Asset management product recommendation method and related device

Abstract

The invention relates to the technical field of computers, and provides an asset management product recommendation method and a related device, which are used for improving the asset security of a target object, and the method comprises the steps: carrying out the resource evaluation of each candidate management product based on the total resource investment of the target object in combination with an exchange resource calculation mode corresponding to each candidate management product, obtaining exchange resource differences among the candidate management products; and then, based on the exchange resource difference, in combination with the object information of the target object, selecting a recommended target management product from the candidate management products. The target management product is determined by using the exchange resource difference, so that the recommended product can better meet the demand of the object, and the asset loss risk caused by selecting the unmatched target management product can be reduced.

Description

Background Art

[0002] With the development of Internet technology, more and more objects are managing assets based on asset management products. Asset management products can be funds, stocks, etc., and asset management products can also be simply referred to as management products.

[0003] In the related art, the product information of each candidate management product is usually provided to the target object to assist the target object in selecting the target management product from the candidate management products.

[0004] However, the selection of target management products depends on the product selection experience of the target object. If the target object has little experience in product selection, the selected target management product may not be suitable for the target object's asset management method, thereby causing asset losses to the target object. Even if the target object has rich experience in product selection, different candidate management products require different exchange resources (exchange resources refer to the resources used when exchanging the product ownership of the candidate management products), and the calculation method of exchange resources is relatively complicated. Therefore, in the case of a large number of candidate management products, the target object may still select an inappropriate target management product, thereby bringing the risk of asset loss. Summary of the invention

[0005] The embodiments of the present application provide an asset management product recommendation method and related devices to ensure the asset security of a target object.

[0006] On the one hand, an embodiment of the present application provides an asset management product recommendation method, comprising:

[0007] Get the total amount of resource input of the target object;

[0008] Based on the total amount of resource input and in combination with the exchange resource calculation method corresponding to each candidate management product, resource evaluation is performed on each candidate management product to obtain the exchange resource difference between the candidate management products;

[0009] Based on the exchange resource difference and in combination with the object information of the target object, a target management product recommended to the target object is selected from the candidate management products.

[0010] On the one hand, an embodiment of the present application provides an asset management product recommendation method, comprising:

[0011] The total amount input interface is presented;

[0012] In response to the input operation triggered by the target object in the total amount input interface, recommended information of the target management product is presented. The target management product is determined based on the exchange resource differences between the candidate management products. The exchange resource differences are determined based on the input total amount of resource input and the exchange resource calculation method corresponding to each candidate management product.

[0013] On the one hand, an embodiment of the present application provides an asset management product recommendation device, including:

[0014] An information acquisition unit, used to acquire the total amount of resource input of the target object;

[0015] An information analysis unit, configured to perform resource evaluation on each candidate management product based on the total resource input and in combination with the exchange resource calculation method corresponding to each candidate management product, to obtain the exchange resource difference between the candidate management products;

[0016] The information recommendation unit is used to select a target management product recommended to the target object from the candidate management products based on the exchange resource difference and in combination with the object information of the target object.

[0017] In a possible implementation, when performing resource evaluation on each candidate management product based on the total resource input and the exchange resource calculation method corresponding to each candidate management product to obtain the exchange resource difference between the candidate management products, the information analysis unit is specifically used to:

[0018] Based on the total amount of resource input and in combination with the exchange resource calculation method corresponding to each candidate management product, resource evaluation is performed on each candidate management product to obtain a reference duration, wherein the reference duration represents: the product holding duration required when the exchange resources of each candidate management product are equal;

[0019] Based on the reference duration, the exchange resource difference is obtained.

[0020] In a possible implementation, the information analysis unit is further used for:

[0021] Based on the target holding time of the target object and the total amount of resource investment, combined with various exchange resource calculation methods, obtaining the exchange resources of each candidate management product within the target holding time;

[0022] The exchange resource difference is obtained based on the reference duration and in combination with the obtained resource difference between the exchange resources.

[0023] In a possible implementation, the information recommendation unit is further used to:

[0024] Based on the target management product and the exchange resource difference, generating recommendation information, wherein the recommendation information is represented in the form of a tag;

[0025] The recommendation information is sent to a terminal device, so that the terminal device presents the recommendation information to the target object.

[0026] In a possible implementation, if the object information includes: each key preference information corresponding to the target object and the information sequence therebetween, the information sequence is used to represent the importance of each key preference information;

[0027] When selecting a target management product recommended to the target object from the candidate management products based on the exchange resource difference and in combination with the object information of the target object, the information recommendation unit is specifically used to:

[0028] Based on the exchange resource differences, combined with the key preference information and the information order therebetween, obtaining matching evaluation information between each of the candidate management products and the target object;

[0029] Based on the obtained matching evaluation information, a candidate management product that meets the set matching condition is selected from the candidate management products as the target management product.

[0030] In a possible implementation, the information recommendation unit is further used to:

[0031] If the custom preference information of the target object is set, obtaining the object information based on the custom preference information;

[0032] If the customized preference information is not set, preference inference is performed based on the historical product usage data and historical product feedback data of the target object to obtain the object information.

[0033] In a possible implementation, when the preference inference is performed based on the historical product usage data and historical product feedback data of the target object and the object information is obtained, the information recommendation unit is specifically used to:

[0034] Based on the historical product feedback data, obtaining the explicit preference characteristics of the target object, and based on the historical product usage data, obtaining the implicit preference characteristics of the target object;

[0035] Based on the explicit preference feature and the implicit preference feature, combined with preset various aspects of labels, the object information is obtained.

[0036] In a possible implementation, when obtaining the object information based on the explicit preference feature and the implicit preference feature in combination with various preset labels, the information recommendation unit is specifically configured to:

[0037] Based on the explicit preference feature and the implicit preference feature, obtain a label evaluation value between each of the aspect labels and the target object, and based on the obtained label evaluation values, screen out an aspect label sequence including k aspect labels from the aspect labels;

[0038] Based on the semantic information corresponding to each of the k aspect labels, corresponding key preference information is generated, and based on the label order of the K aspect labels in the aspect label sequence, the information order between the generated key preference information is obtained.

[0039] In a possible implementation, when obtaining the explicit preference characteristics of the target object based on the historical product feedback data, the information recommendation unit is specifically configured to:

[0040] Extracting each word from the historical product feedback data, and encoding each word to obtain word features corresponding to each word;

[0041] Based on the association between the aspect labels and the words, information aggregation is performed in combination with the word features corresponding to each of the words to obtain the aspect label features corresponding to each of the aspect labels;

[0042] Based on the association between the product and the aspect label, information aggregation is performed in combination with the aspect label features corresponding to each aspect label to obtain the product features corresponding to each historical management product in the historical product feedback data;

[0043] Based on the association relationship between products and objects, information aggregation is performed in combination with the product features corresponding to each of the historical management products to obtain the explicit preference features of the target object.

[0044] In a possible implementation, when performing resource evaluation on each candidate management product based on the total resource input and the exchange resource calculation method corresponding to each candidate management product to obtain the exchange resource difference between the candidate management products, the information analysis unit is specifically used to:

[0045] When the total amount of resource investment reaches the investment threshold set for the target object, resource evaluation is performed based on the total amount of resource investment and each exchange resource calculation method to obtain the exchange resource difference; wherein the investment threshold is the first threshold or the second threshold, and the first threshold is higher than the second threshold.

[0046] On the one hand, an embodiment of the present application provides an asset management product recommendation device, including:

[0047] A first presentation unit, used to present a total amount input interface;

[0048] The second presentation unit is used to present recommendation information of the target management product in response to the input operation triggered by the target object in the total amount input interface. The target management product is determined based on the exchange resource differences between the candidate management products. The exchange resource differences are determined based on the input total amount of resource input and the exchange resource calculation method corresponding to each candidate management product.

[0049] In a possible implementation, when presenting the total amount input interface, the first presentation unit is specifically configured to:

[0050] Presenting a total amount input interface for a designated management product, where the designated management product is a candidate management product among the candidate management products except the target management product;

[0051] After presenting the recommended information of the target management product in the total amount input interface, the second presenting unit is further used to:

[0052] In response to the difference viewing operation triggered by the target object, a difference viewing interface is presented, the difference viewing interface including: exchange resource difference and product switching components;

[0053] In response to the triggering operation of the target object on the product switching component, a total amount input interface for the target management product is presented.

[0054] In a possible implementation, the second presentation unit is further configured to:

[0055] In response to a product information viewing operation triggered by the target object for the target recommended product, presenting a product information interface of the target recommended product; the product information interface includes a product comparison component;

[0056] In response to a product information viewing operation triggered by the target object for the product comparison component, the exchange resource difference is presented in the product information interface.

[0057] On the one hand, an electronic device is provided, comprising a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the above method.

[0058] In one aspect, a computer-readable storage medium is provided, which includes a computer program. When the computer program is run on an electronic device, the computer program is used to enable the electronic device to execute the steps of any of the above methods.

[0059] On the one hand, a computer program product is provided, which includes a computer program stored in a computer-readable storage medium. A processor of an electronic device reads and executes the computer program from the computer-readable storage medium, so that the electronic device performs the steps of any of the above methods.

[0060] In the embodiments of the present application, firstly, the target object is assisted in asset management by recommending target management products. In this way, product selection is not dependent on the target object's product selection experience, thereby reducing the risk of asset loss caused by selecting inappropriate target management products.

[0061] Secondly, the target management products are selected based on the exchange resource differences. Through the exchange resource differences, a clear decision-making basis is provided to the target users, helping them to predict and select cost-effective management products, thereby reducing the possibility of asset loss and protecting the asset security of the target users.

[0062] Finally, based on the total amount of resource input of the target object and the exchange resource calculation method of each candidate management product, the exchange resource difference between the candidate management products is calculated. Since the exchange resource calculation method is usually more complicated and usually changes over time, resource evaluation based on the total amount of resource input of the target object can not only improve the efficiency of product recommendations, but also improve the quality of product recommendations, making the recommended products more in line with the needs of the object, thereby improving the user experience.

[0063] Other features and advantages of the present application will be described in the following description, and partly become apparent from the description, or be understood by practicing the present application. The purpose and other advantages of the present application can be realized and obtained by the structures specifically pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0065] Figure 1 A schematic diagram of an application scenario provided in an embodiment of the present application;

[0066] Figure 2 A flowchart of the asset management product recommendation method provided in an embodiment of the present application;

[0067] Figure 3 A schematic diagram of the breakeven points of Class A shares and Class C shares provided in the embodiments of the present application;

[0068] Figure 4 A schematic diagram of the HHRG provided in the examples of the present application;

[0069] Figure 5 This is a schematic diagram of the structure of the PRT provided in the embodiments of the present application;

[0070] Figure 6 A schematic diagram of a process for obtaining object information provided in an embodiment of the present application;

[0071] Figure 7 A flowchart of a method for obtaining object information provided in an embodiment of the present application;

[0072] Figure 8 A schematic diagram of the structure of the label generation model provided in the embodiments of the present application;

[0073] Fig. 9 A schematic diagram of the flow of the model training method provided in the embodiments of the present application;

[0074] Fig.10 A schematic diagram of a WAPU meta-path and a WAU meta-path provided in an embodiment of the present application;

[0075] Fig.11 A logical diagram of the model training process provided in the embodiments of the present application;

[0076] Fig.12 This is a schematic diagram of the first interface provided in the embodiment of the present application;

[0077] Fig.13 This is a schematic diagram of the second interface provided in the embodiment of the present application;

[0078] Fig.14 A logical diagram of the asset management recommendation process provided in an embodiment of the present application;

[0079] Fig.15 A schematic diagram of the structure of an asset management product recommendation device provided in an embodiment of the present application;

[0080] Fig.16 A schematic diagram of the structure of an asset management product recommendation device provided in an embodiment of the present application;

[0081] Fig.17 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0082] In order to make the purpose, technical scheme and advantages of the present application clearer, the technical scheme in the embodiment of the present application will be clearly and completely described below in conjunction with the drawings in the embodiment of the present application. Obviously, the described embodiment is only a part of the embodiment of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application. In the absence of conflict, the embodiments in the present application and the features in the embodiments can be combined with each other arbitrarily. In addition, although the logical order is shown in the flow chart, in some cases, the steps shown or described can be performed in an order different from that here.

[0083] It will be understood that in the following specific implementations of the present application, it may involve obtaining user-related data such as historical product feedback data and historical behavior data. When the various embodiments of the present application are applied to specific products or technologies, relevant licenses or consents need to be obtained, and the collection, use and processing of relevant data need to comply with relevant laws, regulations and standards of relevant countries and regions.

[0084] In the related art, the product information of each candidate management product is usually provided to the target object to assist the target object in selecting the target management product from the candidate management products.

[0085] However, the selection of target management products depends on the product selection experience of the target object. If the target object has little experience in product selection, the selected target management product may not be suitable for the target object's asset management method, thereby causing asset losses to the target object. Even if the target object has rich experience in product selection, different candidate management products require different exchange resources, and the calculation method of exchange resources is relatively complicated. Therefore, in the case of a large number of candidate management products, the target object may still select an inappropriate target management product, thereby bringing the risk of asset loss.

[0086] In the embodiments of the present application, firstly, the target object is assisted in asset management by recommending target management products. In this way, product selection is not dependent on the target object's product selection experience, thereby reducing the risk of asset loss caused by selecting inappropriate target management products.

[0087] Secondly, the target management products are selected based on the exchange resource differences. Through the exchange resource differences, a clear decision-making basis is provided to the target users, helping them to predict and select cost-effective management products, thereby reducing the possibility of asset loss and protecting the asset security of the target users.

[0088] Finally, based on the total amount of resource input of the target object and the exchange resource calculation method of each candidate management product, the exchange resource difference between the candidate management products is calculated. Since the exchange resource calculation method is usually more complicated and usually changes over time, resource evaluation based on the total amount of resource input of the target object can not only improve the efficiency of product recommendation, but also improve the quality of product recommendation, so that the recommended products are more in line with the needs of the object, thereby improving the user experience. In addition, the exchange resource calculation method usually changes over time. The recommendation process not only focuses on current functional requirements, but also considers future scalability and maintainability, ensuring the long-term effectiveness and adaptability of product recommendation technology.

[0089] The following briefly introduces the application scenarios to which the technical solutions of the embodiments of the present application can be applied. It should be noted that the application scenarios introduced below are only used to illustrate the embodiments of the present application and are not limited. In the specific implementation process, the technical solutions provided by the embodiments of the present application can be flexibly applied according to actual needs.

[0090] See also Figure 1 As shown, it is a schematic diagram of an application scenario provided by an embodiment of the present application. In this scenario, a terminal device 101 and a server 102 may be included.

[0091] The terminal device 101 may be, for example, a mobile phone, a tablet computer, a laptop computer, a desktop computer, a smart TV, a smart vehicle-mounted device, a smart wearable device, a smart TV, an aircraft, or any other device involved in asset management product recommendations.

[0092] The terminal device 101 may be installed with a client or plug-in for asset management. The client may be a software client, or a web page, a small program, or other client, and the specific type of the client is not limited. In the following, only the client is used as an example for explanation.

[0093] In some implementations, the terminal device 101 has front-end interactive capabilities. The terminal device 101 adopts a responsive design, adapts to a variety of devices and screen sizes, and can use front-end technologies such as HTML5, CSS3, and JavaScript.

[0094] In some implementations, interactive prompt windows and prompt windows may be presented in the terminal device 101. The interactive prompt windows and prompt windows may be implemented through CSS animation and JavaScript events to achieve dynamic display of interactive resource differences and front-end interaction effects.

[0095] In some implementations, the terminal device 101 uses Ajax technology to perform data interaction with the server 102, so that real-time exchange of resource differences can be achieved without refreshing the page.

[0096] Server 102 is a backend server corresponding to the client. Server 102 can 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 databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms, but is not limited to these.

[0097] In some implementations, server 102 has sufficient computing power and storage space to support large-scale object access and data processing.

[0098] In some implementations, the server 102 may use cloud services or a distributed system architecture to ensure scalability and high availability of the system.

[0099] In some implementations, the server 102 uses a stable server-side language to process business logic and database interaction, such as Java, Python, etc.

[0100] In some implementations, the server 102 communicates data with the front end via a RESTful application programming interface (API) to provide services such as exchange resource calculation, resource intersection point prediction, and investment advice.

[0101] In some implementations, the server 102 is used to implement one or more functions such as object input parsing, resource computing algorithm execution, data query and response data packaging.

[0102] Both the server 102 and the terminal device 101 may include one or more processors, memories, and interactive I / O interfaces, etc. In addition, the server 102 may also be configured with a database, which may be used to store various historical product usage data, etc. Among them, the memories of the server 102 and the terminal device 101 may also store program instructions required to be executed in the asset management product recommendation method provided in the embodiment of the present application, and these program instructions, when executed by the processor, can be used to implement the asset management product recommendation process provided in the embodiment of the present application.

[0103] The terminal device 101 and the server 102 may be directly or indirectly connected to each other through one or more networks. The network may be a wired network or a wireless network. For example, the wireless network may be a mobile cellular network or a wireless fidelity (Wireless-Fidelity, WIFI) network. Of course, it may also be other possible networks, which are not limited in the embodiments of the present application. Of course, Figure 1 The figure is only an example. In fact, the number of terminal devices and servers is not limited and is not specifically limited in the embodiments of the present application.

[0104] In some implementations, object information and product information are stored in a database, such as a relational database management system PostgreSQL. By designing a reasonable database model, data consistency and integrity can be ensured. In addition, data access speed and system performance can be improved through the cache mechanism Redis.

[0105] In some implementations, the HTTPS protocol is used to encrypt data transmission between the terminal device 101 and the server 102 to protect the security of the object data.

[0106] In some implementations, the server 102 also has an identity authentication and authorization mechanism to ensure that only authorized objects have the right to access sensitive data.

[0107] In some implementations, the terminal device 101 presents a total amount input interface for a specified management product, and in response to an input operation triggered by a target object in the total amount input interface, presents recommended information of the target management product in the total amount input interface, and the recommended information of the target management product can be presented in the form of a prompt window, but is not limited thereto; thereafter, in response to a difference viewing operation triggered by the target object, presents a difference viewing interface, and the difference viewing interface includes: exchange resource differences. In addition, the difference viewing interface may also include: a product switching component, and the terminal device 101 may also present a total amount input interface for the target management product in response to a triggering operation of the target object on the product switching component.

[0108] The target management product can be determined by the server 102 according to the exchange resource difference between the candidate management products, and the exchange resource difference is determined according to the input total resource input and the exchange resource calculation method corresponding to each candidate management product. Each candidate management product includes a designated management product and a target management product.

[0109] Exemplarily, after the target object enters the total amount of resource investment in the total amount input interface for the designated management product, the terminal device 101 sends the total amount of resource investment entered by the target object to the server 102. The server 102 triggers the calculation of the exchange resource difference between the designated management product and other management products based on the total amount of resource investment entered by the target object, wherein the other management products may be other fund shares with the same name as the designated management product. After the server 102 calculates the exchange resource difference, it sends the exchange resource difference to the terminal device 101, so that the terminal device 101 presents the exchange resource difference to the target object in the prompt window. The prompt window includes a "detailed comparison" button. When the target object triggers the "detailed comparison" button, the terminal device responds to the difference viewing operation triggered by the target object and presents a difference viewing interface including the exchange resource difference. The exchange resource difference is presented in the form of a chart. The difference viewing interface may also include the exchange resource calculation method corresponding to the designated management product and other management products, that is, detailed cost information.

[0110] In some implementations, after the target object inputs the total amount of resource investment, the terminal device 101 sends the total amount of resource investment to the server 102 through a secure data transmission protocol. After receiving the total amount of resource investment, the server 102 calls the calculation service to obtain the exchange resource difference between each candidate management product, and generates recommendation information for the target management product based on the calculated exchange resource difference. The recommendation information can be expressed in the form of a label. The exchange resource difference and the recommendation information of the target management product are returned to the terminal device 101 through the API, and the terminal device 101 dynamically updates the front-end interface based on the returned data.

[0111] In some implementations, the terminal device 101 presents a product management interface corresponding to a management product combination, and the management product combination includes multiple management products; the terminal device 101 responds to the exchange resource comparison operation triggered by the target object for other management product combinations and management product combinations, and presents the exchange resource difference between other management product combinations and management product combinations. Among them, the other management product combination contains some different management products or all different management products from the management product combination. The other management product combination can be selected by the target object, or the system can select it from the historical management product combinations of the target object.

[0112] The difference in exchange resources between other management product combinations and management product combinations is determined by the server 1 according to the total amount of resource input and the exchange resource calculation method corresponding to each management product in other management product combinations and management product combinations.

[0113] Exemplarily, after the target object inputs the total amount of resource investment in the total amount input interface, the terminal device 101 sends the total amount of resource investment input by the target object to the server 102. The server 102 triggers the calculation of the exchange resource difference between other management product combinations and the management product combination according to the total amount of resource investment input by the target object. After the server 102 calculates the exchange resource difference, it sends the exchange resource difference to the terminal device 101, so that the terminal device 101 presents the exchange resource difference to the target object.

[0114] Currently, in actual application, specific configuration can be performed according to the situation, and this application does not make any specific limitations here.

[0115] See also Figure 2 As shown, it is a flow chart of an asset management product recommendation method provided in an embodiment of the present application. The process is applied to a server or a terminal device. Taking the application to a server as an example, the specific process is as follows:

[0116] S201. Obtain the total amount of resource input of the target object.

[0117] In the embodiment of the present application, the target object refers to the object that needs to be managed. The total amount of resource investment refers to the amount of resources transferred by the target object when purchasing the candidate management product.

[0118] In some implementations, the total amount of resource investment may be input by the target object, for example, the target object enters the total amount of resource investment in the total amount input interface to hold a certain candidate management product or multiple candidate management products. The total amount of resource investment may also be determined by the server based on the target object's historical total amount of resource investment. For example, the server uses the average of the target object's historical total amount of resource investment within a certain time range as the target object's total amount of resource investment.

[0119] S202: Based on the total amount of resource input and in combination with the exchange resource calculation method corresponding to each candidate management product, perform resource evaluation on each candidate management product to obtain the exchange resource difference between the candidate management products.

[0120] In the embodiments of the present application, only the candidate management product is taken as a fund as an example for illustration. In actual application, the candidate management product may also be other products used for asset management.

[0121] In a possible implementation, since the same fund can provide different shares for target objects to choose from, and different shares differ in at least one aspect such as the calculation method of exchange resources, the transfer time of exchange resources, and the holding period, a fund share can be used as a candidate management product. Exemplary share types include but are not limited to: Class A shares, Class B shares, Class C shares, and Class D shares, wherein Class A shares and Class C shares transfer exchange resources when purchased, Class B shares and Class D shares transfer exchange resources when redeemed, Class A shares and Class B shares are both suitable for long-term holding, and Class C shares and Class D shares are both suitable for short-term holding.

[0122] In one possible implementation, different shares of the same fund can be used as candidate management products for resource evaluation, similar shares of different funds can be used as candidate management products for resource evaluation, and non-similar shares of different funds can be used as candidate management products for resource evaluation. There is no restriction on this.

[0123] In one possible implementation, multiple fund shares can be regarded as a candidate management product. The multiple fund shares can belong to the same fund or different funds. In other words, a combination of multiple fund shares can be regarded as a candidate management product, and resource evaluation is performed on it together with a combination of other fund shares or a certain fund share.

[0124] In a possible implementation, each candidate management product may include management products that are all selected by the target object, or may include management products that are partially selected by the target object.

[0125] In one possible implementation, the exchange resource difference includes any one or more of the following information: resource difference; resource difference between the exchange resources of each candidate management product within the target holding time; exchange resources of each candidate management product within the target holding time, but is not limited to this.

[0126] As a first example, the exchange resource differences between the candidate management products include: reference duration, which represents: the product holding time required when the exchange resources of the candidate management products are equal, that is, when the target object holds each candidate management product for a reference duration, the exchange resources of each candidate management product are equal, and the reference duration can also be called a resource difference point or breakeven point. The unit of the reference duration can be year or day, but is not limited to this.

[0127] The reference duration may be obtained in the following manner, but is not limited to: the server evaluates the resources of each candidate management product based on the total amount of resource input and the exchange resource calculation method corresponding to each candidate management product to obtain the reference duration.

[0128] Taking the candidate management products including Class A shares and Class C shares of xx fund as an example, assuming that the exchange resources of Class A shares include subscription fees, management fees, custody fees and redemption fees, and the exchange resources of Class C shares include management fees, custody fees, sales service fees and redemption fees, the server determines the breakeven point where the exchange resources of Class A shares and Class C shares are equal based on the exchange resource calculation method of Class A shares and the exchange resource calculation method of Class C shares.

[0129] Y is used to represent the breakeven point in years, and D is used to represent the breakeven point in days, where D = Y*total number of days in the year.

[0130] The calculation method for exchange resources of Class A shares is: Exchange resources of Class A shares = subscription rate*total resource investment+management fee rate*total resource investment*Y+custody fee rate*total resource investment*Y+A redemption rate*total resource investment.

[0131] The calculation method for exchange resources of Class C shares is: Exchange resources of Class C shares = management fee rate*total resource investment*Y+custody fee rate*total resource investment*Y+sales service fee rate*total resource investment*Y+C redemption rate*total resource investment.

[0132] In the above-mentioned exchange resource calculation method, the subscription rate refers to the proportion of resources transferred at one time when purchasing shares; the management fee rate refers to the proportion of resources transferred to the fund management company, usually calculated on an annual basis; the custody fee rate refers to the proportion of resources transferred to the fund custodian bank, usually calculated on an annual basis; the A redemption rate refers to the proportion of resources transferred when redeeming Class A shares; the C redemption rate refers to the proportion of resources transferred when redeeming Class C shares; the sales service fee rate refers to the resources unique to Class C shares, usually calculated on an annual basis.

[0133] Assuming that the management fee rates of Class A shares and Class C shares are the same, and the custody fee rates of Class A shares and Class C shares are the same, when the exchange resources of Class A shares and Class C shares are equal, Y = (subscription rate + A redemption rate - C redemption rate) / sales service rate.

[0134] For example, see Figure 3 As shown, it is a schematic diagram of the breakeven points of Class A shares and Class C shares provided in an embodiment of the present application. The horizontal axis in the figure represents the holding period of the candidate management product. It is assumed that the breakeven point is 361 days. When the holding period of the candidate management product exceeds 361 days, the exchange resources of Class A shares are equal to the exchange resources of Class C shares. When the holding period of the candidate management product exceeds 361 days, the exchange resources of Class A shares are lower than the exchange resources of Class C shares. When the holding period of the candidate management product is less than 361 days, the exchange resources of Class C shares are lower than the exchange resources of Class A shares.

[0135] The server determines the breakeven point by comparing the exchange resources of Class A shares with the exchange resources of Class C shares, and then determines the target management product based on the breakeven point to help the target object decide which fund shares with exchange resource calculation method are more economical under different holding periods, thereby avoiding the risk of resource loss as much as possible and improving asset security.

[0136] In a possible implementation, considering that the values ​​of various parameters in the exchange resource calculation method may be different under different holding periods, the server evaluates the resources of each candidate management product based on the total amount of resource investment and the exchange resource calculation method corresponding to each candidate management product under the target holding period to obtain a reference period. The target holding period can be input by the target object or inferred by the server based on the target object's historical product holding period.

[0137] As a second example, the exchange resource difference includes: the resource difference between the exchange resources of each candidate management product within the target holding time.

[0138] The target holding time may be the holding time input by the target object, or may be estimated by the server based on the historical holding time of the target object. For example, the server may take the average of the historical holding times within a certain time range as the target holding time.

[0139] The server can obtain the exchange resources of each candidate management product within the target holding time based on the target holding time and the total resource investment of the target object, combined with various exchange resource calculation methods, and obtain the resource difference of each exchange resource based on the obtained exchange resources.

[0140] The exchange resources of each candidate management product are calculated based on the target holding time using a corresponding exchange resource calculation method.

[0141] Still taking the Class A and Class C shares of xx fund as an example, assuming that the total resource investment is 50,000 and the target holding period is 365 days. Within the target holding period, the exchange resources of Class A shares = subscription rate * total resource investment + management fee rate * total resource investment * target holding period + custody fee rate * total resource investment * target holding period + A redemption rate * total resource investment = 0.12% * 50,000 + 0.5% * 50,000 * 1 + 0.1% * 50,000 * 1 + 0 * 50,000 = 360.

[0142] During the target holding period, the exchange resources of Class C shares = management fee rate * total resource investment * target holding period + custody fee rate * total resource investment * target holding period + sales service fee rate * total resource investment * target holding period + C redemption rate * total resource investment = 0.5% * 50000 * 1 + 0.1% * 50000 * 1 + 0.2% * 50000 * 1 + 0 * 50000 = 400.

[0143] Therefore, based on the target holding time and total resource investment, the server calculates the exchange resources of Class A shares and Class C shares respectively, and obtains the resource difference between Class A shares and Class C shares. Resource difference = 400-360 = 40, that is, when Class A shares and Class C shares are held for 365 days, the exchange resources of Class A shares are 40 less than the exchange resources of Class C shares.

[0144] The server obtains the resource difference of the exchange resources of Class A shares and Class C shares by comparing their respective computing resources. The server uses the specific resource difference to guide the selection of subsequent target management products, so as to assist the target objects in choosing which fund shares with the exchange resource calculation method are more economical under the same holding period, thereby avoiding the risk of resource loss as much as possible and improving asset security.

[0145] As a third example, the exchange resource difference includes: the exchange resource of each reference exchange resource within the target holding time. The calculation method of the exchange resource of each reference exchange resource is the same as above, which will not be repeated here.

[0146] As a fourth example, resource differences include: breakeven points, and the exchange resources of each reference exchange resource within the target holding time.

[0147] Specifically, the server performs resource evaluation on each candidate management product based on the total resource investment and the exchange resource calculation method corresponding to each candidate management product to obtain a reference duration, and obtains an exchange resource difference based on the reference duration and the resource difference between the obtained exchange resources.

[0148] As the calculation process of breakeven points and the calculation process of exchange resources are specifically described above, they will not be repeated here.

[0149] S203: Based on the exchange resource difference and in combination with the object information of the target object, a target management product recommended to the target object is selected from each candidate management product.

[0150] Among them, the object information of the target object is used to characterize the asset management preferences of the target object, such as the risk tolerance of the target object, the preferred product holding period, the preferred total amount of resource investment, etc.

[0151] Exemplarily, S203 includes the following steps:

[0152] Firstly, based on the exchange resource differences and in combination with the object information of the target object, the matching evaluation information between each candidate management product and the target object is obtained;

[0153] Secondly, based on the obtained matching evaluation information, a candidate management product that meets the set matching conditions is selected from the candidate management products as the target management product.

[0154] The matching evaluation information between a candidate management product and a target object is used to represent the matching degree between the candidate management product and the target object. The matching evaluation information between a candidate management product and a target object can also be called the candidate management product corresponding to the candidate management product. The matching evaluation information can be expressed in numerical values, or in grades or text descriptions. The number of target management products can be one or more.

[0155] In a possible implementation, if the object information includes: each key preference information corresponding to the target object and the information order between them, and the information order is used to characterize the importance of each key preference information, then the server obtains the matching evaluation information between each candidate management product and the target object based on the exchange resource difference and the information order between each key preference information. The key preference information may include one or more behavior tags, and each key preference information and the information order between them may be represented by a behavior tag sequence, each behavior tag in the behavior tag sequence represents a key preference information, and the arrangement order between each behavior tag in the behavior tag sequence represents the information order.

[0156] In the embodiment of the present application, the process of obtaining each key preference information and the order of information therebetween includes but is not limited to the following two situations:

[0157] Case 1: If the custom preference information of the target object is set, the server obtains the object information of the target object based on the custom preference information.

[0158] For example, the client can provide a preference information template to the target object, and the preference information template is used for the target object to fill in the content of each key preference information. After the target object fills in the preference information template, the client uses the filled preference information template as the custom preference information of the target object, and transmits the custom preference information to the server. After receiving the custom preference information, the server directly uses the custom preference information as the object information of the target object.

[0159] In this way, accurate personalized preference information can be obtained, which is conducive to improving the accuracy of the target management products determined subsequently, making the target management products closer to the asset management preferences of the target objects, thereby improving the user experience.

[0160] Case 2: If the custom preference information of the target object is not set, the server performs preference inference based on the historical product usage data and historical product feedback data of the target object to obtain the object information of the target object.

[0161] That is to say, when the target object has not set its own preference information, the target object’s historical product usage data and historical product evaluation data can be used to infer the target object’s object information, thereby ensuring that the recommended target management products are in line with the target object’s asset management preferences in the absence of customized preference information.

[0162] In a possible implementation, the server performs preference inference based on the historical product usage data and historical product feedback data of the target object to obtain the object information of the target object, including but not limited to the following steps:

[0163] Step A: The server obtains the explicit preference characteristics of the target object based on historical product feedback data.

[0164] Among them, explicit preference features are used to characterize the preferences explicitly expressed by the target object, which are usually obtained through the feedback data of the target object, such as ratings, comments, and collections of management products.

[0165] In a possible implementation, step A specifically includes the following steps:

[0166] A1. The server extracts each word from the historical product feedback data, encodes each word, and obtains the word features corresponding to each word.

[0167] A2. The server aggregates information based on the association between aspect labels and words and the word features corresponding to each word to obtain aspect label features corresponding to each aspect label.

[0168] A3. The server aggregates information based on the association between products and aspect tags and aspect tag features corresponding to each aspect tag, and obtains product features corresponding to each historical management product in the historical product feedback data.

[0169] A4. Based on the association between products and objects, the server aggregates information in combination with the product features corresponding to each historical management product to obtain the explicit preference features of the target object.

[0170] Among them, the historical product feedback data includes: feedback data triggered by the target object for each historical management product, for example, the target object’s evaluation of three historical management products. The three historical management products include Product 1, Product 2 and Product 3. Among them, the evaluation for Product 1 is “very suitable for long-term investment, very good”, the evaluation for Product 2 is “sound investment”, and the evaluation for Product 3 is “short-term high return”.

[0171] Each historical management product may be a management product purchased by the target object within a certain time range in the past.

[0172] Aspect tags are used to represent information about a certain aspect of a management product, such as fund type, risk level, various ratings, etc.

[0173] In a possible implementation, a Hierarchical Heterogeneous User Usage Graph (HHRG) may be used to represent: the association relationship between aspect labels and words, the association relationship between products and aspect labels, and the association relationship between products and objects.

[0174] See also Figure 4 As shown, it is a schematic diagram of an HHRG provided in an embodiment of the present application. The HHRG includes four layers, from top to bottom: object (user) layer, product (product) layer, aspect label (aspect) layer, word (word) layer. Each layer contains several nodes. The nodes contained in the object layer are called object nodes, and each object node represents an object; the nodes contained in the product layer are called product nodes, and each product node represents a management product; the nodes contained in the aspect label layer are called aspect label nodes, and each aspect label node represents an aspect label; the nodes contained in the word layer are called word nodes, and each word node represents a word. The connection relationship between the object node and the product node represents the association relationship between the product and the object, the connection relationship between the product node and the aspect label node represents the association relationship between the product and the aspect label, and the connection relationship between the aspect label node and the word node represents the association relationship between the aspect label and the word. Taking the connection relationship between the object node and the product node as an example, if an object node and a product node have an edge connection, then the object represented by the object node and the management product represented by the product node have an association relationship.

[0175] The nodes of each layer in HHRG can be pre-set, and the connection relationship between the nodes can be automatically learned through the model training process.

[0176] In one possible implementation, information aggregation can be implemented using but not limited to heterogeneous graph ATtention networks (HGAT).

[0177] In the embodiment of the present application, the process of obtaining explicit preference features can also be referred to as an information propagation process based on the word-aspect label-product-object (Word-Aspect-Product-User, WAPU) meta-path, that is, based on historical product feedback data, starting from the word layer, the information is first aggregated to the aspect label layer, then to the product layer, and finally to the object layer to obtain the explicit preference features of the target object.

[0178] In HHRG, for the object layer, product layer, and aspect label layer, the neighbor nodes of each node in the layer refer to the heterogeneous nodes in the lower layer. u Represents the set of neighbor nodes of the object at the product layer. The explicit preference characteristics of the target object can be expressed by the following formula (1):

[0179]

[0180] in, represents the explicit preference feature of object u in WAPU, Indicates the initial value of the explicit preference feature. The initial value can be preset. represents the product features of product p in WAPU, product p belongs to N of object u u The neighbor nodes in N u The product characteristics of each neighbor node in .

[0181] Similarly, for the nodes in the product layer and aspect label layer in HHRG, the corresponding feature representation can be obtained according to the neighbor node set of the node.

[0182] Through HHRG, the information interaction between objects and products is enhanced. Combined with the node information update method based on HGAT and the information propagation method based on WAPU meta-path, the explicit personalized preferences of the target objects can be obtained.

[0183] Compared with mining personalized preference information directly from historical product feedback data, the WAPU meta-path aggregates product information first and then obtains explicit preference features through the object's feedback on the product. The explicit preference feature provides direct object feedback, which helps to quickly and accurately understand the preferences of the target object.

[0184] In one possible implementation, the server can use the Explicit Feedback Tracker for Users (REPT) to obtain explicit preference features, that is, the server inputs historical product feedback data into REPT to obtain the explicit preference features of the target object. REPT obtains explicit preference features based on HHRG and the information propagation method of WAPU meta-path.

[0185] Feature learning based on WAPU meta-path can obtain object preferences from a wider range of data and explore the potential relationship between objects and products through the interaction between them, thereby enhancing the user's preference representation and alleviating the sparsity problem of sample data.

[0186] Step B: Obtain the implicit preference characteristics of the target object based on historical product usage data.

[0187] Implicit preference features are used to characterize the preferences indirectly expressed by the target object through its behavior, such as purchasing behavior, collection behavior, clicking behavior, etc.

[0188] In one possible implementation, the server inputs historical product usage data into an Implicit Feedback Tracker for Users (BIPT) to obtain implicit preference features.

[0189] Step C: Based on explicit preference features and implicit preference features, combined with preset labels of various aspects, obtain key preference information and the order of information between them.

[0190] In a possible implementation, step C specifically includes the following steps:

[0191] Step C1: Based on the explicit preference features and the implicit preference features, the various aspect labels are evaluated to obtain the label evaluation values ​​corresponding to each aspect label, and based on the obtained label evaluation values, an aspect label sequence containing K aspect labels is screened out.

[0192] Step C2: Generate corresponding key preference information based on the semantic information corresponding to each of the k aspect labels, and obtain the information order between the generated key preference information based on the label order of the k aspect labels in the aspect label sequence.

[0193] The label evaluation value of each aspect label is used to evaluate the matching degree between the aspect label and the target object.

[0194] In one possible implementation, a ranking-aware Tag Generator (PRT) can be used to obtain key preference information and the information order between them based on explicit preference features and implicit preference features, that is, the explicit preference features and implicit preference features are input into the PRT to obtain key preference information and the information order between them.

[0195] Since the object's preference in key preference information is consistent with the object's preference in aspect labels, the order of key preference information can be determined by generating ordered aspect labels. Specifically, in PRT, the generation of object information can be divided into two stages: ordered aspect label generation (first stage) and ordered key preference information generation (second stage). In the first stage, an aspect label sequence containing k aspect labels is generated based on explicit preference features and implicit preference features, i.e., step C1. In the second node, the key preference information sequence is generated using the aspect label sequence, i.e., step C2.

[0196] In a possible implementation, the first stage can be implemented in the following ways, but not limited to:

[0197] First, based on explicit preference features and implicit preference features, the server predicts the preference of each aspect label in the preset aspect label set and predicts the corresponding label evaluation value;

[0198] Next, the server selects k aspect labels from the aspect labels based on the label evaluation values ​​corresponding to each aspect label, and sorts the selected k aspect labels according to the label evaluation values ​​to obtain an aspect label sequence.

[0199] Exemplarily, the server performs feature fusion based on explicit preference features and implicit preference features to obtain the object features of the target object. Then, the server calculates the similarity between the aspect label features corresponding to each aspect label and the object features for each aspect label, and uses the calculated similarity as the corresponding label evaluation value. In the embodiments of the present application, the feature fusion method and the similarity calculation method are not limited.

[0200] In the embodiment of the present application, the k aspect labels selected are the k aspect labels with the highest matching degree with the target object among the aspect labels. Exemplarily, the larger the value of the label evaluation value, the higher the matching degree between the aspect label and the target object, and the first k aspect labels with the highest label evaluation value are selected. When the selected k aspect labels are subsequently sorted, the k aspect labels are sorted in descending order according to the label evaluation value to obtain an aspect label sequence.

[0201] For example, assuming that the preset aspect label set contains 100 aspect labels, the value of k is 5, and the server selects the top 5 aspect labels with the highest label evaluation values ​​from the aspect labels based on the label evaluation values ​​corresponding to each of the 100 aspect labels, and sorts the selected 5 aspect labels according to the label evaluation values ​​to obtain an aspect label sequence.

[0202] In a possible implementation, the second stage can be implemented in the following ways, but not limited to:

[0203] First, the server generates key preference information describing each aspect tag among the k aspect tags, that is, the key preference information corresponding to the aspect tag;

[0204] Next, the server uses the tag order of the k aspect tags in the aspect tag sequence as the information order between the k key preference information;

[0205] Finally, based on the generated k key preference information and the order of the information between them, a key preference information sequence is obtained.

[0206] Among them, the key preference information corresponding to an aspect label can be composed of one or more behavior labels.

[0207] In one possible implementation, PRT is implemented using the Transformer architecture. The Transformer model is mainly used to process sequence data. Each input sequence consists of a series of time steps, and each time step corresponds to a token or data point. In the Transformer model, the decoding time (t) represents the tth position in the input sequence.

[0208] At the t-th decoding moment of PRT, the object features (including explicit preference features and implicit preference features) and the decoder input (x t ), input into a fusion layer, and then send the fusion information into the decoder for generation, so as to obtain the label evaluation values ​​corresponding to each aspect of the label. This process can be expressed by the following formula (2) and formula (3):

[0209]

[0210] p(y t )=TransDec([e1,e2,…,e t ]) Formula (3)

[0211] in, represents the explicit preference features obtained from REPT, represents the implicit preference features obtained from BIPT; et represents the fused feature vector at time t, e t is based on and x t Fusion; y t represents the behavior or preference of the target object at time t, which can be a specific action, such as clicking, purchasing, rating, etc., or a more abstract preference state of the target object; p(y t ) represents y t The probability distribution of x is the label evaluation value corresponding to each of the labels. The probability distribution is usually calculated by a probability model, which estimates the probability of the target object taking a specific behavior based on historical data and the current context; t represents the contextual features or other relevant information at time t, x t This information includes but is not limited to timestamps, device types, geographic locations, historical behaviors of the target object, etc. This information helps to understand and predict y t .

[0212] Fusion represents a fusion layer, which can be designed as a linear layer when applied; TransDec represents a Transformer decoder (which can be referred to as a decoder for short), which can include three components: a multi-head self-attention mechanism, a multi-head cross-attention mechanism, and a feed forward neural network (FFN) layer, but is not limited to these. When the multi-head cross-attention mechanism is used, the historical product usage data and the historical product feedback data are integrated, that is, at the decoding time t, the state vector e of the decoder is used t As the query (Query), the vector representation (h enc ) as the key and value.

[0213] In the first generation stage of PRT, the aspect label sequence is generated. Assume that (a1, a2, …, a k ) represents the generated aspect label sequence, a1 represents the first aspect label, a2 represents the second aspect label, and a k represents the k-th aspect label.

[0214] At the first decoding time t1, the product representation obtained by the information propagation method of the WAPU meta-path in REPT is used as the initial input of the decoder, that is, Then substitute x1 into the above formula (2) and formula (3) to obtain the generation probability (i.e., label evaluation value) of each aspect label in the predefined aspect label set, that is, y1 = a i (y1 includes all aspects of the label), and p(a i ) represents the set of generation probabilities corresponding to each of the various labels. In order to make the generation probability reflect the object's preference for various labels, in the process of model training, the ranking loss (Rank Loss) is introduced in the generation of the first stage. The calculation method of the ranking loss is detailed below.

[0215] In the second generation stage of PRT, according to the generated aspect label sequence, a corresponding behavior label is generated for each aspect label, that is, for a1, a2, ..., a k For each aspect label in , a corresponding behavior label is generated in turn, where the number of behavior labels can be one or more. The behavior label corresponding to the aspect label is a label used to describe the aspect label. For example, if the aspect label is color, the label may be red, green, yellow, etc., which are labels describing colors.

[0216] An aspect label and its corresponding at least one behavior label can be called an aspect label-behavior label sequence. For each aspect label-behavior label sequence, the aspect label is regarded as the first label of the aspect label-behavior label sequence, and then other words (i.e., behavior labels) in the aspect label-behavior label sequence are generated in sequence. In PRT, aspect labels and their corresponding behavior labels can be generated at the same time, and finally spliced ​​into aspect label-behavior label sequences in order.

[0217] Exemplarily, at time t, the decoder input can be expressed by formula (4):

[0218]

[0219] Among them, a i represents the aspect label, i represents the i-th aspect label-behavior label sequence, y i,t-1 Indicates a i The corresponding aspect label-behavior label sequence is the t-1th label.

[0220] That is, at t = 1, As the encoder input; at t=2, the feature representation of each aspect label is used as the decoder input; at t=3, the feature representation of the first behavior label of each aspect label is used as the decoder input; at t=4, the feature representation of the second behavior label of each aspect label is used as the decoder input.

[0221] See also Figure 5As shown, PRT mainly consists of the following parts: input layer, output layer, N encoders and N decoders. Figure 5 In the description, only N=1 is used as an example. Below, these parts are described separately.

[0222] (1) Input layer

[0223] The input of the input layer includes input data 1 and input data 2, wherein input data 1 includes: explicit preference features and implicit preference features, and the target text is Xt.

[0224] The input layer includes embedding layer 1 and embedding layer 2, wherein embedding layer 1 is used to obtain the embedding representation vector and the position representation vector of input data 1, and the final word segmentation representation vector is obtained by linear addition of the two representation vectors. Embedding layer 2 is used to obtain the embedding representation vector and the position representation vector of input data 2, and the final word segmentation representation vector is obtained by linear addition of the two representation vectors.

[0225] (3) Encoder

[0226] The encoder part is composed of N encoders stacked together, each of which includes two sub-layer connection structures. The first sub-layer connection structure includes: Masked Multi-Head Attention layer and Residual Connection and Normalization. The second sub-layer connection structure includes: FeedForward layer and Residual Connection and Normalization layer.

[0227] Exemplarily, an encoder includes: a multi-head attention mechanism layer 1, a residual connection and normalization (LayerNormalization) layer 1, a feedforward layer 1, and a residual connection and normalization layer 2.

[0228] Among them, the multi-head attention mechanism layer 1 is used to capture the dependencies between different positions in the input sequence by performing multi-head attention calculation on the input sequence. The multi-head attention mechanism layer 1 is implemented using the self-attention mechanism.

[0229] The residual connection and normalization layer 1 accelerates the training of the model and improves the performance of the model by performing a residual connection between the output of the multi-head attention mechanism layer 1 and the input sequence and normalizing the result after the residual connection.

[0230] The feed-forward layer 1 is used to enhance the feature expression capability of each position by performing a fully connected layer calculation on the representation of each position in the output of the residual connection and normalization layer 1.

[0231] The residual connection and normalization layer 2 is used to perform a residual connection on the output of the feedforward layer 1 and the output of the normalization layer 1, and to normalize the result after the residual connection.

[0232] (4) Decoder

[0233] The decoder part is composed of N decoders stacked together, and each encoder includes three sub-layer connection structures. The first sub-layer connection structure includes: a masked multi-head attention mechanism layer and a residual connection and a normalization layer; the second sub-layer connection structure includes: a multi-head attention mechanism layer and a residual connection and a normalization layer; the second sub-layer connection structure includes: a feedforward layer and a residual connection and a normalization layer

[0234] Exemplarily, an encoder includes: a multi-head attention mechanism layer with a mask mechanism, a residual connection and normalization layer 3, a multi-head attention mechanism layer 2, a residual connection and normalization layer 4, a feedforward layer 2 and a residual connection and normalization layer 5.

[0235] In the multi-head attention mechanism layer with mask mechanism, the dependencies between different positions in the input sequence are captured by performing multi-head attention calculation on the input sequence. The mask mechanism is used to avoid using future information in the attention mechanism. The multi-head attention mechanism layer with mask mechanism is implemented using the self-attention mechanism.

[0236] The residual connection and normalization layer 3 is used to perform residual connection on the output of the multi-head attention mechanism layer with a mask mechanism and the input sequence, and normalize the result after the residual connection.

[0237] The multi-head attention mechanism layer 2 is used to perform multi-head attention calculation on the output of the residual connection and normalization layer 3 and the output of the residual connection and normalization layer 2. The multi-head attention mechanism layer 2 is implemented using a cross attention mechanism.

[0238] The residual connection and normalization layer 4 is used to perform residual connection on the output of the multi-head attention mechanism layer 2 and the output of the residual connection and normalization layer 3, and normalize the result after the residual connection.

[0239] The feed-forward layer 2 is used to compute the fully connected layer by performing the representation of each position in the output of the residual connection and normalization layer 4.

[0240] The residual connection and normalization layer 5 is used to perform a residual connection on the output of the feedforward layer 2 and the output of the residual connection and normalization layer 4, and normalize the result after the residual connection.

[0241] (5) Output layer

[0242] In a possible implementation, the output layer may be composed of a linear transformation layer and a Softmax function layer, wherein the linear transformation layer maps the output of the decoder to a vector space, and the Softmax function converts each element in the vector space into a probability value, each probability value representing a label evaluation value.

[0243] In PRT, a two-stage generation task is designed. In the first stage, personalized information is introduced to sequentially generate aspect labels of interest to the object, and then in the second stage, the behavior labels corresponding to the aspect labels are sequentially generated. In this way, the task of generating personalized behavior labels by PRT is completed.

[0244] In one possible implementation, based on the exchange resource difference and the key preference information sequence (each key preference information and the information order between them), when obtaining the matching evaluation information between each candidate management product and the target object, the candidate management products that match the key preference information can be screened out from each candidate management product based on the exchange resource difference and each key preference information. Then, according to the information order, the screened-out candidate management products that match the key preference information are sorted, and based on the sorting result, the matching evaluation information between each candidate management product and the target object is obtained.

[0245] See also Figure 6 As shown, it is a schematic diagram of an object information acquisition process provided in an embodiment of the present application. First, the server obtains the explicit preference features of the target object using REPT based on historical product feedback data and HHRG. Next, the server uses BIPT to obtain the implicit preference characteristics of the target object based on historical product usage data. Afterwards, the server will and Input into PRT to obtain each key preference information and the information order between them. In the first stage of PRT, based on explicit preference features and implicit preference features, preference prediction is performed on each aspect label in the preset aspect label set, and the corresponding label evaluation values ​​are predicted. Based on the label evaluation values ​​corresponding to each aspect label, k aspect labels are selected from the aspect labels, and the selected k aspect labels are sorted according to the label evaluation values ​​to obtain the aspect label sequence. In the second stage of PRT, for each aspect label among the k aspect labels, key preference information describing the aspect label is generated, that is, the key preference information corresponding to the aspect label; the label order of the k aspect labels in the aspect label sequence is used as the information order between the k key preference information; based on the generated k key preference information and the information order between them, a key preference information sequence is obtained.

[0246] Recommendations based on key preference information can help target objects more conveniently make decisions that are in line with their own asset management goals and preferences, thereby greatly enhancing the user experience and providing an efficient, secure and object-friendly asset management decision-making assistance tool.

[0247] See also Figure 7 As shown, it is a flow chart of a method for obtaining object information provided in an embodiment of the present application. The flow is applied to a server, and the specific flow is as follows:

[0248] S701. Extract each word from the historical product feedback data, encode each word, and obtain the word features corresponding to each word.

[0249] S702: Based on the association between the aspect labels and the words, information aggregation is performed in combination with the word features corresponding to each word to obtain the aspect label features corresponding to each aspect label.

[0250] S703: Based on the association between the product and the aspect label, information is aggregated in combination with the aspect label features corresponding to each aspect label to obtain the product features corresponding to each historical management product in the historical product feedback data.

[0251] S704: Based on the association relationship between products and objects, information is aggregated in combination with the product features corresponding to each historical management product to obtain explicit preference features of the target object.

[0252] S705: Obtain implicit preference characteristics of the target object based on historical product usage data.

[0253] S706. Evaluate each aspect label based on the explicit preference feature and the implicit preference feature to obtain label evaluation values ​​corresponding to each aspect label, and select an aspect label sequence including K aspect labels based on the obtained label evaluation values.

[0254] S707 , generating corresponding key preference information based on the semantic information corresponding to each of the k aspect tags, and obtaining a key preference information sequence based on the tag order of the k aspect tags in the aspect tag sequence.

[0255] S708: Use the key preference information sequence as the object information of the target object.

[0256] In a possible implementation, setting the matching condition includes but is not limited to at least one of the following conditions: the matching evaluation information exceeds a set evaluation information threshold; the value of the matching evaluation information represents the highest degree of matching with the target object.

[0257] In a possible implementation, the server may also generate recommendation information based on target management products and exchange resource differences, where the recommendation information is represented in the form of tags; and send the product recommendation information to the client so that the client presents the product recommendation information to the target object.

[0258] In order to facilitate the target object to quickly understand the recommended information, the recommended information can be represented in the form of tags. The recommended information in the form of tags can also be called a decision tag, but is not limited to this.

[0259] For example, the recommendation information may be a decision label such as “Long-term investment recommendation share A” or “Short-term investment recommendation category C”.

[0260] Since the labels are expressed in concise and clear language, they can help the target audience quickly and accurately grasp the investment advice, thereby avoiding the purchase of inappropriate products due to semantic errors, which in turn may lead to the target audience's personal asset losses.

[0261] In a possible implementation, after obtaining the total amount of resource investment, the server can directly evaluate the resources of each candidate management product based on the total amount of resource investment and the exchange resource calculation method corresponding to each candidate management product to obtain the exchange resource difference between the candidate management products.

[0262] In a possible implementation, after the server obtains the total amount of resource input, when the total amount of resource input reaches the input amount threshold set for the target object, the server performs resource evaluation based on the total amount of resource input and each exchange resource calculation method to obtain the exchange resource difference. The input amount threshold is the first threshold or the second threshold, and the first threshold is higher than the second threshold.

[0263] Exemplarily, the first threshold is a robust threshold, and the second threshold is an advanced threshold.

[0264] The robust threshold is used to trigger the resource exchange comparison prompt when the total amount of resource input by the target object reaches the threshold, that is, to provide the resource exchange difference in the client. The setting of the robust threshold takes into account the general situation of the market and the average investment scale of most objects, and aims to provide a benchmark investment advice for a wide range of objects. The setting of the robust threshold helps the object to obtain a relatively safe and conservative cost comparison result when investing, and is suitable for the object group with low risk appetite.

[0265] The advanced threshold is designed for those with higher risk tolerance. When the total amount of resource input by the subject exceeds the advanced threshold, a more detailed and in-depth exchange resource analysis and investment advice will be provided in the client. The setting of the advanced threshold allows the subject to obtain more personalized and accurate investment decision support based on their investment preferences and market conditions. The introduction of the advanced threshold enables the system to adapt to the complex needs of different advanced thresholds and provide more flexible investment strategies.

[0266] In some implementations, the server may periodically recalculate and update the behavior tags to reflect market changes and adjustments to the fund exchange resource structure.

[0267] In some implementations, while the target object holds the target management product, the server may continuously detect and timely update the investment advice.

[0268] Below, the model training process of REPT, BIPT and PRT mentioned in the embodiment of the present application is explained. In the embodiment of the present application, a multi-task learning framework is adopted to minimize the loss function of each module, each module includes REPT, BIPT, PRT, and the loss function of each module includes: contrast loss in REPT, selection loss in BIPT, sorting loss and generation loss in PRT.

[0269] See also Figure 8 As shown, it is a structural diagram of a label generation model provided in an embodiment of the present application. The label generation model is used to generate a corresponding key preference information sequence based on the historical product feedback data and historical product usage data of the target object. The label generation model includes: a first sub-model for generating explicit preference features, a second sub-model for generating implicit preference features, and a third sub-model for generating key preference information sequences. Exemplarily, the first sub-model may be REPT, the second sub-model may be BIPT, and the third sub-model may be PRT, REPT is used to obtain the corresponding explicit preference features based on the historical product feedback data; BIPT is used to obtain the corresponding implicit preference features based on the historical product usage data; PRT is used to obtain the key preference information sequence based on the implicit preference features and the explicit preference features.

[0270] See also Fig. 9 As shown, it is a flow chart of a model training method provided in an embodiment of the present application. The model training method is applied to a server. The model training process is an iterative training process. During the iterative training process, all sample data are divided into specified batches, and training is performed based on the sample data of each batch. Since the steps performed for training each batch in each iteration are similar, the training for one batch is used as an example for explanation here.

[0271] S901. Obtain corresponding explicit preference features based on historical product feedback data of each sample object, and obtain contrast loss based on the explicit preference features and true explicit preference features of each sample object.

[0272] The sample data includes: historical product feedback data and historical product usage data of each sample object. The contrast loss is used to measure the loss in the first sub-model.

[0273] In one possible implementation, the server can obtain the explicit preference features corresponding to each sample object based on the historical product feedback data of each sample object through the information aggregation method of the WAPU meta-path, and then obtain the contrast loss in REPT based on the explicit preference features corresponding to each sample object and the real explicit preference features corresponding to each sample object.

[0274] In one possible implementation, although the information aggregation method based on the WAPU meta-path enables REPT to obtain the explicit preference features of the target object in sparse data, the implementation of the WAPU meta-path requires the use of a large amount of historical usage data, and the explicit preference features are aggregated from words to aspect labels, then from aspect labels to products, and finally from products to objects. The information propagation path is very long, which will bring certain noise both in terms of data and implementation methods, resulting in deviations in the learned explicit preference features. In order to solve this problem, in the embodiment of the present application, the Word-Aspect-User (WAU) meta-path is introduced to assist the learning of explicit preference features in REPT, thereby alleviating the noise brought by the WAPU meta-path.

[0275] Among them, the information propagation path of the WAU meta-path is: from word aggregation to aspect labels, and then directly from aspect labels to objects. Specifically, the information aggregation process of the WAU meta-path is: the server extracts each word from the historical product feedback data, encodes each word, obtains the word features corresponding to each word, and then aggregates information based on the reference word set associated with each of the preset aspect labels, combined with the word features corresponding to each reference word, to obtain the aspect label features corresponding to each of the aspect labels. Finally, the server obtains the explicit preference features of the target object based on the aspect label features corresponding to each of the aspect labels.

[0276] Compared with the information propagation path of the WAPU meta-path, the information propagation path of the WAU meta-path is shorter, which can avoid the error accumulation from the product layer to the object layer, thereby making the data error smaller in the aggregation process.

[0277] In the information aggregation process based on WAU meta-path, the goal of REPT is to mine object preferences based on the historical data of sample objects and obtain explicit preference features of sample objects. In HHRG, the edge between the object and aspect label is determined based on whether the historical data of the sample object contains the aspect label. If the historical data of the sample object contains the aspect label, there is an edge connection relationship between the object and the aspect label, otherwise there is no edge connection relationship between the object and the aspect label. Due to the sparsity of the data, the representation information obtained based on the WAU meta-path may be very limited, but it is relatively more accurate. By using this representation to supervise the meta-path propagation process of WAPU, the explicit preference features finally obtained by the WAPU meta-path will be more accurate and comprehensive.

[0278] In S901, the server may obtain the explicit preference features corresponding to each sample object through the information aggregation method of the WAU meta-path based on the historical product feedback data of each sample object, and then obtain the selection loss in REPT based on the explicit preference features corresponding to each sample object and the product features. The product features are obtained by using the historical product feedback data of each sample object through the information aggregation method of the WAPU meta-path.

[0279] Specifically, in REPT, contrastive learning is used to enhance the feature representation of objects at the product level of WAPU and the user level of WAU. This process is based on the assumption that if a sample object likes a product, then the information contained in the product is consistent with the sample object's preference information to some extent. For any product p, the set of sample objects that like product p (for example, write a good review for product p) is denoted as u + (u + It can also be called a positive example object), and the set of other sample objects is represented as u - (u - It can also be called a negative example object). In contrastive learning, the product features of product p in WAUP are compared with u in WAU. + The object features of the sample objects in are close, and as far away from u as possible - The object features of the sample objects in . This process can be expressed by formula (5):

[0280]

[0281] Among them, L c represents the contrast loss in REPT, represents the product characteristics of product p in WAUP, Represents object u in WAU + The object characteristics, u * Indicates u + and u - The sample objects in Indicates u in WAU * The object features of “·” represent the distance.

[0282] For example, see Fig.10 As shown in the figure, in the meta-path propagation path of WAPU, HHRG includes a word layer, an aspect label layer, a product layer, and an object layer, wherein the product layer includes multiple products (represented by ellipses), and the multiple products include product p. In the meta-path propagation path of WAU, HHRG includes a word layer, an aspect label layer, and an object layer, wherein the object layer includes 5 sample objects. For product p, among the 5 sample objects, sample object 1, sample object 2, and sample object 3 belong to u. + , sample objects 4 and 5 belong to u - When calculating the sub-contrast loss corresponding to product p, the sub-contrast loss corresponding to product p is obtained based on the feature distances between the object features of the five sample objects and the product features of product p in WAUP, combined with the feature distances between sample object 1, sample object 2, and each of the sample objects and the product features. Similarly, for each sample product in the product layer in the meta-path propagation path of WAPU, the corresponding sub-contrast loss can be obtained, and then the sub-contrast losses of each sample product are summarized to obtain the contrast loss in REPT.

[0283] By minimizing the feature distance between the product features of product p and the object features of the positive object (assuming that the smaller the distance, the closer the semantics), the contrast loss can make the semantic representation of the product features close to that of the object features of the positive object. After being supervised by the object features in WAU, the product features in WAUP will be more accurate. Therefore, the object features aggregated from the product features in WAUP can also be more accurate. At the same time, the application of contrastive learning further enhances the information interaction between products, thereby capturing more fine-grained information connections, helping to alleviate noise and enhance feature representation.

[0284] S902. Obtain corresponding implicit preference features based on the historical product usage data of each sample object, and obtain selection loss based on the implicit preference features and true implicit preference features of each sample object.

[0285] The selection loss is used to measure the loss in the second sub-model.

[0286] S903. Based on the explicit preference features and implicit preference features corresponding to each sample object, a corresponding key preference information sequence is obtained, and based on the obtained key preference information sequence corresponding to each sample object and the real key preference information sequence, a ranking loss and a generation loss are obtained. The ranking loss is used to measure the loss of the first stage in the third sub-model, and the generation loss is used to measure the loss of the second stage in the third sub-model.

[0287] (1) Sorting loss

[0288] In a first possible implementation, the server performs the following operations for each sample object: based on the explicit preference features and implicit preference features of the sample object, evaluates various labels to obtain label evaluation values ​​corresponding to various labels, and obtains the sub-ranking loss of the sample object based on the label evaluation values ​​corresponding to various labels and the true evaluation values ​​corresponding to various labels. After obtaining the sub-ranking loss of each sample object, the server obtains the ranking loss in the PRT based on the sub-ranking loss of each sample object.

[0289] A is used to represent the aspect label set, which contains preset aspect labels. i represents the label evaluation value of the i-th aspect label in the aspect label sequence generated by PRT, using l i represents the true evaluation value of the i-th aspect label in the aspect label sequence generated by PRT, G is used to represent the set of label evaluation values, and L is used to represent the set of true evaluation values.

[0290] The ranking loss is calculated using the following formula (6):

[0291]

[0292] Among them, L(G,L) represents the sorting loss, l i > j , σ is a hyperparameter, and n represents the number of aspect labels contained in the aspect label set.

[0293] The ranking loss in the first possible implementation satisfies l i > j , but g i <g j The (i, j) pair is penalized, which reduces the amount of calculation to a certain extent. However, since the loss function cannot be directly used as an evaluation index of the model, the model trained with the traditional regression loss function may not be able to achieve good performance in the sorting index. Ideally, the sorting algorithm should use the sorting index as a learning goal, but the sorting index is usually discrete and cannot be directly used as a loss function for model optimization. In view of the above problems, a second possible implementation method is proposed in the embodiment of the present application.

[0294] In the second possible implementation, a ranking indicator is introduced into the traditional regression loss: Normalized Discounted Cumulative Gain (NDCG) to reweight each (i, j) in the sequence so that the model is optimized in the direction of improving the ranking indicator.

[0295] The ranking loss is calculated using the following formula (7):

[0296]

[0297] Among them, L r represents the ranking loss, and ΔNDGG represents the change value of the ranking indicator NDCG when aspect label i and aspect label j are exchanged.

[0298] By introducing the ranking loss, in the first stage of PRT, a label sequence of aspects of interest to the object can be generated sequentially. The higher the generation probability, the higher the degree of interest.

[0299] (2) Generation loss

[0300] In a possible implementation, the server performs the following operations for each sample object: based on the semantic information corresponding to each of the k aspect labels, the first k aspect labels in each aspect label are generated, each corresponding to at least one key preference information, and based on the order between the first k aspect labels, the information order is obtained, and based on the generated key preference information and the information order between them, combined with at least one real key preference information corresponding to each of the first k aspect labels and the information order between them, the sub-generation loss of the sample object is obtained. Then, the server obtains the generation loss in the PRT based on the sub-generation loss of each sample object. The first k aspect labels are determined according to the label evaluation values ​​of the aspect labels.

[0301] Exemplarily, the generation loss is calculated using the following formula (8):

[0302]

[0303] Among them, y represents the predicted behavior label, that is, the generated key preference information, y * Indicates the real behavior label, represents the probability of generating a true key preference information sequence, and T represents the maximum number of key preference information.

[0304] S904: Obtain a total loss value based on the contrast loss, selection loss, sorting loss, and generation loss.

[0305] Exemplarily, the total loss can be calculated by the following formula (9):

[0306] L=λ1L c +λ2L s +λ3L r +λ4L g Formula (9)

[0307] Among them, L represents the total loss value, λ1, λ2, λ3, and λ4 are all hyperparameters, among which λ1 represents the weight of contrast loss, λ2 represents the weight of selection loss, λ3 represents the weight of sorting loss, and λ4 represents the weight of generation loss.

[0308] S905: Determine whether the model convergence condition is met. If so, output the trained label generation model, otherwise execute S906.

[0309] In the embodiment of the present application, the model convergence condition may include at least one of the following conditions:

[0310] (1) The total loss value is not greater than the preset loss value threshold.

[0311] (2) The number of iterations reaches the preset upper limit.

[0312] S906: Adjust model parameters based on the total loss value.

[0313] If the above model convergence conditions are met, it is determined that the model has met the convergence conditions and the training is completed. Otherwise, it is determined that the model has not yet met the model convergence conditions, then it is necessary to continue to adjust the model parameters and use the adjusted model to enter the next training process, that is, jump to step S901 to execute.

[0314] See also Fig.11 As shown, it is a logical diagram of a model training process provided in an embodiment of the present application. First, based on the historical product feedback data of each sample object, the server obtains the explicit preference features corresponding to each sample object through the information aggregation method of the WAU meta-path, and then, based on the explicit preference features corresponding to each sample object, combined with the historical product feedback data of each sample object, the product features obtained through the information aggregation method of the WAPU meta-path are used to obtain the selection loss in REPT.

[0315] Next, the server obtains the corresponding implicit preference features based on the historical product usage data of each sample object, and obtains the selection loss in BIPT based on the implicit preference features and true implicit preference features of each sample object.

[0316] Next, the server obtains the corresponding aspect label sequence based on the explicit preference features and implicit preference features corresponding to each sample object, and uses the aspect label sequence to generate the corresponding key preference information sequence, and obtains the generation loss based on the obtained aspect label sequence corresponding to each sample object, and obtains the generation loss based on the generated key preference information sequence corresponding to each sample object.

[0317] In some embodiments, the client may present a total amount input interface, and in response to an input operation triggered by the target object in the total amount input interface, present recommendation information of a target recommended product, wherein the target recommended product is determined based on the exchange resource differences between the candidate management products, and the exchange resource differences are determined based on the input total amount of resource input and the exchange resource calculation method corresponding to each candidate management product.

[0318] The total amount input interface is an interface for inputting the total amount of resource input. The recommended information can be presented in the form of a prompt box, pop-up window, etc. in the total amount input interface, or in an independent interface, without limitation.

[0319] In a possible implementation, the client presents a total quantity input interface for a designated management product, where the designated management product is a candidate management product other than the target management product among the candidate management products. In response to an input operation triggered by a target object in the total quantity input interface, the client presents the recommendation information of the target recommended product in the total quantity input interface, and may also present a difference viewing interface in response to a difference viewing operation triggered by the target object for the target recommended product, wherein the difference viewing interface includes: an exchange resource difference and a product switching component; and presenting a total quantity input interface for the target recommended product in response to a trigger operation of the target object for the product switching component.

[0320] As an example, the client can present a difference viewing component in the total amount input interface while presenting the target recommended products. When the target object triggers the difference viewing component, the client determines the triggering operation of the target object for the difference viewing component.

[0321] The exchange resource differences may be presented in the form of text, charts, tables, etc., but are not limited thereto.

[0322] For example, see Fig.12 As shown, it is a schematic diagram of the interface of an asset management product recommendation process provided in an embodiment of the present application. Fig.12The operation interface 11, the operation interface 12 and the operation interface 13 are included. The operation interface 11 is a total amount input interface, which is triggered by the target object for the Class A shares of the xx fund. The total amount input interface includes an input component for receiving the total amount of resource investment. When the target object inputs the total amount of resource investment through the input component, the client responds to the input operation and presents the recommended information "hold for less than 361 days, buy Class C funds, the fee is lower" on the operation interface 11. In addition, the operation interface 11 also includes a "view details" button, which is equivalent to a difference viewing component. After the target object clicks the "view details" button, the client responds to the difference viewing operation triggered by the target object for the target recommended product and jumps from the operation interface 11 to the operation interface 122.

[0323] The operation interface 12 is a difference viewing interface, which includes a chart, a "continue to buy Class A shares" button, and a "buy Class C shares" button, wherein the chart is used to characterize the difference in exchange resources, and the "buy Class B shares" button is a product switching component. In the chart, the horizontal axis in the figure represents the holding period of the candidate management product. Assume that the breakeven point is 361 days. When the holding period of the candidate management product exceeds 361 days, the exchange resources of Class A shares are equal to the exchange resources of Class C shares. When the holding period of the candidate management product exceeds 361 days, the exchange resources of Class A shares are lower than the exchange resources of Class C shares. When the holding period of the candidate management product is less than 361 days, the exchange resources of Class C shares are lower than the exchange resources of Class A shares. When the target object clicks the "buy Class C shares" button, the client responds to the trigger operation of the target object on the product switching component and presents the operation interface 13.

[0324] The operation interface 13 is an interface for inputting the total amount of Class C shares, and the interface for inputting the total amount of Class C shares includes an input component for receiving the total amount of resource input.

[0325] In one possible implementation, in response to a product information viewing operation triggered by the target object for the target recommended product, the client presents a product information interface for the target recommended product, wherein the product information interface includes a product comparison component; in response to a product information viewing operation triggered by the target object for the product comparison component, the exchange resource differences are presented in the product information interface.

[0326] For example, see Fig.13 As shown, it is a schematic diagram of the interface of an asset management product recommendation process provided in an embodiment of the present application. Fig.13 The operation interface 21, the operation interface 22 and the operation interface 23 are included.

[0327] Operation interface 21 is a product information viewing operation triggered by the target object for Class C shares of xx fund, presenting the product information interface of Class C shares of xx fund. The product information interface includes a "View AC Rate Comparison" button, which is equivalent to a product comparison component. When the target object clicks the "View AC Rate Comparison" button, the client jumps from operation interface 21 to operation interface 22. The product information interface also includes product-related information such as the buying rules and withdrawal rules of Class C shares of xx fund.

[0328] Operation interface 22 is for viewing product information. In the product information interface, the differences in exchange resources are presented in the form of a chart. In addition, the calculation basis of the exchange resources corresponding to Class A shares of xx fund and Class C shares of xx fund are also presented.

[0329] In addition, the operation interface 22 also includes a "View Fund PK Button". The "View Fund PK Button" can also be regarded as a comparison component. The client responds to the trigger operation of the target object on the comparison component and presents the operation interface 23. The operation interface 23 includes detailed resource transfer information of each management product being compared.

[0330] It should be understood that the operations in the embodiments of the present application (such as trigger operations, input operations, etc.) may include contact operations such as clicking, long pressing, and sliding, and may also include non-contact operations such as voice and gestures, without limitation.

[0331] Below, the present application is described in conjunction with a specific embodiment.

[0332] See also Fig.14 As shown, the asset management recommendation process includes the following steps:

[0333] S141. The target object enters the total amount of resource investment and the target holding period in the total amount input interface for xx Fund A presented on the client.

[0334] S142: The client determines whether the total amount of resource investment reaches the investment threshold set for the target object. If so, the subsequent comparison process is performed, otherwise the process ends.

[0335] S143: The client prompts the target object with a share comparison prompt.

[0336] S144. The client inputs the total amount of resource investment and the target holding period of the target object for xx Fund A and sends them to the server.

[0337] S145. The server calculates the difference in exchange resources.

[0338] Assume that xx Fund has four shares, A, B, C, and D. Among them, the server takes xx Fund A, xx Fund B, xx Fund C, and xx Fund D as candidate management products, and obtains the exchange resource difference between xx Fund A, xx Fund B, xx Fund C, and xx Fund D based on the exchange resource calculation method corresponding to each of xx Fund A, xx Fund B, xx Fund C, and xx Fund D. The exchange resource difference includes: the reference duration of each candidate management product. The calculation method of the exchange resource difference is specifically referred to above.

[0339] S146. The server selects a target management product.

[0340] Based on the difference in exchange resources and the key preference information sequence of the target object, the server selects xx Fund C as the target management product from xx Fund A, xx Fund B, xx Fund C, and xx Fund D. The method for selecting the target management product is described above.

[0341] S147. Generate recommendation information for target management products.

[0342] The server generates recommendation information for xx Fund C. The recommendation information for xx Fund C includes a decision label "Short-term investment recommendation Class C" and may also include a decision label "Long-term investment recommendation Class A".

[0343] S148. The server sends the recommendation information to the client.

[0344] S149: The client presents recommendation information.

[0345] The server sends the recommendation information of xx Fund C to the client, so that the client presents the recommendation information of xx Fund C in the total amount input interface.

[0346] Furthermore, the total amount input interface also includes a "view details" button. When the target object clicks the "view details" button, the client presents a difference viewing interface, which includes: exchange resource differences and product switching components.

[0347] Furthermore, the client may also present a total quantity input interface for the target recommended product in response to the triggering operation of the target object on the product switching component.

[0348] In addition, the system can continuously update the target object's key preference information sequence through the target object's product feedback and usage data, thereby providing personalized investment advice and enhancing user experience.

[0349] In the embodiment of the present application, the exchange resource comparison and label recommendation mechanism significantly improves the transparency and convenience of decision-making, optimizes resource gain, enhances user experience, and facilitates the selection of resource products by the subject. At the same time, the system's highly personalized services, rapid response capabilities, and regular update timeliness ensure the accuracy and adaptability of investment advice, strengthen risk management, promote the popularization of financial information, support flexible strategy recommendations, and increase the subject's stickiness to the platform, which generally improves the intelligence level of the resource management product recommendation process and the overall satisfaction of the subject.

[0350] Based on the same inventive concept, the present application embodiment provides an asset management product recommendation device. Fig.15 As shown, it is a schematic diagram of the structure of the asset management product recommendation device 1500, which may include:

[0351] Information acquisition unit 1501, used to acquire the total amount of resource input of the target object;

[0352] The information analysis unit 1502 is used to perform resource evaluation on each candidate management product based on the total resource input and the exchange resource calculation method corresponding to each candidate management product, so as to obtain the exchange resource difference between the candidate management products;

[0353] The information recommendation unit 1503 is configured to select a target management product recommended to the target object from the candidate management products based on the exchange resource difference and in combination with the object information of the target object.

[0354] In a possible implementation, when performing resource evaluation on each candidate management product based on the total resource input and the exchange resource calculation method corresponding to each candidate management product to obtain the exchange resource difference between the candidate management products, the information analysis unit 1502 is specifically used to:

[0355] Based on the total amount of resource input and in combination with the exchange resource calculation method corresponding to each candidate management product, resource evaluation is performed on each candidate management product to obtain a reference duration, wherein the reference duration represents: the product holding duration required when the exchange resources of each candidate management product are equal;

[0356] Based on the reference duration, the exchange resource difference is obtained.

[0357] In a possible implementation, the information analysis unit 1502 is further configured to:

[0358] Based on the target holding time of the target object and the total amount of resource investment, combined with various exchange resource calculation methods, obtaining the exchange resources of each candidate management product within the target holding time;

[0359] The exchange resource difference is obtained based on the reference duration and in combination with the obtained resource difference between the exchange resources.

[0360] In a possible implementation, the information recommendation unit 1503 is further configured to:

[0361] Based on the target management product and the exchange resource difference, generating recommendation information, wherein the recommendation information is represented in the form of a tag;

[0362] The recommendation information is sent to a terminal device, so that the terminal device presents the recommendation information to the target object.

[0363] In a possible implementation, if the object information includes: each key preference information corresponding to the target object and the information sequence therebetween, the information sequence is used to represent the importance of each key preference information;

[0364] When selecting a target management product recommended to the target object from the candidate management products based on the exchange resource difference and in combination with the object information of the target object, the information recommendation unit 1503 is specifically used to:

[0365] Based on the exchange resource differences, combined with the key preference information and the information order therebetween, obtaining matching evaluation information between each of the candidate management products and the target object;

[0366] Based on the obtained matching evaluation information, a candidate management product that meets the set matching condition is selected from the candidate management products as the target management product.

[0367] In a possible implementation, the information recommendation unit 1503 is further configured to:

[0368] If the custom preference information of the target object is set, obtaining the object information based on the custom preference information;

[0369] If the customized preference information is not set, preference inference is performed based on the historical product usage data and historical product feedback data of the target object to obtain the object information.

[0370] In a possible implementation, when the preference inference is performed based on the historical product usage data and historical product feedback data of the target object and the object information is obtained, the information recommendation unit 1503 is specifically used to:

[0371] Based on the historical product feedback data, obtaining the explicit preference characteristics of the target object, and based on the historical product usage data, obtaining the implicit preference characteristics of the target object;

[0372] Based on the explicit preference feature and the implicit preference feature, combined with preset various aspects of labels, the object information is obtained.

[0373] In a possible implementation, when the object information is obtained based on the explicit preference feature and the implicit preference feature in combination with various preset labels, the information recommendation unit 1503 is specifically configured to:

[0374] Based on the explicit preference feature and the implicit preference feature, obtain a label evaluation value between each of the aspect labels and the target object, and based on the obtained label evaluation values, screen out an aspect label sequence including k aspect labels from the aspect labels;

[0375] Based on the semantic information corresponding to each of the k aspect labels, corresponding key preference information is generated, and based on the label order of the K aspect labels in the aspect label sequence, the information order between the generated key preference information is obtained.

[0376] In a possible implementation, when obtaining the explicit preference characteristics of the target object based on the historical product feedback data, the information recommendation unit 1503 is specifically configured to:

[0377] Extracting each word from the historical product feedback data, and encoding each word to obtain word features corresponding to each word;

[0378] Based on the association between the aspect labels and the words, information aggregation is performed in combination with the word features corresponding to each of the words to obtain the aspect label features corresponding to each of the aspect labels;

[0379] Based on the association between the product and the aspect label, information aggregation is performed in combination with the aspect label features corresponding to each aspect label to obtain the product features corresponding to each historical management product in the historical product feedback data;

[0380] Based on the association relationship between products and objects, information aggregation is performed in combination with the product features corresponding to each of the historical management products to obtain the explicit preference features of the target object.

[0381] In a possible implementation, when performing resource evaluation on each candidate management product based on the total resource input and the exchange resource calculation method corresponding to each candidate management product to obtain the exchange resource difference between the candidate management products, the information analysis unit 1502 is specifically used to:

[0382] When the total amount of resource investment reaches the investment threshold set for the target object, resource evaluation is performed based on the total amount of resource investment and each exchange resource calculation method to obtain the exchange resource difference; wherein the investment threshold is the first threshold or the second threshold, and the first threshold is higher than the second threshold.

[0383] Based on the same inventive concept, the present application embodiment provides an asset management product recommendation device. Fig.16 As shown, it is a schematic diagram of the structure of the asset management product recommendation device 1600, which may include:

[0384] The first presentation unit 1601 is used to present a total amount input interface;

[0385] The second presentation unit 1602 is used to present the recommended information of the target management product in response to the input operation triggered by the target object in the total amount input interface. The target management product is determined based on the exchange resource differences between the candidate management products. The exchange resource differences are determined based on the input total amount of resource input and the exchange resource calculation method corresponding to each candidate management product.

[0386] In a possible implementation, when presenting the total amount input interface, the first presenting unit 1601 is specifically used to:

[0387] Presenting a total amount input interface for a designated management product, where the designated management product is a candidate management product among the candidate management products except the target management product;

[0388] After presenting the recommended information of the target management product in the total amount input interface, the second presenting unit 1602 is further used to:

[0389] In response to the difference viewing operation triggered by the target object, a difference viewing interface is presented, the difference viewing interface including: exchange resource difference and product switching components;

[0390] In response to the triggering operation of the target object on the product switching component, a total amount input interface for the target management product is presented.

[0391] In a possible implementation, the second presenting unit 1602 is further configured to:

[0392] In response to a product information viewing operation triggered by the target object for the target recommended product, presenting a product information interface of the target recommended product; the product information interface includes a product comparison component;

[0393] In response to a product information viewing operation triggered by the target object for the product comparison component, the exchange resource difference is presented in the product information interface.

[0394] For the convenience of description, the above parts are divided into modules (or units) according to their functions and described separately. Of course, when implementing this application, the functions of each module (or unit) can be implemented in the same or multiple software or hardware.

[0395] Regarding the device in the above embodiment, the specific manner in which each unit executes the request has been described in detail in the embodiment of the method, and will not be elaborated here.

[0396] Those skilled in the art will appreciate that various aspects of the present application may be implemented as a system, method or program product. Therefore, various aspects of the present application may be specifically implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software, which may be collectively referred to as "circuit", "module" or "system" herein.

[0397] Based on the same inventive concept, the embodiment of the present application also provides an electronic device. In one embodiment, the electronic device can be a server or a terminal device. Fig.17 As shown, it is a schematic diagram of the structure of a possible electronic device provided in an embodiment of the present application, Fig.17 In the embodiment, the electronic device 1700 includes: a processor 1710 and a memory 1720.

[0398] Among them, the memory 1720 stores a computer program that can be executed by the processor 1710, and the processor 1710 can execute the steps of the above-mentioned asset management product recommendation method by executing the instructions stored in the memory 1720.

[0399] The memory 1720 may be a volatile memory, such as a random-access memory (RAM); the memory 1720 may also be a non-volatile memory, such as a read-only memory (ROM), a flash memory, a hard disk drive (HDD) or a solid-state drive (SSD); or the memory 1720 may be any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 1720 may also be a combination of the above memories.

[0400] The processor 1710 may include one or more central processing units (CPU) or a digital processing unit, etc. The processor 1710 is configured to implement the asset management product recommendation method when executing the computer program stored in the memory 1720 .

[0401] In some embodiments, the processor 1710 and the memory 1720 may be implemented on the same chip, and in some embodiments, they may also be implemented on separate chips.

[0402] The specific connection medium between the processor 1710 and the memory 1720 is not limited in the embodiment of the present application. In the embodiment of the present application, the processor 1710 and the memory 1720 are connected via a bus as an example. Fig.17 The connections between the other components are only for illustration and are not intended to be limiting. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of description, Fig.17 The diagram shows that only one thick line is used, but this does not mean that there is only one bus or only one type of bus.

[0403] Based on the same inventive concept, an embodiment of the present application provides a computer-readable storage medium, which includes a computer program. When the computer program is executed on an electronic device, the computer program is used to enable the electronic device to execute the steps of the above-mentioned asset management product recommendation method. In some possible implementations, various aspects of the asset management product recommendation method provided by the present application can also be implemented in the form of a program product, which includes a computer program. When the program product is executed on an electronic device, the computer program is used to enable the electronic device to execute the steps in the above-mentioned asset management product recommendation method. For example, the electronic device can execute the following steps: Figure 2 Follow the steps shown in .

[0404] The program product may use any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable disk, a hard disk, a RAM, a ROM, an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (Compact Disk Read Only Memory, CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0405] The program product of the embodiment of the present application may adopt a CD-ROM and include a computer program, and can be run on an electronic device. However, the program product of the present application is not limited thereto, and in this document, a readable storage medium may be any tangible medium containing or storing a computer program, which may be used by or in combination with a command execution system, apparatus or device.

[0406] A readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, wherein a readable computer program is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A readable signal medium may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a computer program for use by or in conjunction with a command execution system, apparatus, or device.

[0407] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0408] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A method for recommending asset management products, characterized in that: include: Get the total amount of resource input of the target object; Based on the total amount of resource input and in combination with the exchange resource calculation method corresponding to each candidate management product, resource evaluation is performed on each candidate management product to obtain the exchange resource difference between the candidate management products; Based on the exchange resource difference and in combination with the object information of the target object, a target management product recommended to the target object is selected from the candidate management products.

2. The method according to claim 1, characterized in that The resource evaluation of each candidate management product is performed based on the total resource input and the exchange resource calculation method corresponding to each candidate management product to obtain the exchange resource difference between the candidate management products, including: Based on the total amount of resource investment and in combination with the exchange resource calculation method corresponding to each candidate management product, resource evaluation is performed on each candidate management product to obtain a reference duration, wherein the reference duration represents: the product holding duration required when the exchange resources of each candidate management product are equal; Based on the reference duration, the exchange resource difference is obtained.

3. The method according to claim 2, characterized in that Before obtaining the exchange resource difference based on the reference duration, the method further includes: Based on the target holding time of the target object and the total amount of resource investment, in combination with various exchange resource calculation methods, obtaining the exchange resources of each candidate management product within the target holding time; Based on the reference duration, obtaining the exchange resource difference includes: The exchange resource difference is obtained based on the reference duration and in combination with the obtained resource difference between the exchange resources.

4. The method according to claim 1, 2 or 3, characterized in that: After selecting a target management product recommended to the target object from the candidate management products based on the exchange resource difference and in combination with the object information of the target object, the method further includes: Based on the target management product and the exchange resource difference, generating recommendation information, wherein the recommendation information is represented in the form of a tag; The recommendation information is sent to a terminal device, so that the terminal device presents the recommendation information to the target object.

5. The method according to claim 1, 2 or 3, characterized in that: If the object information includes: each key preference information corresponding to the target object and the information sequence therebetween, the information sequence is used to represent the importance of each key preference information; Then, based on the exchange resource difference and in combination with the object information of the target object, selecting a target management product recommended to the target object from the candidate management products includes: Based on the exchange resource differences, combined with the key preference information and the information order therebetween, obtaining matching evaluation information between each of the candidate management products and the target object; Based on the obtained matching evaluation information, a candidate management product that meets the set matching condition is selected from the candidate management products as the target management product.

6. The method according to claim 5, characterized in that The object information is obtained by: If the custom preference information of the target object is set, obtaining the object information based on the custom preference information; If the customized preference information is not set, preference inference is performed based on the historical product usage data and historical product feedback data of the target object to obtain the object information.

7. The method according to claim 6, characterized in that The performing preference inference based on the historical product usage data and historical product feedback data of the target object to obtain the object information includes: Based on the historical product feedback data, obtaining the explicit preference characteristics of the target object, and based on the historical product usage data, obtaining the implicit preference characteristics of the target object; Based on the explicit preference feature and the implicit preference feature, combined with preset various aspects of labels, the object information is obtained.

8. The method according to claim 7, characterized in that The obtaining of the object information based on the explicit preference feature and the implicit preference feature in combination with various preset labels includes: Based on the explicit preference feature and the implicit preference feature, obtain a label evaluation value between each of the aspect labels and the target object, and based on the obtained label evaluation values, screen out an aspect label sequence including k aspect labels from the aspect labels; Based on the semantic information corresponding to each of the k aspect labels, corresponding key preference information is generated, and based on the label order of the K aspect labels in the aspect label sequence, the information order between the generated key preference information is obtained.

9. The method according to claim 7, characterized in that The obtaining the explicit preference characteristics of the target object based on the historical product feedback data includes: Extracting each word from the historical product feedback data, and encoding each word to obtain word features corresponding to each word; Based on the association between the aspect labels and the words, information aggregation is performed in combination with the word features corresponding to each of the words to obtain the aspect label features corresponding to each of the aspect labels; Based on the association between the product and the aspect label, information aggregation is performed in combination with the aspect label features corresponding to each aspect label to obtain the product features corresponding to each historical management product in the historical product feedback data; Based on the association relationship between products and objects, information aggregation is performed in combination with the product features corresponding to each of the historical management products to obtain the explicit preference features of the target object.

10. The method according to claim 1, 2 or 3, characterized in that: The step of performing resource evaluation on each candidate management product based on the total resource input and in combination with the exchange resource calculation method corresponding to each candidate management product to obtain the exchange resource difference between the candidate management products includes: When the total amount of resource investment reaches the investment threshold set for the target object, resource evaluation is performed based on the total amount of resource investment and each exchange resource calculation method to obtain the exchange resource difference; wherein the investment threshold is the first threshold or the second threshold, and the first threshold is higher than the second threshold.

11. A method for recommending asset management products, characterized in that: include: The total amount input interface is presented; In response to the input operation triggered by the target object in the total amount input interface, recommended information of the target management product is presented. The target management product is determined based on the exchange resource differences between the candidate management products. The exchange resource differences are determined based on the input total amount of resource input and the exchange resource calculation method corresponding to each candidate management product.

12. The method according to claim 11, characterized in that The presenting total amount input interface includes: Presenting a total amount input interface for a designated management product, where the designated management product is a candidate management product among the candidate management products except the target management product; After presenting the recommended information of the target management product in the total amount input interface, the method further includes: In response to the difference viewing operation triggered by the target object, a difference viewing interface is presented, the difference viewing interface including: exchange resource difference and product switching components; In response to the triggering operation of the target object on the product switching component, a total amount input interface for the target management product is presented.

13. The method according to claim 11, characterized in that Also includes: In response to a product information viewing operation triggered by the target object for the target recommended product, presenting a product information interface of the target recommended product; The product information interface includes a product comparison component; In response to a product information viewing operation triggered by the target object for the product comparison component, the exchange resource difference is presented in the product information interface.

14. An asset management product recommendation device, characterized in that: include: An information acquisition unit, used to acquire the total amount of resource input of the target object; An information analysis unit, configured to perform resource evaluation on each candidate management product based on the total resource input and in combination with the exchange resource calculation method corresponding to each candidate management product, to obtain the exchange resource difference between the candidate management products; The information recommendation unit is used to select a target management product recommended to the target object from the candidate management products based on the exchange resource difference and in combination with the object information of the target object.

15. An asset management product recommendation device, characterized in that: include: A first presentation unit, used to present a total amount input interface; The second presentation unit is used to present recommendation information of the target management product in response to the input operation triggered by the target object in the total amount input interface. The target management product is determined based on the exchange resource differences between the candidate management products. The exchange resource differences are determined based on the input total amount of resource input and the exchange resource calculation method corresponding to each candidate management product.

16. An electronic device, characterized in that: It includes a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of any method described in claims 1 to 10, or executes the steps of any method described in claims 11 to 13.

17. A computer-readable storage medium, characterized in that: It includes a computer program. When the computer program is run on an electronic device, the computer program is used to enable the electronic device to execute the steps of any method described in claims 1 to 10, or execute the steps of any method described in claims 11 to 13.

18. A computer program product, characterized in that It includes a computer program, which is stored in a computer-readable storage medium. The processor of the electronic device reads and executes the computer program from the computer-readable storage medium, so that the electronic device performs the steps of any method described in claims 1 to 10, or performs the steps of any method described in claims 11 to 13.

Images