Product recommendation method and apparatus, device, and readable storage medium

Abstract

Embodiments of the present application relate to the technical field of computers, and provide a product recommendation method and a related product. The product recommendation method comprises: on the basis of a target operation performed by a user on a target page, acquiring a plurality of product recommendation strategies corresponding to a target recommendation slot on the target page; on the basis of first user information of the user from a product consignment platform and second user information of the user from a financial institution platform, performing federated prediction of recommendation effects of the product recommendation strategies for the user, so as to obtain recommendation effect information corresponding to the product recommendation strategies; on the basis of the recommendation effect information corresponding to the product recommendation strategies, determining a target product recommendation strategy from among the plurality of product recommendation strategies; and on the basis of the target product recommendation strategy, generating product recommendation information for the target recommendation slot. The method and product provided by the embodiments of the present application can improve the accuracy of product recommendations.

Description

Product recommendation method, device, equipment and readable storage medium Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method for product recommendation and related products. Background Art

[0002] With the development of third-party platforms, such as third-party payment platforms, the use of third-party platforms as consignment channels for product sales has gradually begun to be applied. In related art, when selling products on consignment, third-party platforms will set product recommendation information on the relevant pages of the platform to recommend the products they are selling on consignment. This product recommendation information is usually determined by the third-party platform based on its own user information. However, the user information owned by third-party platforms is limited, resulting in poor product recommendation accuracy. Summary of the Invention

[0003] The embodiments of the present application provide a product recommendation method and related products to improve the accuracy of product recommendations on a product distribution platform.

[0004] In a first aspect, an embodiment of the present application provides a product recommendation method, which is applied to a product distribution platform, and the method includes: obtaining multiple product recommendation strategies corresponding to a target recommendation position in the target page based on the user's target operation on the target page; based on the first user information of the user of the product distribution platform and the second user information of the user of the financial institution platform, performing a federal prediction on the recommendation effect of the product recommendation strategy on the user to obtain recommendation effect information corresponding to the product recommendation strategy; determining a target product recommendation strategy among the multiple product recommendation strategies based on the recommendation effect information corresponding to the product recommendation strategy; and generating product recommendation information for the target recommendation position based on the target product recommendation strategy.

[0005] In a second aspect, an embodiment of the present application provides a product recommendation device, which is applied to a product distribution platform, and the device includes: a strategy acquisition module, which is used to obtain multiple product recommendation strategies corresponding to a target recommendation position in the target page according to the user's target operation on the target page; an effect prediction module, which is used to perform a federal prediction of the recommendation effect of the product recommendation strategy on the user based on the first user information of the user of the product distribution platform and the second user information of the user of the financial institution platform, and obtain recommendation effect information corresponding to the product recommendation strategy; a target strategy determination module, which is used to determine a target product recommendation strategy from the multiple product recommendation strategies according to the recommendation effect information corresponding to the product recommendation strategy; and a recommendation information generation module, which is used to generate product recommendation information for the target recommendation position according to the target product recommendation strategy.

[0006] In a third aspect, an embodiment of the present application provides a computer device comprising a processor and a memory for storing instructions executable by the processor; wherein the processor is configured to execute the instructions to implement the product recommendation method provided in the first aspect of the embodiment of the present application.

[0007] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having computer instructions stored thereon, wherein the computer instructions are configured to, when executed by a processor of a computer device, enable the computer device to execute the product recommendation method provided in the first aspect of the embodiment of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in one or more of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0009] FIG1 is a schematic diagram of an application scenario of a product recommendation method provided in an embodiment of the present application.

[0010] FIG2 is a flow chart of a product recommendation method provided in an embodiment of the present application.

[0011] FIG3 is a recommendation example of the product recommendation method shown in FIG2 .

[0012] FIG4 is a module diagram of a product recommendation device provided in an embodiment of the present application.

[0013] FIG5 is a schematic structural diagram of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0014] In order to enable those skilled in the art to better understand the technical solutions in one or more of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of one or more embodiments of the present application, rather than all embodiments. Based on one or more embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0015] It should be noted that, in the absence of conflict, one or more embodiments and features in the embodiments of the present application may be combined with each other. The embodiments of the present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0016] FIG1 is a schematic diagram of an application scenario of a product recommendation method provided by an embodiment of the present application. The application scenario shown in FIG1 includes a user device 100 and a product distribution platform 200. The product distribution platform is used to distribute products on behalf of others. It can be a third-party platform, such as a third-party payment platform. Users can browse the page of the product distribution platform 200 through the user device 100. When the user browses to the target page, the product distribution platform 200 can display product recommendation information for the user in the preset recommendation position of the target page to recommend products to the user.

[0017] The product recommendation method provided in the embodiment of the present application is used to generate product recommendation information, which can be used in the product distribution platform shown in Figure 1. The product can be a financial product, such as a financial product such as a fund product, and the product distribution platform can be a platform for distributing financial products, such as a third-party payment platform.

[0018] FIG2 is a product recommendation method provided in an embodiment of the present application. Now, with reference to FIG2 , the product recommendation method provided in an embodiment of the present application will be described in detail.

[0019] As shown in FIG2 , the product recommendation method provided in the embodiment of the present application may include the following processing.

[0020] S102: Acquire multiple product recommendation strategies corresponding to target recommendation positions in the target page according to the user's target operation on the target page.

[0021] The target page is a page with a preset recommendation slot, which is an area on the page used to display recommended information. The target recommendation slot is a recommendation slot on the target page used to display product recommendations. The product recommendations in this slot can be generated in response to a user's target action on the target page, such as entering the target page.

[0022] The product recommendation information displayed in the target recommendation slot is generated based on the target product recommendation strategy corresponding to the user and changes with the user. The target product recommendation strategy corresponding to the user is the recommendation strategy that has the best recommendation effect on the user and is selected from the multiple product recommendation strategies corresponding to the target recommendation slot. There are multiple product recommendation strategies corresponding to the target recommendation slot, and the product recommendation strategies can be recommendation strategies for various products, such as recommendation strategies for various financial products. The multiple product recommendation strategies corresponding to the target recommendation slot can include recommendation strategies corresponding to multiple different products.

[0023] The product recommendation strategy is used to generate product recommendation information, which may include the first product feature information of the product, such as the product name, issuing organization, rate of return and other information. The product recommendation strategy can be obtained from the recommendation strategy library based on the recommendation position information of the target recommendation position, wherein the recommendation position information is, for example, the scenario information of the recommendation position, and the recommendation strategy library is a pre-built database that contains multiple product recommendation strategies and their corresponding scenarios. In specific implementation, the target recommendation position in the target page can be determined based on the user's entry operation on the target page, and the recommendation strategy database can be retrieved based on the recommendation position information of the target recommendation position, thereby obtaining multiple product recommendation strategies corresponding to the target recommendation position.

[0024] S104, based on the first user information of the user on the product distribution platform and the second user information of the user on the financial institution platform, a federal prediction is made on the recommendation effect of the product recommendation strategy on the user to obtain recommendation effect information corresponding to the product recommendation strategy.

[0025] The first user information is user information on the product distribution platform, and the second user information is user information on the financial institution platform. These two user information types have different feature data. The product distribution platform is a platform used for product distribution, such as a third-party payment platform. The first user information on the product distribution platform can be user profile information on the payment platform, such as the user's education level, occupation, region, payment preferences, etc. The financial institution can be the institution that issues the product, and the second user information on the financial institution platform can be the user's financial information, such as the user's product purchase and transaction information.

[0026] It is understandable that a product distribution platform can distribute multiple products. A product can be distributed on multiple product distribution platforms. For a product distribution platform, it does not have the purchase information of users on other distribution platforms. It is impossible to make a relatively accurate prediction of the recommendation effect of the product recommendation strategy based solely on the product distribution platform's own data. However, the financial institution that issues the product will have data such as the user's financial information on other distribution platforms. The prediction of the product recommendation strategy on the user's recommendation effect depends largely on the number and quality of features of the user information on which the prediction is based. By conducting a federal prediction of the recommendation effect of the product recommendation strategy based on the user information of the product distribution platform and the user information of the financial institution platform, it is possible to predict a more accurate recommendation effect based on more comprehensive user information, thereby making the product recommendation information more targeted to the user and improving the recommendation effect on the user, for example, the conversion rate of the recommendation information.

[0027] "Federated Prediction" refers to the use of local models of multiple participants to make predictions without centralizing or sharing the original data. Federated Prediction is closely related to "Federated Learning". Federated Learning is a machine learning framework that allows multiple participants to retain their own data locally while jointly training a global model. This method can use multiple data sources for machine learning while protecting user privacy and data security. Since the user information required to predict the recommendation effect of a product recommendation strategy on a user is the user's sensitive financial data, such as the first user information of the above-mentioned product distribution platform and the second user information of the above-mentioned financial institution platform, these data are the user's private data. By making a federated prediction of the recommendation effect of a product recommendation strategy on a user based on the first user information of the user on the product distribution platform and the second user information of the user on the financial institution platform, it is possible to protect the privacy of user information, enrich the richness of the financial data based on which the recommendation effect of the product recommendation strategy on the user is predicted, and improve the accuracy of the recommendation without leaving the domain.

[0028] In specific implementations, the product distribution platform and the financial institution platform can conduct federated learning to obtain a target recommendation effect prediction model. This target recommendation effect prediction model is used to perform privacy calculations on the recommendation effects of product recommendation strategies on users based on the first user information of users on the product distribution platform and the second user information of users on the financial institution platform, without data leaving the domain, and predict the recommendation effect information corresponding to the product recommendation strategy. The product distribution platform can use this target recommendation effect prediction model to predict the recommendation effect information of the product recommendation strategy on users.

[0029] The target recommendation effect prediction model is a federated learning model, which can be obtained through federated learning between the product distribution platform and the financial institution platform. The participants in this federated learning are the product distribution platform and the financial institution platform. The federated learning method can adopt an appropriate federated learning framework. In one implementation, the process of obtaining the target recommendation effect model through federated learning between the product distribution platform and the financial institution platform can include the following steps.

[0030] Step 1: Receive the public key sent by the coordinator.

[0031] The coordinator can be an agreed central server that is connected to the product distribution platform and the financial institution platform. The public key is generated by the coordinator, who stores the private key corresponding to the public key. The private key can decrypt the data encrypted by the public key.

[0032] Step 2: Encrypt the sample user information of the product distribution platform based on the public key to obtain the first encrypted user information, align the encrypted samples of the first encrypted user information with the second encrypted user information of the financial institution platform, and determine the intersection users of the product distribution platform and the financial institution platform.

[0033] The second encrypted user information is obtained by encrypting the sample user information of the financial institution platform using the public key. The intersection users are the users shared by the sample users of the financial institution's agency sales side and the sample users of the financial institution platform.

[0034] Step 3: Based on the first user information of the intersection user and the recommendation effect information of the sample product recommendation strategy on the intersection user, the target recommendation effect prediction model to be trained is encrypted model training to obtain the gradient and loss of the target recommendation effect prediction model.

[0035] Among them, the first user information of the intersection user is the user information of the intersection user on the product distribution platform, which includes: the user portrait information of the user on the product distribution platform.

[0036] Step 4: Upload the gradient and loss of the target recommendation effect prediction model of the product distribution platform to the coordinator, so that the coordinator can aggregate the gradient and loss of the target recommendation effect prediction model of the product distribution platform and the gradient and loss of the financial institution platform model to update the model parameters.

[0037] The gradient and loss of the financial institution platform model are uploaded to the coordinator by the financial institution platform. These gradient and loss are obtained by the financial institution platform through encrypted training of the financial institution platform model based on the second user information of the intersection user. The second user information of the intersection user is the user information of the intersection user on the financial institution platform, which includes the user's financial information.

[0038] Step 5: Update the model parameters of the target recommendation effect prediction model based on the model parameters returned by the coordinator.

[0039] After obtaining the target recommendation effect prediction model through federated learning with the financial institution platform, the product distribution platform can use the target recommendation effect prediction model to predict the recommendation effect information of the product recommendation strategy on the user. The prediction process may include the following steps.

[0040] Step 1: Input the user's first user information and product recommendation strategy into the target recommendation effect prediction model to generate first encrypted intermediate result information corresponding to the product recommendation strategy.

[0041] Among them, the target recommendation effect prediction model can extract the first product feature information of the product from the product recommendation strategy, and generate first encrypted intermediate result information based on the first user information and the first product feature information.

[0042] Step 2: Send the user identification and the first encrypted intermediate result information of the user to the coordinator, so that the coordinator generates aggregate information based on the first encrypted intermediate result information and the second encrypted intermediate result information; wherein, the second encrypted intermediate result information is returned to the coordinator by the financial institution platform, and the second encrypted intermediate result information is generated by the financial institution platform based on the second user information of the user, and the second user information of the user is obtained by the financial institution platform based on the user identification sent by the coordinator.

[0043] Step 3: Based on the aggregated information returned by the coordinator, generate recommendation effect information corresponding to the product recommendation strategy.

[0044] S106: Determine a target product recommendation strategy from the multiple product recommendation strategies based on the recommendation effect information corresponding to the product recommendation strategies.

[0045] The recommendation effect information is used to indicate the recommendation effect corresponding to the product recommendation strategy, which is, for example, the recommendation effect score corresponding to the product recommendation strategy. The target product recommendation effect is, for example, the product recommendation strategy with the highest recommendation effect score among multiple product recommendation strategies corresponding to the target recommendation position. The higher the recommendation effect score, the better the recommendation effect.

[0046] S108: Generate product recommendation information for the target recommendation position according to the target product recommendation strategy.

[0047] Product recommendation information can include product descriptions and product links corresponding to the target product recommendation strategy. The product links can link to the product's purchase page. By displaying the product recommendation information to users via the target recommendation slot, the product can be recommended to them. Users can further click on the product link in the product recommendation information to jump to the corresponding purchase page and make further purchases.

[0048] It is understandable that for financial products such as funds, the predictive effectiveness of their product recommendation strategies on user recommendations depends largely on the richness of the user's financial information. Furthermore, to enrich user financial information, the financial institution platform may include: multiple financial institution platforms, which may include first-category financial institution platforms and second-category financial institution platforms. First-category financial institutions may be institutions that issue financial products, such as wealth management product institutions. The second user information of the first-category financial institution platforms may include: user financial information, such as user purchase and transaction information of wealth management products; second-category financial institutions may be authoritative financial institutions, such as banks. The second user information of the second-category financial institution platforms may include: user asset information, such as the user's assets, liabilities, credit rating, etc. Thus, by combining user financial information from different financial institutions, a more accurate federated prediction of the recommendation effectiveness of financial product recommendation strategies for users can be made, further improving the accuracy of the recommendation prediction and making the recommendation information more targeted to users.

[0049] Accordingly, when a federated prediction is made on the recommendation effect of a product recommendation strategy on a user based on the first user information of a user on the product distribution platform and the second user information of users on multiple financial institution platforms, the target recommendation effect prediction model based on the prediction is also obtained through federated learning by the product distribution platform and the multiple financial institution platforms. The target recommendation effect prediction model is used to perform a federated prediction on the recommendation effect of a product recommendation strategy on a user based on the first user information of a user on the product distribution platform and the second user information of users on multiple financial institution platforms without data leaving the domain. The federated learning method used for training and prediction of the target recommendation effect prediction model is similar to the federated learning method of the aforementioned embodiment.

[0050] In specific implementation, the training method of the target recommendation effect prediction model may include the following steps.

[0051] Step 1: Receive the public key sent by the coordinator.

[0052] Step 2: Encrypt the sample user information of the product distribution platform using the public key to obtain first encrypted user information. Align the first encrypted user information with the second encrypted user information of multiple financial institution platforms to determine the intersection of users between the product distribution platform and the multiple financial institution platforms. The second encrypted user information of each financial institution platform is obtained by encrypting the sample user information of each financial institution platform using the public key. Intersection users are users shared by the product distribution platform and each financial institution platform.

[0053] Step 3: Based on the sample first user information of the intersection user and the recommendation effect information of the sample product recommendation strategy on the intersection user, the target recommendation effect prediction model to be trained is encrypted model training to obtain the gradient and loss of the target recommendation effect prediction model.

[0054] Among them, the sample first user information of the intersection user is similar to the first user information, which can be: the user portrait information of the user on the product distribution platform.

[0055] Step 4: Upload the gradient and loss of the target recommendation effect prediction model of the product distribution platform to the coordinator, so that the coordinator aggregates the gradient and loss of the target recommendation effect prediction model of the product distribution platform and the gradient and loss of the financial institution platform model, and updates the model parameters. Among them, the gradient and loss of the financial institution platform model are uploaded to the coordinator by the financial institution platform, and the gradient and loss are obtained by the financial institution platform through encrypted training of the financial institution platform model based on the sample second user information of the intersection user. The sample second user information of the intersection user is similar to the second user information, and may include: the user's financial management information.

[0056] Step 5: Update the model parameters of the target recommendation effect prediction model based on the model parameters returned by the coordinator.

[0057] The prediction process of predicting the recommendation effect information of the product recommendation strategy on the user through the target recommendation effect prediction model may include the following steps.

[0058] Step 1: Input the user's first user information and product recommendation strategy into the target recommendation effect prediction model to generate first encrypted intermediate result information corresponding to the product recommendation strategy.

[0059] Step 2: Send the user identification and the first encrypted intermediate result information of the user to the coordinator, so that the coordinator generates aggregate information based on the first encrypted intermediate result information and the second encrypted intermediate result information of multiple financial institution platforms; wherein, the second encrypted intermediate result information of each financial institution platform is returned to the coordinator by each financial institution platform, and the second encrypted intermediate result information of each financial institution platform is generated by each financial institution platform based on the second user information of the user, and the second user information of the user is obtained by each financial institution platform based on the user identification sent by the coordinator.

[0060] Step 3: Based on the aggregated information returned by the coordinator, generate recommendation effect information corresponding to the product recommendation strategy.

[0061] It is understandable that the second product feature information of the product, such as the scale and sales of the product, also has a relatively important impact on the success of the recommendation. Therefore, when predicting the recommendation effect of the product recommendation strategy on the user, the second product feature information of the product corresponding to the product recommendation strategy can be considered. In some implementations, a federal prediction of the recommendation effect of the product recommendation strategy on the user can be made based on the first user information of the user of the product distribution platform, the product feature information of the product corresponding to the product recommendation strategy, and the second user information of the user of the financial institution platform. Among them, the second product feature information can be provided by the product distribution platform. In specific implementation, the second product feature information can be obtained from the product distribution platform.

[0062] Accordingly, the target recommendation effect prediction model underlying the federated prediction is also derived through federated learning between the product distribution platform and the financial institution platform based on the first user information, the product feature information, and the target recommendation effect prediction model. The federated learning method for this target recommendation effect prediction model is similar to the aforementioned federated learning method.

[0063] In specific implementation, the federated learning process of the target recommendation effect prediction model may include the following steps.

[0064] Step 1: Receive the public key sent by the coordinator.

[0065] Step 2: Encrypt the sample user information of the product distribution platform using the public key to obtain first encrypted user information. Align the first encrypted user information with the second encrypted user information of the financial institution platform to determine the intersection of users between the product distribution platform and the financial institution platform. The second encrypted user information is obtained by the financial institution platform by encrypting the sample user information of the financial institution platform using the public key. Intersection users are users shared by both the sample users of the financial institution distribution side and the sample users of the financial institution platform.

[0066] Step 3: Based on the first user information of the intersection user, the second product feature information of the product corresponding to the sample product recommendation strategy, and the recommendation effect information of the sample product recommendation strategy on the intersection user, the target recommendation effect prediction model to be trained is encrypted model training to obtain the gradient and loss of the target recommendation effect prediction model.

[0067] Among them, the first user information of the intersection user and the second product feature information of the product corresponding to the sample product recommendation strategy can be used as the input of the target recommendation effect prediction model to be trained, and the recommendation effect information of the sample product recommendation strategy for the intersection user can be used as the output label corresponding to the input to calculate the gradient and loss of the target recommendation effect prediction model.

[0068] Step 4: Upload the gradient and loss of the target recommendation effect prediction model of the product distribution platform to the coordinator, so that the coordinator can aggregate the gradient and loss of the target recommendation effect prediction model of the product distribution platform and the gradient and loss of the financial institution platform model to update the model parameters.

[0069] The gradient and loss of the financial institution platform model are uploaded to the coordinator by the financial institution platform. These gradient and loss are obtained by the financial institution platform through encrypted training of the financial institution platform model based on the second user information of the intersection user. The second user information of the intersection user is the user information of the intersection user on the financial institution platform, which includes the user's financial information.

[0070] Step 5: Update the model parameters of the target recommendation effect prediction model based on the model parameters returned by the coordinator.

[0071] The prediction process of predicting the recommendation effect information of the product recommendation strategy on the user through the target recommendation effect prediction model may include the following steps.

[0072] Step 1: Input the user's first user information, the second product feature information of the product corresponding to the product recommendation strategy, and the product recommendation strategy into the target recommendation effect prediction model to generate the first encrypted intermediate result information corresponding to the product recommendation strategy.

[0073] Among them, the target recommendation effect prediction model can extract the first product feature information of the product from the product recommendation strategy, splice the first product feature information with the second product feature information to obtain the product feature information of the product, and generate the first encrypted intermediate result information corresponding to the product recommendation strategy based on the user's first user information and the product feature information of the product.

[0074] Step 2: Send the user identification and the first encrypted intermediate result information of the user to the coordinator, so that the coordinator generates aggregate information based on the first encrypted intermediate result information and the second encrypted intermediate result information of multiple financial institution platforms; wherein, the second encrypted intermediate result information of each financial institution platform is returned to the coordinator by each financial institution platform, and the second encrypted intermediate result information of each financial institution platform is generated by each financial institution platform based on the second user information of the user, and the second user information of the user is obtained by each financial institution platform based on the user identification sent by the coordinator.

[0075] Step 3: Based on the aggregated information returned by the coordinator, generate recommendation effect information corresponding to the product recommendation strategy.

[0076] A product recommendation strategy is used to generate product recommendation information. In some cases, the product recommendation strategy may include information from multiple dimensions used to generate the product recommendation information, such as product feature information of the product to be recommended, recommendation copy information, etc. Furthermore, to more accurately evaluate the recommendation effectiveness of the product recommendation strategy, the recommendation effectiveness information may include scores from multiple dimensions of the product recommendation strategy, such as the product dimension score and the copy dimension score. Based on the scores from multiple dimensions of the product recommendation strategy, such as the product dimension score and the copy dimension score, a target product recommendation strategy may be selected from the product recommendation strategies.

[0077] During specific implementation, the product recommendation strategies can be sorted in a first order based on their product dimension scores, and the candidate product recommendation strategies can be determined based on their order in the first sorting. Then, the candidate product recommendation strategies can be sorted in a second order based on their copy dimension scores, and the target product recommendation strategy can be determined based on their order in the second sorting. Among them, the candidate product recommendation strategies are, for example, product recommendation strategies whose order in the first sorting is before the set order, and the first sorting is, for example, sorting in descending order based on the product dimension scores, where the higher the product dimension score, the better the recommendation effect. The target product recommendation strategy is, for example, the candidate product recommendation strategy with the highest order in the second sorting, and the second sorting is, for example, sorting in descending order based on the copy dimension scores, where the higher the copy dimension score, the better the recommendation effect.

[0078] Accordingly, based on the first user information of the user of the product distribution platform and the second user information of the user of the financial institution platform, when a federal prediction is made on the recommendation effect of the product recommendation strategy on the user, the predicted recommendation effect information includes the corresponding product dimension score and copy dimension score of the product recommendation strategy. This information can be predicted by the corresponding target recommendation effect prediction model. The federated learning method of the target recommendation effect prediction model is similar to the federated learning method of the target recommendation effect prediction model provided in the aforementioned embodiment. In specific implementation, the training method of the target recommendation effect prediction model may include the following steps.

[0079] Step 1: Receive the public key sent by the coordinator.

[0080] Step 2: Encrypt the sample user information of the product distribution platform based on the public key to obtain the first encrypted user information, align the encrypted samples of the first encrypted user information with the second encrypted user information of the financial institution platform, and determine the intersection users of the product distribution platform and the financial institution platform.

[0081] Step 3: Based on the first user information of the intersection user and the sample product recommendation strategy for the sample recommendation effect information of the intersection user, the target recommendation effect prediction model to be trained is encrypted model training to obtain the gradient and loss of the target recommendation effect prediction model.

[0082] Among them, the sample recommendation effect information includes: product dimension score and copy dimension score. The sample product recommendation strategy includes product feature information and recommended copy information of the product. The first user information of the intersection user and the sample product recommendation strategy can be input into the target recommendation effect prediction model to be trained. The target recommendation effect prediction model to be trained can extract the product feature information and recommended copy information of the product from the sample product recommendation strategy, and predict the product dimension score and copy dimension score of the intersection user based on the first user information of the intersection user and the product feature information and recommended copy information of the product. According to the predicted product dimension score and copy dimension score and the product dimension score and copy dimension score in the sample recommendation effect information, the gradient and loss of the target recommendation effect prediction model are calculated.

[0083] Step 4: Upload the gradient and loss of the target recommendation effect prediction model of the product distribution platform to the coordinator, so that the coordinator can aggregate the gradient and loss of the target recommendation effect prediction model of the product distribution platform and the gradient and loss of the financial institution platform model to update the model parameters.

[0084] Step 5: Update the model parameters of the target recommendation effect prediction model based on the model parameters returned by the coordinator.

[0085] The prediction process of predicting the recommendation effect information of the product recommendation strategy on the user through the target recommendation effect prediction model may include the following steps.

[0086] Step 1: Input the user's first user information and product recommendation strategy into the target recommendation effect prediction model to generate first encrypted intermediate result information corresponding to the product recommendation strategy.

[0087] Among them, the target recommendation effect prediction model can extract the product feature information and copy feature information of the product from the product recommendation strategy, and generate the first encrypted intermediate result information based on the user's first user information and the product feature information and copy feature information of the product.

[0088] Step 2: Send the user identification and the first encrypted intermediate result information of the user to the coordinator, so that the coordinator generates aggregate information based on the first encrypted intermediate result information and the second encrypted intermediate result information of multiple financial institution platforms; wherein, the second encrypted intermediate result information of each financial institution platform is returned to the coordinator by each financial institution platform, and the second encrypted intermediate result information of each financial institution platform is generated by each financial institution platform based on the second user information of the user, and the second user information of the user is obtained by each financial institution platform based on the user identification sent by the coordinator.

[0089] Step 3: Based on the aggregated information returned by the coordinator, generate recommendation effect information corresponding to the product recommendation strategy.

[0090] The products in the above embodiments may be appropriate financial products, such as various financial products, etc. The following uses fund products as an example of financial products to further illustrate the product recommendation method provided in the embodiments of the present application.

[0091] Figure 3 illustrates an example product recommendation method for the product recommendation method shown in Figure 2. In this example, the product is a fund product, the product distribution platform is a third-party payment platform, and there are multiple financial institutions, including fund institutions and banks. As shown in Figure 3, the method for recommending fund products may include the following process.

[0092] S202: According to the user's target operation on the target page, multiple fund product recommendation strategies corresponding to the target recommendation positions in the target page are obtained.

[0093] S204, based on the first user information of the user in the third-party payment platform, the second product feature information of the fund product corresponding to the fund product recommendation strategy, the second user information of the user in the fund institution platform, and the second user information of the user in the bank institution platform, a federal prediction is made on the recommendation effect of the fund product recommendation strategy on the user to obtain the recommendation effect information corresponding to the fund product recommendation strategy.

[0094] The fund product recommendation strategy, the first user information of the user on the third-party payment platform, and the second product feature information of the fund product corresponding to the fund product recommendation strategy can be input into the target recommendation effect prediction model to obtain first encrypted intermediate result information. Based on the first encrypted intermediate result information, the second encrypted intermediate result information of the fund institution platform, and the second encrypted intermediate result information of the bank institution platform, recommendation effect information corresponding to the fund product recommendation strategy is generated. The target recommendation effect prediction model is obtained by federated learning of the third-party payment platform, the fund institution platform, and the financial institution platform. There can be one or more fund institution platforms, and one or more financial institution platforms.

[0095] The fund product recommendation strategy includes the first product feature information of the fund product, such as the fund product's rate of return, and the copy feature information of the recommendation copy, such as copy style, keywords, etc. The second product feature information of the fund product includes: the sales information of the fund product. The first user information includes: the user portrait information of the user on the third-party payment platform. The second user information of the fund institution platform includes the user's financial management information, such as the user's holdings and transaction information. The second user information of the banking institution platform includes the user's asset information, such as asset holdings and liabilities. Based on the above information, the recommendation effect information of the fund product recommendation strategy for the user is predicted, which can obtain a more accurate prediction effect based on the rich feature information.

[0096] S206 , determining a target fund product recommendation strategy from a plurality of fund product recommendation strategies according to the recommendation effect information corresponding to the fund product recommendation strategy.

[0097] Recommendation effectiveness information includes product and content scores. These scores can be combined to comprehensively evaluate the effectiveness of the fund product recommendation strategy. The target fund product recommendation strategy can then be determined based on the comprehensive evaluation results.

[0098] S208: Generate fund product recommendation information for the target recommendation position according to the target fund product recommendation strategy.

[0099] S210: Display the fund product recommendation information in the target recommendation position.

[0100] It should be noted that in this example, the product recommendation strategies corresponding to the target recommendation positions are all fund product recommendation strategies. In this example, a fund product recommendation strategy suitable for the user is selected from the fund product recommendation strategies. However, the present application is not limited to this. In other embodiments, the product recommendation strategies corresponding to the target recommendation positions may also include recommendation strategies for different products, such as fund product recommendation strategies and recommendation strategies for other financial products. A product recommendation strategy suitable for the user can be selected from the recommendation strategies for different products.

[0101] Corresponding to the method provided in the above embodiment, based on the same technical concept, the embodiment of the present application also provides a product recommendation device for executing the product recommendation method provided in the embodiment of the present application, and the device can be used in a product distribution platform. Figure 4 is a module schematic diagram of the product recommendation device provided in the embodiment of the present application. As shown in Figure 4, the financial product recommendation device includes: a strategy acquisition module 10, which is used to obtain multiple product recommendation strategies corresponding to the target recommendation position in the target page according to the user's target operation on the target page; an effect prediction module 20, which is used to perform a federal prediction of the recommendation effect of the product recommendation strategy on the user based on the first user information of the user of the product distribution platform and the second user information of the user of the financial institution platform, and obtain the recommendation effect information corresponding to the product recommendation strategy; a target strategy determination module 30, which is used to determine the target product recommendation strategy among the multiple product recommendation strategies according to the recommendation effect information corresponding to the product recommendation strategy; a recommendation information generation module 40, which is used to generate product recommendation information for the target recommendation position according to the target product recommendation strategy.

[0102] It should be noted that the product recommendation device provided in the embodiment of the present application and the product recommendation method provided in the embodiment of the present application are based on the same inventive concept. Therefore, the specific implementation of the system embodiment can refer to the implementation of the corresponding method mentioned above, and the repeated parts will not be repeated.

[0103] Corresponding to the product recommendation method provided in the above embodiment, based on the same technical concept, the embodiment of the present application also provides a computer device, which is used to execute the product recommendation method provided in the embodiment of the present application, as shown in Figure 5.

[0104] Computer devices can vary significantly depending on their configuration or performance. They may include one or more processors and memory, which may store one or more applications or data. The memory may be either ephemeral or persistent. Applications stored in the memory may include one or more modules, each of which may include a series of computer-executable instructions for the computer device. Furthermore, the processor may be configured to communicate with the memory to execute the series of computer-executable instructions in the memory on the computer device. Computer devices may also include one or more power supplies, one or more wired or wireless network interfaces, one or more input / output interfaces, one or more keyboards, and the like.

[0105] In a specific embodiment, a computer device includes a memory and one or more programs, wherein the one or more programs are stored in the memory, and the one or more programs may include one or more modules, and each module may include a series of computer-executable instructions in the computer device, and is configured to be executed by one or more processors. The one or more programs include instructions for performing the following computer-executable instructions: according to the user's target operation on the target page, obtaining multiple product recommendation strategies corresponding to the target recommendation position in the target page; based on the first user information of the user on the product distribution platform and the second user information of the user on the financial institution platform, performing a federal prediction on the recommendation effect of the product recommendation strategy on the user to obtain recommendation effect information corresponding to the product recommendation strategy; according to the recommendation effect information corresponding to the product recommendation strategy, determining a target product recommendation strategy among the multiple product recommendation strategies; according to the target product recommendation strategy, generating product recommendation information for the target recommendation position.

[0106] In one implementation, the recommendation effect information includes: the product dimension score and the copy dimension score of the product recommendation strategy; determining the target product recommendation strategy among the multiple product recommendation strategies based on the recommendation effect information corresponding to the product recommendation strategy includes: performing a first sorting of the product recommendation strategies according to the product dimension scores of the product recommendation strategies, and determining the candidate product recommendation strategies according to the order of the product recommendation strategies in the first sorting; performing a second sorting of the candidate product recommendation strategies according to the copy dimension scores of the candidate product recommendation strategies, and determining the target product recommendation strategy according to the order of the candidate product recommendation strategies in the second sorting.

[0107] In one implementation, the first user information of the product distribution platform includes: user portrait information; the financial institution platform includes: multiple financial institution platforms, the multiple financial institution platforms include a first type of financial institution platform and a second type of financial institution platform, the second user information of the first type of financial institution platform includes: the user's financial management information; the second user information of the second type of financial institution platform includes: the user's asset status information.

[0108] In one implementation, the first user information of the user on the product distribution platform and the second user information of the user on the financial institution platform are used to make a federal prediction of the recommendation effect of the product recommendation strategy on the user, including: based on the first user information of the user on the product distribution platform, the product feature information of the product corresponding to the product recommendation strategy of the product distribution platform, and the second user information of the users on the multiple financial institution platforms, a federal prediction of the recommendation effect of the product recommendation strategy on the user.

[0109] In one implementation, the product feature information of the product corresponding to the product recommendation strategy includes: first product feature information and second product feature information; the first product feature information is extracted from the product recommendation strategy, and the second product feature information is provided by the product distribution platform.

[0110] In one implementation, the recommendation effect of the product recommendation strategy on the user is federally predicted based on the first user information of the user on the product distribution platform and the second user information of the user on the financial institution platform, including: inputting the first user information of the user and the product recommendation strategy into a target recommendation effect prediction model to generate first encrypted intermediate result information corresponding to the product recommendation strategy; wherein, the target recommendation effect prediction model is obtained by federal learning between the product distribution platform and the financial institution platform; sending the user identifier of the user and the first encrypted intermediate result information to the coordinator, so that the coordinator generates aggregate information based on the first encrypted intermediate result information and the second encrypted intermediate result information; wherein, the second encrypted intermediate result information is returned to the coordinator by the financial institution platform, and the second encrypted intermediate result information is generated by the financial institution platform based on the second user information of the user, and the second user information of the user is obtained by the financial institution platform based on the user identifier sent by the coordinator; based on the aggregate information returned by the coordinator, the recommendation effect information corresponding to the product recommendation strategy is generated.

[0111] In one implementation, the federated learning method of the target recommendation effect prediction model includes: receiving a public key sent by a coordinator; encrypting the sample user information of the product distribution platform based on the public key to obtain first encrypted user information, performing encrypted sample alignment on the first encrypted user information and the second encrypted user information of the financial institution platform to obtain the intersection user of the product distribution platform and the financial institution platform; performing encrypted model training on the target recommendation effect prediction model to be trained based on the first user information of the intersection user on the product distribution platform and the recommendation effect information of the sample recommendation strategy on the intersection user to obtain the gradient and loss of the target recommendation effect prediction model; uploading the gradient and loss of the target recommendation effect prediction model to the coordinator, so that the coordinator aggregates the gradient and loss of the target recommendation effect prediction model of the product distribution platform and the gradient and loss of the financial institution platform model to update the model parameters; the gradient and loss of the financial institution platform model are uploaded to the coordinator by the financial institution platform, and the gradient and loss are obtained by the financial institution platform through encrypted training of the financial institution platform model based on the second user information of the intersection user; and updating the model parameters of the target recommendation effect prediction model according to the model parameters returned by the coordinator.

[0112] It should be noted that the computer device provided in the embodiment of the present application is used to execute the product recommendation method provided in the embodiment of the present application. It is based on the same inventive concept as the product recommendation method provided in the aforementioned embodiment. Therefore, the specific implementation of the computer device embodiment can refer to the implementation of the aforementioned corresponding method, and the repetitive parts will not be repeated.

[0113] Corresponding to the product recommendation method provided in the embodiment of the present application, based on the same technical concept, the embodiment of the present application also provides a storage medium for storing computer-executable instructions.

[0114] In a specific embodiment, the storage medium can be a U disk, an optical disk, a hard disk, etc., and the computer executable instructions stored in the storage medium can implement the following process when executed by the processor: according to the user's target operation on the target page, obtain multiple product recommendation strategies corresponding to the target recommendation position in the target page; based on the first user information of the user on the product distribution platform and the second user information of the user on the financial institution platform, perform a federal prediction of the recommendation effect of the product recommendation strategy on the user to obtain the recommendation effect information corresponding to the product recommendation strategy; according to the recommendation effect information corresponding to the product recommendation strategy, determine the target product recommendation strategy among the multiple product recommendation strategies; according to the target product recommendation strategy, generate product recommendation information for the target recommendation position.

[0115] In one implementation, the recommendation effect information includes: the product dimension score and the copy dimension score of the product recommendation strategy; determining the target product recommendation strategy among the multiple product recommendation strategies based on the recommendation effect information corresponding to the product recommendation strategy includes: performing a first sorting of the product recommendation strategies according to the product dimension scores of the product recommendation strategies, and determining the candidate product recommendation strategies according to the order of the product recommendation strategies in the first sorting; performing a second sorting of the candidate product recommendation strategies according to the copy dimension scores of the candidate product recommendation strategies, and determining the target product recommendation strategy according to the order of the candidate product recommendation strategies in the second sorting.

[0116] In one implementation, the first user information of the product distribution platform includes: user portrait information; the financial institution platform includes: multiple financial institution platforms, the multiple financial institution platforms include a first type of financial institution platform and a second type of financial institution platform, the second user information of the first type of financial institution platform includes: the user's financial management information; the second user information of the second type of financial institution platform includes: the user's asset status information.

[0117] In one implementation, the first user information of the user on the product distribution platform and the second user information of the user on the financial institution platform are used to make a federal prediction of the recommendation effect of the product recommendation strategy on the user, including: based on the first user information of the user on the product distribution platform, the product feature information of the product corresponding to the product recommendation strategy of the product distribution platform, and the second user information of the users on the multiple financial institution platforms, a federal prediction of the recommendation effect of the product recommendation strategy on the user.

[0118] In one implementation, the product feature information of the product corresponding to the product recommendation strategy includes: first product feature information and second product feature information; the first product feature information is extracted from the product recommendation strategy, and the second product feature information is provided by the product distribution platform.

[0119] In one implementation, the recommendation effect of the product recommendation strategy on the user is federally predicted based on the first user information of the user on the product distribution platform and the second user information of the user on the financial institution platform, including: inputting the first user information of the user and the product recommendation strategy into a target recommendation effect prediction model to generate first encrypted intermediate result information corresponding to the product recommendation strategy; wherein, the target recommendation effect prediction model is obtained by federal learning between the product distribution platform and the financial institution platform; sending the user identifier of the user and the first encrypted intermediate result information to the coordinator, so that the coordinator generates aggregate information based on the first encrypted intermediate result information and the second encrypted intermediate result information; wherein, the second encrypted intermediate result information is returned to the coordinator by the financial institution platform, and the second encrypted intermediate result information is generated by the financial institution platform based on the second user information of the user, and the second user information of the user is obtained by the financial institution platform based on the user identifier sent by the coordinator; based on the aggregate information returned by the coordinator, the recommendation effect information corresponding to the product recommendation strategy is generated.

[0120] In one implementation, the federated learning method of the target recommendation effect prediction model includes: receiving a public key sent by a coordinator; encrypting the sample user information of the product distribution platform based on the public key to obtain first encrypted user information, performing encrypted sample alignment on the first encrypted user information and the second encrypted user information of the financial institution platform to obtain the intersection user of the product distribution platform and the financial institution platform; performing encrypted model training on the target recommendation effect prediction model to be trained based on the first user information of the intersection user on the product distribution platform and the recommendation effect information of the sample recommendation strategy on the intersection user to obtain the gradient and loss of the target recommendation effect prediction model; uploading the gradient and loss of the target recommendation effect prediction model to the coordinator, so that the coordinator aggregates the gradient and loss of the target recommendation effect prediction model of the product distribution platform and the gradient and loss of the financial institution platform model to update the model parameters; the gradient and loss of the financial institution platform model are uploaded to the coordinator by the financial institution platform, and the gradient and loss are obtained by the financial institution platform through encrypted training of the financial institution platform model based on the second user information of the intersection user; and updating the model parameters of the target recommendation effect prediction model according to the model parameters returned by the coordinator.

[0121] It should be noted that the computer-executable instructions stored in the storage medium provided in the embodiment of the present application are used to execute the product recommendation method provided in the embodiment of the present application. It is based on the same inventive concept as the product recommendation method provided in the aforementioned embodiment. Therefore, the specific implementation of the storage medium embodiment can refer to the implementation of the aforementioned corresponding method, and the repetitive parts will not be repeated.

[0122] The foregoing description describes specific embodiments of the present application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different from that described in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order shown or the sequential order to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0123] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the embodiments of the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the application can adopt the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0124] The present application is described with reference to the flow chart and / or block diagram of the method, device (system), and computer program product according to the embodiment of the present application. It should be understood that each flow process and / or box in the flow chart and / or block diagram and the combination of the flow process and / or box in the flow chart and / or block diagram can be realized by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processing machine or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a device for realizing the function specified in one flow chart flow or multiple flows and / or one box or multiple boxes of the block diagram.

[0125] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a product including an instruction device that implements the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0126] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0127] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0128] Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0129] Computer-readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory media such as modulated data signals and carrier waves.

[0130] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0131] Embodiments of the present application may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, and the like that perform specific tasks or implement specific abstract data types. One or more embodiments of the present application may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communications network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0132] The various embodiments in this application are described in a progressive manner. Similar parts between the various embodiments can be referred to in conjunction with each other. Each embodiment focuses on the differences between the other embodiments. In particular, the system embodiment is generally similar to the method embodiment, so the description is relatively simple. For relevant parts, refer to the partial description of the method embodiment.

[0133] The foregoing description is merely an example of the present invention and is not intended to limit the present invention. Persons skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be included within the scope of the claims herein.

Claims

1. A product recommendation method, comprising: According to the user's target operation on the target page, multiple product recommendation strategies corresponding to the target recommendation position in the target page are obtained; Based on the first user information of the user on the product distribution platform and the second user information of the user on the financial institution platform, a federated prediction is performed on the recommendation effect of the product recommendation strategy on the user to obtain recommendation effect information corresponding to the product recommendation strategy; Determining a target product recommendation strategy from the multiple product recommendation strategies based on the recommendation effect information corresponding to the product recommendation strategies; According to the target product recommendation strategy, product recommendation information for the target recommendation position is generated.

2. The product recommendation method according to claim 1, wherein the recommendation effect information includes: The product dimension score and the copy dimension score of the product recommendation strategy; determining the target product recommendation strategy from the multiple product recommendation strategies based on the recommendation effect information corresponding to the product recommendation strategy, including: performing a first ranking of the product recommendation strategies according to the product dimension scores of the product recommendation strategies, and determining candidate product recommendation strategies according to the order of the product recommendation strategies in the first ranking; The candidate product recommendation strategies are ranked in a second order according to the copy dimension scores of the candidate product recommendation strategies, and the target product recommendation strategy is determined according to the order of the candidate product recommendation strategies in the second order.

3. The product recommendation method according to claim 1, wherein the first user information of the product distribution platform includes: User profile information; The financial institution platform includes: multiple financial institution platforms, the multiple financial institution platforms include a first type of financial institution platform and a second type of financial institution platform, the second user information of the first type of financial institution platform includes: the user's financial management information; the second user information of the second type of financial institution platform includes: the user's asset information.

4. The product recommendation method according to claim 3, wherein the method comprises: performing a federated prediction of the recommendation effect of the product recommendation strategy on the user based on the first user information of the user on the product distribution platform and the second user information of the user on the financial institution platform, comprising: Based on the first user information of the user of the product distribution platform, the product feature information of the product corresponding to the product recommendation strategy, and the second user information of the users of the multiple financial institution platforms, a federal prediction is made on the recommendation effect of the product recommendation strategy on the user.

5. The product recommendation method according to claim 1, wherein the product feature information of the product corresponding to the product recommendation strategy includes: First product feature information and second product feature information; The first product feature information is extracted from the product recommendation strategy, and the second product feature information is provided by the product distribution platform.

6. The product recommendation method according to claim 1, wherein the method of performing a federated prediction of the recommendation effect of the product recommendation strategy on the user based on the first user information of the user on the product distribution platform and the second user information of the user on the financial institution platform comprises: Inputting the first user information of the user and the product recommendation strategy into a target recommendation effect prediction model to generate first encrypted intermediate result information corresponding to the product recommendation strategy; wherein the target recommendation effect prediction model is obtained by federated learning between the product distribution platform and the financial institution platform; sending the user identifier of the user and the first encrypted intermediate result information to the coordinator, so that the coordinator generates aggregate information based on the first encrypted intermediate result information and the second encrypted intermediate result information; wherein the second encrypted intermediate result information is returned to the coordinator by the financial institution platform, and the second encrypted intermediate result information is generated by the financial institution platform based on the second user information of the user, and the second user information of the user is obtained by the financial institution platform based on the user identifier sent by the coordinator; Based on the aggregated information returned by the coordinator, recommendation effect information corresponding to the product recommendation strategy is generated.

7. The product recommendation method according to claim 6, wherein the federated learning method of the target recommendation effect prediction model comprises: Receive the public key sent by the coordinator; Encrypting sample user information of the product distribution platform based on the public key to obtain first encrypted user information, performing encrypted sample alignment on the first encrypted user information and the second encrypted user information of the financial institution platform to obtain the intersection user of the product distribution platform and the financial institution platform; Based on the first user information of the intersection user on the product distribution platform and the recommendation effect information of the sample recommendation strategy on the intersection user, performing encrypted model training on the target recommendation effect prediction model to be trained to obtain the gradient and loss of the target recommendation effect prediction model; The gradient and loss of the target recommendation effect prediction model are uploaded to the coordinator, so that the coordinator aggregates the gradient and loss of the target recommendation effect prediction model of the product distribution platform and the gradient and loss of the financial institution platform model to update the model parameters; the gradient and loss of the financial institution platform model are uploaded to the coordinator by the financial institution platform, and the gradient and loss are obtained by the financial institution platform through encrypted training of the financial institution platform model based on the second user information of the intersection user; The model parameters of the target recommendation effect prediction model are updated according to the model parameters returned by the coordinator.

8. A product recommendation device, applied to a product distribution platform, comprising: A strategy acquisition module is used to acquire multiple product recommendation strategies corresponding to target recommendation positions on the target page according to the user's target operation on the target page; an effect prediction module, configured to perform a federated prediction of the recommendation effect of the product recommendation strategy on the user based on the first user information of the user on the product distribution platform and the second user information of the user on the financial institution platform, and obtain recommendation effect information corresponding to the product recommendation strategy; a target strategy determination module, configured to determine a target product recommendation strategy from among the plurality of product recommendation strategies based on recommendation effect information corresponding to the product recommendation strategies; The recommendation information generation module is used to generate product recommendation information for the target recommendation position according to the target product recommendation strategy.

9. A computer device, characterized in that: include: processor; as well as a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that Computer instructions are stored thereon, which are configured to, when executed by a processor of a computer device, enable the computer device to perform the method according to any one of claims 1 to 7.

Images