (Gradient Boosting Decision Tree), XGBoost (eXtreme Gradient Boosting), LightGBM and other machine learning models. The meta learner can be trained using the commonly used generalized linear regression (GLR).

[0100] In some preferred embodiments, the base learners selected in the first layer are generally complex nonlinear transformations, so there is no need to use complex transformations in the selection of the second layer classifier. One or more of GBDT, XGBoost, and LightGBM can be selected to construct the first layer base learner, and the logistic regression with a simple model and good generalization performance can be used as the second layer meta learner.

[0101] According to the embodiments of the present disclosure, by combining the base learner and the meta learner, the advantages of various machine learning models can be integrated, the shortcomings and deficiencies of a single model can be overcome, the recommendation accuracy can be improved, and the prediction accuracy of the underlying asset ratio can be effectively improved.

[0102] The following is an explanation of the training process of each asset prediction sub-module based on the Stacking weighted fusion strategy. Since the input of the second-layer meta-learner is generated by the output of each base learner in the first layer, if the training set of the base learner is directly used as the training set of the second-layer meta-learner, the data is repeatedly learned by the two-layer learner, which can easily cause overfitting problems. In order to minimize the risk of overfitting of the model, the cross-validation method is usually used in the first layer to obtain the input features of the second layer. The meta-learner fully integrates the model advantages of each base learner in the first layer to build the model, thereby improving the effect of the model.

[0103] First, a weighted fusion strategy is implemented. The weighted algorithms GBDT-XGBoost, GBDT-LightGBM, and XGboost-LightGBM of the two models are constructed. Taking the GBDT-XGBoost algorithm as an example, the original features are first input into the GBDT model to obtain the prediction result 1, and then the original features are input into the XGBoost model to obtain the prediction result 2. The weights of the two models are determined by a simple weighted method. The prediction results are multiplied by the corresponding weights and added to obtain the final prediction result. The optimal combination model is fitted by multiple training of the data set to obtain the optimal prediction accuracy of the recommended model. Other similar algorithm steps are the same as above. Secondly, a weighted fusion model of the three models based on the GBDT-XGBoost-LightGBM algorithm is constructed. The weighted fusion results of the two models are fused by the mean average or median average method. The AUC indicators obtained by the two methods are compared, and the three-model fusion method with the largest AUC is selected. Finally, the corresponding model weights of the optimal results are determined by comparing the AUC change curves of different fusion models.

[0104] Secondly, the Stacking fusion strategy is implemented, and the processed data set is divided into a training set and a test set, of which 70% is used as a training set and 30% is used as a test set; on the training set, the grid search plus cross-validation method is used to tune the hyperparameters of the base learners of the first layer in Stacking to obtain the optimal hyperparameter combination of each model; then different fusion models (GBDT-XGBoost, GBDT-LightGBM, XGboost-LightGBM, GBDT-XGBoost-LightGBM) are constructed using the weighted fusion method mentioned above as the first-layer base learners, and the first-layer model combination is trained using a 4-fold cross-validation method on the training set, 3 of which are used as training sets, and the remaining 1 is used as a test set for prediction, and this process is repeated until the prediction results of the 4 base learners are combined in a predetermined order as the training set for the second-layer meta-learner. At the same time, the test set is predicted by the four base learners in the first layer, and the average of the predicted samples is taken. Then the predicted samples are combined as the test set of the second layer meta-learner; finally, the newly generated training set of each base learner in the first layer of Stacking is applied to the second layer meta-learner and trained using a logistic regression model to obtain the underlying asset splitting model of funds and wealth management products based on the Stacking weighted fusion strategy.

[0105] According to the above training process, financial product samples that meet different risk levels are used to select and train to obtain R1 sub-model, R2 sub-model and R3 sub-model. For example, financial product samples can be stored in the form of <user feature set, financial product name, underlying asset ratio, risk level>.

[0106] Finally, the test set is input into the asset prediction sub-model based on the Stacking weighted fusion strategy to output the user's underlying asset ratio. According to the user's risk level, funds and wealth management products with a ratio within the underlying asset range of the corresponding product (ratio deviation range is 0-10%) are selected, and the top five products in sales are recommended to customers based on the purchase situation in the past month.

[0107] Fig. 9 The flowchart of the silent client activation method according to the embodiment of the present disclosure is schematically shown.

[0108] like Fig. 9 As shown, this embodiment includes:

[0109] In operation S910, a silent client set is determined;

[0110] Silent customers can be identified from various dimensions, such as the activity level of the corresponding dimension in a specific time period, such as online channel dimension, transaction dimension and financial product dimension.

[0111] In terms of online channels, they are divided into silent customers on the bank’s platform and silent customers on the off-bank platform, which are determined based on whether transactions are made on online channels within a specific period of time. It mainly targets silent customers of commercial banks’ mobile banking, with a narrow coverage. It not only fails to pay attention to other credit card life platforms within the bank, but also ignores other social media platforms outside the bank. In recent years, the commercial banks’ on-bank platforms have been gradually restricted in terms of customer acquisition, while off-bank platforms have considerable advantages in terms of rich promotional methods and a wide range of audiences. Therefore, commercial banks should pay more attention to the off-bank ecology, such as WeChat public accounts, Douyin official accounts and other platforms.

[0112] In terms of transactions, silent customers are identified based on whether they have any transactions in a specific period of time. Such customers have a certain amount of funds retained in the bank, have a certain sense of identity and trust in the bank, and have high potential value after activation.

[0113] In the dimension of financial products, it can be determined based on the lack of purchase behavior within a specific time period, such as fixed investment or purchase of new products. For example, as two advanced investment products, funds and wealth management products have a large number of corresponding silent customers in commercial banks. If the silent activation of funds and wealth management products can be achieved, the customer's gold upgrade effect can be better achieved, which will greatly increase the intermediate income of commercial banks. For this type of silent customers, we must first rely on the situation where the customer has completed the risk assessment. Therefore, this disclosure further divides the silent customers of financial products into those who have not completed the risk assessment and those who have completed the risk assessment.

[0114] For silent customers of financial products who have not completed reputation assessment, the first thing to do is to have marketing customers complete risk assessment. For customers who have completed the assessment, the previous method of directly recommending specific products is not applicable to silent customers, because silent customers lack relevant product purchase records, and it is impossible to judge the customer's product preferences. It is also difficult to find other customers with similar preferences, and traditional collaborative filtering models based on customers or products cannot work. Therefore, for silent customers of financial products with different reputation levels, different machine learning models are established from the perspective of predicting the proportion of underlying assets.

[0115] In operation S920, at least one silent customer level to which each user belongs is determined based on transaction information of each user in the silent customer set;

[0116] Determine at least one silent customer level for the specific behaviors of each of the above dimensions within a specific time period;

[0117] In operation S930, for different silent customer levels, users in the levels are activated in sequence according to different orders and using different activation strategies;

[0118] The silent customer level includes the silent financial product level. For users in the silent financial product level, any one or more embodiments of the above financial product recommendation method are executed to construct an activation strategy.

[0119] According to the embodiments of the present disclosure, the hierarchical activation strategy can more accurately identify and respond to the needs and preferences of different silent customer groups. The layer-by-layer activation method allows financial institutions to intelligently allocate marketing resources based on the importance and potential value of customer levels. By implementing differentiated activation strategies and personalized services, the customer experience is significantly improved, the specific needs of customers are met, and the efficiency of activation work is improved.

[0120] Silent customer levels also include online channel silent levels and transaction silent levels. For silent customers in a single level, adopt corresponding activation strategies. As follows.

[0121] Silent customers in online channels can be divided into silent customers on in-bank platforms and silent customers on off-bank platforms. In-bank platforms can be divided into mobile banking, credit card life platforms, etc.; off-bank platforms can be divided into third-party platforms such as commercial bank WeChat public accounts and Douyin official accounts. The two types of silent customers in online channels can be further subdivided according to the silence time into: silent customers who have not logged in for the past month but have logged in for the past three months, silent customers who have not logged in for the past four months but have logged in for the past seven months, silent customers who have not logged in for the past eight months but have logged in for the past 12 months, and silent customers who have not logged in for the past 12 months or more.

[0122] For this type of customers, we choose to activate them in the corresponding silent channels. The activation strategy is: after logging in to the corresponding platform, they are given the opportunity to draw coupons. If they are silent in both the bank's internal and external channels, after logging in to the bank's internal platform, they are given the opportunity to log in to the external platform to draw coupons. For silent customers with four different silent periods divided by time above, the shorter the silent period, the greater the probability of the customer being activated. Therefore, for customers with short silent periods, we give them greater discounts and conduct marketing in layers.

[0123] Silent transaction customers can be divided into silent debit card transaction customers and silent credit card transaction customers. The two types of silent transaction customers can be further divided into silent customers who have not made transactions in the past 1 month but have traded in the past 3 months, silent customers who have not made transactions in the past 4 months but have traded in the past 7 months, silent customers who have not made transactions in the past 8 months but have traded in the past 12 months, and silent customers who have not made transactions in the past 12 months or more.

[0124] For this type of customers, we choose to activate them in the corresponding silent medium, and the activation strategy is to give them online payment coupons for the corresponding medium. For the silent customers with four different silent periods divided above, the shorter the silent period, the greater the probability of the customer being activated. Therefore, for customers with short silent periods, we give them greater discounts and conduct marketing in layers.

[0125] Financial product silent customers can determine one or more target financial products for recommendation according to one or more embodiments corresponding to the above financial product recommendation method.

[0126] It can be seen that there are multiple levels of silent customer stratification methods. For example, according to the marketing success rate in previous practices, the activation priority is online channel silence > transaction silence > financial product silence. In operation S930, they are activated in this order. In addition, different customers may become silent customers of other levels after marketing, so it is possible to monitor the dynamic changes in the levels of different customers in real time. After the silent customers are converted to silent customers of a higher level, the activation strategy of the customers can be changed in time to achieve dynamic updates of the activation strategy. For example, operations S910 to S930 can be executed cyclically in a timed or actively triggered manner.

[0127] For complex silent customers with at least two levels, corresponding activation strategies will also be adopted, as follows.

[0128] For customers who are silent in online channels and silent in transactions, we first market them as silent customers in online channels. After the corresponding channels are activated, we continue to recommend debit card or credit card payment discounts in online channels, and continue to complete transaction activation on the basis of maintaining the active online channels.

[0129] For customers who are silent on online channels and financial products, we first market them as silent customers on online channels. After the corresponding channels are activated, we continue to recommend the corresponding products through the models recommended by the online channels, and continue to complete product activation on the basis of maintaining the activity of the online channels.

[0130] For customers who are silent on transactions or financial products, we first market to them as silent customers. After the corresponding media is activated, we continue to recommend corresponding products on the payment success interface.

[0131] For customers who are silent on online channels, transactions and financial products, after recommending the silent channels, we will follow up with recommendations for corresponding payment discounts and then recommend corresponding financial products.

[0132] According to the embodiments of the present disclosure, silent customers are first divided into three levels of single silent customers with gradually increasing silence levels, namely online channel silent customers, transaction silent customers, and financial product silent customers, as well as complex silent customers that are cross-linked with single silent customers. Through expert rules and machine learning models, combined with multi-round contact strategies, silent customers are activated layer by layer. After silent customers at different levels complete the level conversion, the converted customers are monitored in real time and the marketing strategy is dynamically changed, so as to realize the multi-dimensional activation of commercial banks from online channels to transactions and then to products, and promote the overall improvement of banking business.

[0133] Based on the above financial product recommendation method, the present disclosure also provides a financial product recommendation device. Fig.10 The device is described in detail.

[0134] Fig.10 The structural block diagram of the financial product recommendation device according to an embodiment of the present disclosure is schematically shown.

[0135] like Fig.10 As shown, the financial product recommendation device 1000 of this embodiment includes an input module 1010 , a prediction module 1020 , a matching module 1040 and a recommendation module 1050 .

[0136] The input module 1010 may perform operation S210, which is used to input the user characteristics of the user into a pre-trained asset prediction model, wherein the asset prediction model is trained based on financial product samples, and the financial product samples include information on the proportion of various underlying assets;

[0137] The prediction module 1020 may perform operation S220 to obtain N types of predicted underlying assets output by the asset prediction model and the proportion information of each type of predicted underlying assets;

[0138] The pre-screening module 1030 may perform operation S230, for querying one or more candidate financial products matching the risk assessment level of the user from a database to form a candidate product set, wherein information of candidate financial products corresponding to different risk assessment levels is pre-stored in the database;

[0139] The matching module 1040 may perform operation S240, for matching M target financial products from a candidate product set based on N types of predicted underlying assets and the proportion information of each type of predicted underlying assets, where the candidate product set includes at least one candidate product and the proportion information of each type of underlying assets of each candidate product, where N and M are both integers greater than or equal to 1;

[0140] The recommendation module 1050 may perform operation S250 for recommending M target financial products to the user.

[0141] In some embodiments, the financial product recommendation device 1000 may further include a candidate product determination module, which may perform operations S310 to S320, which will not be described in detail herein.

[0142] In some embodiments, the prediction module 1020 may perform operations S810 to S830, which will not be described in detail herein.

[0143] Based on the above silent customer activation method, the present disclosure also provides a silent customer activation device. Fig.11 The device is described in detail.

[0144] Fig.11 The structural block diagram of the financial product recommendation device according to an embodiment of the present disclosure is schematically shown.

[0145] like Fig.11 As shown, the silent client activation device 1100 of this embodiment includes a first determination module 1110 , a second determination module 1120 and a silent client activation module 1130 .

[0146] The first determining module 1110 may perform operation S910 for determining a silent client set;

[0147] The second determining module 1120 may perform operation S920 for determining at least one silent customer level to which each user belongs based on transaction information of each user in the silent customer set;

[0148] The silent customer activation module 1130 may perform operation S930 for activating users in different silent customer levels in sequence according to different orders and using different activation strategies;

[0149] The silent customer level includes the silent financial product level. For users in the silent financial product level, any one or more embodiments of the above financial product recommendation method are executed to construct an activation strategy.

[0150] For the parts not mentioned in the device part, they can be understood with reference to the various embodiments of the above method. That is, the device part includes modules for executing the various steps of any one of the method embodiments described above. In addition, the implementation methods, technical problems solved, functions implemented, and technical effects achieved of each module / unit / subunit, etc. in the device part embodiment are respectively the same or similar to the implementation methods, technical problems solved, functions implemented, and technical effects achieved of each corresponding step in the method part embodiment, and will not be repeated here.

[0151] According to an embodiment of the present disclosure, any multiple modules in the financial product recommendation device 1000 or the silent customer activation device 1100 can be combined into one module for implementation, or any one of the modules can be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules can be combined with at least part of the functions of other modules and implemented in one module.

[0152] According to an embodiment of the present disclosure, at least one of the financial product recommendation device 1000 or the silent customer activation device 1100 may be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application specific integrated circuit (ASIC), or may be implemented by hardware or firmware in any other reasonable manner of integrating or packaging the circuit, or implemented in any one of the three implementation methods of software, hardware and firmware or in an appropriate combination of any of them. Alternatively, at least one of the financial product recommendation device 1000 or the silent customer activation device 1100 may be at least partially implemented as a computer program module, and when the computer program module is executed, the corresponding function may be executed.

[0153] Fig.12 A block diagram of an electronic device suitable for implementing a financial product recommendation method or a silent customer activation method according to an embodiment of the present disclosure is schematically shown.

[0154] like Fig.12 As shown, the electronic device 1200 according to an embodiment of the present disclosure includes a processor 1201, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1202 or a program loaded from a storage part 1208 to a random access memory (RAM) 1203. The processor 1201 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or a related chipset and / or a dedicated microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 1201 may also include an onboard memory for caching purposes. The processor 1201 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present disclosure.

[0155] In RAM 1203, various programs and data required for the operation of electronic device 1200 are stored. Processor 1201, ROM 1202 and RAM 1203 are connected to each other via bus 1204. Processor 1201 performs various operations of the method flow according to the embodiment of the present disclosure by executing the program in ROM 1202 and / or RAM 1203. It should be noted that the program can also be stored in one or more memories other than ROM 1202 and RAM 1203. Processor 1201 can also perform various operations of the method flow according to the embodiment of the present disclosure by executing the program stored in one or more memories.

[0156] According to an embodiment of the present disclosure, the electronic device 1200 may further include an input / output (I / O) interface 1205, which is also connected to the bus 1204. The electronic device 1200 may further include one or more of the following components connected to the I / O interface 1205: an input portion 1206 including a keyboard, a mouse, etc.; an output portion 1207 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage portion 1208 including a hard disk, etc.; and a communication portion 1209 including a network interface card such as a LAN card, a modem, etc. The communication portion 1209 performs communication processing via a network such as the Internet. A drive 1210 is also connected to the I / O interface 1205 as needed. A removable medium 1211, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 1210 as needed, so that a computer program read therefrom is installed into the storage portion 1208 as needed.

[0157] The present disclosure also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or may exist independently without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the method according to the embodiment of the present disclosure is implemented.

[0158] According to an embodiment of the present disclosure, a computer-readable storage medium may be a non-volatile computer-readable storage medium, for example, may include but is not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, an apparatus, or a device. For example, according to an embodiment of the present disclosure, a computer-readable storage medium may include the ROM 1202 and / or RAM 1203 described above and / or one or more memories other than ROM 1202 and RAM 1203.

[0159] The embodiment of the present disclosure also includes a computer program product, which includes a computer program, and the computer program contains program code for executing the method shown in the flowchart. When the computer program product is run in a computer system, the program code is used to enable the computer system to implement the method provided by the embodiment of the present disclosure.

[0160] The above functions defined in the system / device of the embodiment of the present disclosure are performed when the computer program is executed by the processor 1201. According to the embodiment of the present disclosure, the system, device, module, unit, etc. described above can be implemented by a computer program module.

[0161] In one embodiment, the computer program may rely on tangible storage media such as optical storage devices, magnetic storage devices, etc. In another embodiment, the computer program may also be transmitted and distributed in the form of signals on a network medium, and downloaded and installed through the communication part 1209, and / or installed from the removable medium 1211. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0162] In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 1209, and / or installed from the removable medium 1211. When the computer program is executed by the processor 1201, the above functions defined in the system of the embodiment of the present disclosure are performed. According to the embodiment of the present disclosure, the system, device, means, module, unit, etc. described above can be implemented by a computer program module.

[0163] According to an embodiment of the present disclosure, the program code for executing the computer program provided by the embodiment of the present disclosure can be written in any combination of one or more programming languages. Specifically, these computing programs can be implemented using high-level process and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, Java, C++, python, "C" language or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on the remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, using an Internet service provider to connect through the Internet).

[0164] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0165] It will be appreciated by those skilled in the art that the features described in the various embodiments and / or claims of the present disclosure may be combined and / or combined in a variety of ways, even if such combinations and / or combinations are not explicitly described in the present disclosure. In particular, the features described in the various embodiments and / or claims of the present disclosure may be combined and / or combined in a variety of ways without departing from the spirit and teachings of the present disclosure. All of these combinations and / or combinations fall within the scope of the present disclosure.

[0166] The embodiments of the present disclosure are described above. However, these embodiments are only for illustrative purposes and are not intended to limit the scope of the present disclosure. Although the embodiments are described above separately, this does not mean that the measures in the various embodiments cannot be used in combination to advantage. The scope of the present disclosure is defined by the attached claims and their equivalents. Without departing from the scope of the present disclosure, those skilled in the art may make a variety of substitutions and modifications, which should all fall within the scope of the present disclosure.

Claims

1. A financial product recommendation method, characterized in that: include: Inputting the user characteristics of the user into a pre-trained asset prediction model, wherein the asset prediction model is trained based on a financial product sample, and the financial product sample includes information on the proportion of various underlying assets; Obtaining N types of predicted underlying assets output by the asset prediction model and the proportion information of each type of predicted underlying assets; Based on the risk assessment level of the user, querying one or more candidate financial products matching the risk assessment level from a database to form a candidate product set, wherein information of candidate financial products corresponding to different risk assessment levels is pre-stored in the database; Based on the N types of predicted underlying assets and the proportion information of each type of predicted underlying assets, M target financial products are matched from the candidate product set, the candidate product set includes at least one candidate product and the proportion information of each type of underlying assets of each candidate product, and N and M are both integers greater than or equal to 1; The M target financial products are recommended to the user.

2. The method according to claim 1, characterized in that Based on the N types of predicted underlying assets and the proportion information of each type of predicted underlying assets, M target financial products are matched from the candidate product set, including: Mapping the N types of predicted underlying assets to the underlying asset categories in each candidate product; For a candidate product having a preset number of categories among the N categories of predicted underlying assets, compare the proportion information of each category of underlying assets in the candidate product with the proportion information of each category of the N categories of predicted underlying assets to obtain a proportion information comparison result; When the comparison result of the proportion information is characterized as similar or identical, it is determined as the target financial product.

3. The method according to claim 1, characterized in that The asset prediction model includes K asset prediction sub-models corresponding to K risk levels one by one, K is greater than or equal to 1, and the inputting of the user characteristics of the user into the pre-trained asset prediction model includes: Inputting the user characteristics of the user into the asset prediction sub-model corresponding to the risk assessment level; Among them, different asset prediction sub-models are trained using samples of financial products that meet different risk levels.

4. The method according to claim 3, characterized in that The risk level includes R1 level, and the K asset prediction sub-models include an R1 level sub-model. When the risk assessment level is R1 level, the N types of predicted underlying assets output by the asset prediction model and the proportion information of each type of predicted underlying assets include: The respective proportion information of the money market underlying assets and the bond underlying assets output by the R1-level sub-model is obtained.

5. The method according to claim 3, characterized in that: The risk level includes R2 level, and the K asset prediction sub-models include an R2 level sub-model. When the risk assessment level is R2 level, the N types of predicted underlying assets output by the asset prediction model and the proportion information of each type of predicted underlying assets are obtained, including: The respective proportion information of the money market underlying assets, bond underlying assets and equity underlying assets output by the R2-level sub-model is obtained.

6. The method according to claim 3, characterized in that The risk level includes R3 and above, the K asset prediction sub-models include an R3 sub-model, and when the risk assessment level is R3 or above, the N types of predicted underlying assets output by the asset prediction model and the proportion of each type of predicted underlying assets are obtained, including: The respective proportion information of the money market underlying assets, bond underlying assets, equity underlying assets and derivative underlying assets output by the R3-level sub-model is obtained.

7. The method according to claim 3, characterized in that Each of the asset prediction sub-models includes a meta-learner and a plurality of base learners, and inputting the user characteristics of the user into the asset prediction sub-model corresponding to the risk assessment level includes: Calling the multiple base learners of the asset prediction sub-model, so that the multiple base learners process the user features of the user respectively, and output multiple first prediction results in one-to-one correspondence; Performing weighted calculation based on the multiple intermediate prediction results and respective weight coefficients to obtain multiple second prediction results; The multiple second prediction results are input into the meta-learner, so that the meta-learner processes the multiple second prediction results, wherein the meta-learner is used to output the N types of predicted underlying assets and the proportion information of each type of predicted underlying assets.

8. A silent customer activation method, characterized in that: include: Identify the silent customer set; Determining at least one silent customer level to which each user belongs based on transaction information of each user in the silent customer set; For different silent customer levels, activate users in each level in different orders and using different activation strategies; The silent customer level includes a silent financial product level. For users of the silent financial product level, the method described in any one of claims 1 to 7 is executed to construct an activation strategy.

9. A financial product recommendation device, characterized in that: include: An input module, used to input the user characteristics of the user into a pre-trained asset prediction model, wherein the asset prediction model is trained based on financial product samples, and the financial product samples include information on the proportion of various underlying assets; A prediction module, used to obtain N types of predicted underlying assets output by the asset prediction model and the proportion information of each type of predicted underlying assets; A pre-screening module, configured to query a database for one or more candidate financial products matching the risk assessment level of the user based on the risk assessment level of the user, to form a candidate product set, wherein information of candidate financial products corresponding to different risk assessment levels is pre-stored in the database; A matching module, configured to match M target financial products from the candidate product set based on the N types of predicted underlying assets and the proportion information of each type of predicted underlying assets, wherein the candidate product set includes at least one candidate product and the proportion information of each type of underlying assets of each candidate product, and N and M are both integers greater than or equal to 1; A recommendation module is used to recommend the M target financial products to the user.

10. A silent customer activation device, characterized in that: include: A first determination module is used to determine a set of silent customers; A second determination module, configured to determine at least one silent customer level to which each user belongs based on transaction information of each user in the silent customer set; The silent customer activation module is used to activate users in different silent customer levels in different orders and using different activation strategies; The silent customer level includes a silent financial product level. For users of the silent financial product level, the method described in any one of claims 1 to 7 is executed to construct an activation strategy.

11. An electronic device, comprising: one or more processors; a memory for storing one or more computer programs, It is characterized in that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 8.

12. A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method according to any one of claims 1 to 8.

13. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images