A product recommendation method and system for simple hybrid privacy protection scenarios

Abstract

The present invention discloses a product recommendation method and system for a simple hybrid privacy protection scenario, and relates to the field of information processing technology. The product recommendation method comprises: after receiving a recommendation request from a target user, obtaining a user ID corresponding to the target user and sending the user ID to an auxiliary party; receiving a user representation fed back by the auxiliary party, and obtaining candidate products based on the user representation and all product representations calculated by a preset model, wherein the user representation is calculated by the auxiliary party based on user attributes obtained using the user ID and the preset model; determining a target product from the candidate products according to preset recommendation rules and sending it to the target user. The present invention can achieve accurate recommendations while meeting the privacy requirements of simple hybrid scenarios.

Description

Technical Field

[0001] The present invention relates to the field of information processing technology, and in particular to a product recommendation method and system for simple hybrid privacy protection scenarios. Background Art

[0002] Simple hybrid privacy-preserving recommendation scenarios typically involve two co-participants: one party stores user-item interaction information and item attributes, while the other party stores the corresponding user attributes. In privacy-preserving scenarios, to enhance privacy, the item and user attributes held by each co-participant are kept private.

[0003] However, existing models usually involve participants sharing product information and only protecting their respective user information and user-product interaction information. Such models cannot meet the privacy requirements of simple hybrid scenarios.

[0004] Therefore, how to achieve accurate recommendations while meeting the privacy requirements of simple hybrid scenarios is a technical problem that technical personnel in this field urgently need to solve. Summary of the Invention

[0005] The present invention provides a product recommendation method and system for simple hybrid privacy protection scenarios, which can achieve accurate recommendations while meeting the privacy requirements of simple hybrid scenarios.

[0006] The present invention provides the following solutions:

[0007] In a first aspect, a product recommendation method for a simple hybrid privacy protection scenario is provided. The product recommendation method is applied to a product recommender and includes:

[0008] After receiving the recommendation request from the target user, obtaining the user ID corresponding to the target user and sending the user ID to the auxiliary party;

[0009] receiving a user representation fed back by the auxiliary party, and obtaining candidate products based on the user representation and all product representations calculated by a preset model, wherein the user representation is calculated by the auxiliary party based on user attributes obtained using the user ID and the preset model;

[0010] A target product is determined from the candidate products according to a preset recommendation rule and sent to the target user.

[0011] Optionally, obtaining candidate products according to the user representation and all product representations calculated by a preset model includes:

[0012] The distance between the user representation and all the product representations is calculated according to a preset algorithm, and the candidate products are obtained according to the distance.

[0013] Optionally, determining the target product from the candidate products according to a preset recommendation rule and sending the target product to the target user includes:

[0014] All the candidate commodities are sorted according to the distance, and the candidate commodities at a preset position are selected as the target commodities and sent to the target user.

[0015] Optionally, the process of constructing the preset model includes:

[0016] Generate training sample sets based on the interaction information between users and products;

[0017] Processing the training sample set to obtain a first sample set and a second sample set, and sending the second sample set to the auxiliary party;

[0018] Extracting product attributes corresponding to the products in the first sample set, and calculating corresponding product representations based on the product attributes;

[0019] receiving a user representation fed back by the auxiliary party, the user representation being obtained by the auxiliary party extracting user attributes corresponding to the user ID in the second sample set and calculating based on the user attributes;

[0020] Calculate an auxiliary party representation gradient based on the product representation and the user representation and send the auxiliary party representation gradient to the auxiliary party, the auxiliary party calculates an auxiliary party neural network representation gradient based on the auxiliary party representation gradient and updates the first part of the preset model based on the auxiliary party neural network representation gradient;

[0021] A recommender neural network representation gradient is calculated based on the product representation and the user representation, and a second part of the preset model is updated based on the recommender neural network representation gradient.

[0022] Optionally, the training sample set includes a plurality of samples consisting of users, commodities clicked or purchased by the users, and a first preset number of commodities not clicked or purchased by the users, and processing the training sample set to obtain the first sample set and the second sample set includes:

[0023] Selecting a second preset number of samples from the training sample set to generate the first sample set;

[0024] The second sample set is generated according to the user in the first sample set.

[0025] Optionally, the calculating the auxiliary party representation gradient according to the product representation and the user representation and sending the auxiliary party representation gradient to the auxiliary party includes:

[0026] Calculating the product representation and the user representation according to a preset calculation formula to obtain an initial auxiliary party representation gradient;

[0027] Noise obfuscation is performed on the initial auxiliary representation gradient to obtain the auxiliary representation gradient.

[0028] Optionally, performing noise obfuscation on the initial auxiliary side representation gradient to obtain the auxiliary side representation gradient includes:

[0029] Determine target noise according to the initial auxiliary representation gradient, and superimpose the target noise on the initial auxiliary representation gradient to perform noise confusion to obtain the auxiliary representation gradient; or

[0030] A noise having the same intensity as the initial auxiliary side representation gradient is selected from a plurality of pre-stored noises, and the noise is superimposed on the initial auxiliary side representation gradient to obtain the auxiliary side representation gradient.

[0031] Secondly, a product recommendation system for a simple hybrid privacy protection scenario is also provided, including:

[0032] An acquisition module, configured to, upon receiving a recommendation request from a target user, acquire a user ID corresponding to the target user and send the user ID to an auxiliary party;

[0033] a calculation module, connected to the acquisition module, configured to receive the user representation fed back by the auxiliary party and acquire candidate products based on the user representation and all product representations calculated by a preset model, wherein the user representation is calculated by the auxiliary party based on user attributes obtained using the user ID and the preset model;

[0034] The determination module is connected to the calculation module and is used to determine the target product from the candidate products according to a preset recommendation rule and send the target product to the target user.

[0035] In a third aspect, a computer device is also provided, comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the computer program is executed by the processor, the product recommendation method for a simple hybrid privacy protection scenario is implemented.

[0036] In a fourth aspect, a computer-readable storage medium is also provided, in which a computer program is stored. When the computer program is executed, the product recommendation method for a simple hybrid privacy protection scenario is implemented.

[0037] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0038] The product recommendation method for a simple hybrid privacy protection scenario provided by the present invention involves two participants, namely a product recommender and an auxiliary party. The product recommender stores the interaction information between users and products, as well as product attributes, and the auxiliary party stores user attributes. In order to improve data privacy, the product recommender does not send product attributes directly to the auxiliary party, and the auxiliary party does not send user attributes directly to the product recommender. In order to realize model construction, complete data is required. Therefore, the product auxiliary party converts user attributes into user representations and sends them to the product recommender. The product recommender converts product attributes into product representations and further obtains the auxiliary party representation gradient based on the product representation before sending it to the auxiliary party. In this way, the product recommender obtains the user representation data required to update the model, and the auxiliary party obtains the product representation-related data required to update the model. That is, the product recommender and the auxiliary party both obtain the data required to change their respective models, and protect their respective data, thereby meeting the privacy requirements of simple hybrid scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0040] Figure 1 This is a flowchart of a product recommendation method for a simple hybrid privacy protection scenario provided by an embodiment of the present invention;

[0041] Figure 2 This is a flowchart of a product recommendation method for a simple hybrid privacy protection scenario provided by another embodiment of the present invention;

[0042] Figure 3 This is a structural block diagram of a product recommendation system for a simple hybrid privacy protection scenario provided by an embodiment of the present invention;

[0043] Figure 4 This is an architectural diagram of a computer device provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0044] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field are within the scope of protection of the present invention.

[0045] The specific implementation scheme provided by the embodiment of the present invention is introduced in detail below.

[0046] The present invention provides a product recommendation method for a simple hybrid privacy protection scenario, involving a product recommender and an auxiliary party, which jointly implement recommendations. The simple hybrid method is to implement product recommendations after the information of the product recommender and the auxiliary party is simply mixed. Before product recommendations, model training is required. After model training, the product recommender stores product representations corresponding to all product attributes, and the auxiliary party stores user representations corresponding to all user attributes. The users involved in the two are consistent. The information interaction between the product recommender and the auxiliary party does not involve user attributes and product attributes, but only involves product representations and user representations, thereby ensuring that product recommendations are completed while meeting the privacy requirements of simple hybrid scenarios.

[0047] Example 1

[0048] This embodiment provides a product recommendation method for a simple hybrid privacy protection scenario, which is applied to the product recommendation side. Figure 1 This is a flowchart of a product recommendation method for a simple hybrid privacy protection scenario provided by an embodiment of the present invention. Figure 1 As shown, the product recommendation method generally includes the following steps:

[0049] S10: After receiving the recommendation request of the target user, obtain the user ID corresponding to the target user and send the user ID to the auxiliary party;

[0050] The product recommender pre-stores users and their corresponding user IDs. The target user is one of the pre-stored users in the product recommender. Upon receiving a recommendation request from the target user, the product recommender obtains the target user's corresponding user ID based on the pre-stored users and their corresponding user IDs and sends the user ID to the assisting party.

[0051] S20: Receive the user representation fed back by the auxiliary party, and obtain candidate products based on the user representation and all product representations calculated by a preset model, wherein the user representation is calculated by the auxiliary party based on user attributes obtained using the user ID and the preset model;

[0052] The preset model is a pre-trained model shared by both the product recommender and the auxiliary party. The product recommender obtains product representations based on product attributes and the preset model, while the auxiliary party obtains user representations based on user attributes and the preset model.

[0053] S30: Determine a target product from the candidate products according to a preset recommendation rule and send the target product to the target user.

[0054] The product recommendation method for a simple hybrid privacy protection scenario provided in this embodiment involves two participants, namely a product recommender and an auxiliary party. The recommender stores the interaction information between users and products, as well as product attributes, and the auxiliary party stores user attributes. In order to improve data privacy, the product recommender does not send product attributes directly to the auxiliary party, and the auxiliary party does not send user attributes directly to the product recommender. In order to realize model construction, complete data is required. Therefore, the auxiliary party converts user attributes into user representations and sends them to the product recommender. The product recommender converts product attributes into product representations and further obtains the auxiliary party representation gradient based on the product representation before sending it to the auxiliary party. In this way, the product recommender obtains the user representation data required to update the model, and the auxiliary party obtains the product representation-related data required to update the model. That is, the product recommender and the auxiliary party both obtain the data required to modify their respective models, and protect their respective data, thereby meeting the privacy requirements of simple hybrid scenarios.

[0055] Figure 2 This is a flowchart of a product recommendation method for a simple hybrid privacy protection scenario provided by another embodiment of the present invention. Figure 2 As shown, obtaining candidate products based on the user representation and all product representations calculated by a preset model includes:

[0056] S21: Calculate the distance between the user representation and all product representations according to a preset algorithm, and obtain the candidate products according to the distance.

[0057] More specifically, the user representation and the representations of all products are normalized data, and the distance is a cosine distance.

[0058] Furthermore, step S30 includes:

[0059] All the candidate commodities are sorted according to the distance, and the candidate commodities at a preset position are selected as the target commodities and sent to the target user.

[0060] The selection of the candidate products in the preset position may be a selection of a specific number of candidate products ranked first, and the specific number can be set as needed, for example, 100. The product recommender obtains the target product and presents it to the target user.

[0061] Preferably, the preset model includes a dual-tower model (DSSM). The product recommender and the auxiliary party perform a secure intersection of user IDs based on PSI technology. The preset model used in the above content can be constructed according to the following process, and the construction work is mainly completed by the product recommender:

[0062] Generate training sample sets based on the interaction information between users and products;

[0063] Processing the training sample set to obtain a first sample set and a second sample set, and sending the second sample set to the auxiliary party;

[0064] Extracting product attributes corresponding to the products in the first sample set, and calculating corresponding product representations based on the product attributes;

[0065] receiving a user representation fed back by the auxiliary party, the user representation being obtained by the auxiliary party extracting user attributes corresponding to the user ID in the second sample set and calculating based on the user attributes;

[0066] Calculate an auxiliary party representation gradient based on the product representation and the user representation and send the auxiliary party representation gradient to the auxiliary party, the auxiliary party calculates an auxiliary party neural network representation gradient based on the auxiliary party representation gradient and updates the first part of the preset model based on the auxiliary party neural network representation gradient;

[0067] A recommender neural network representation gradient is calculated based on the product representation and the user representation, and a second part of the preset model is updated based on the recommender neural network representation gradient.

[0068] Of course, it is understandable that before building the preset model, the product recommender and the auxiliary party must first align the users and obtain a list of user IDs of the common users.

[0069] Furthermore, after updating the preset model, the product recommendation party calculates the convergence index and sends the convergence index to the auxiliary party. Both parties judge whether the preset model has converged based on the convergence index. If not, the above process needs to be repeated until the preset model converges and the model construction is completed.

[0070] Specifically, the training sample set includes a plurality of samples consisting of users, products clicked or purchased by the users, and a first preset number of products that the users did not click or purchase. The first preset number can be selected according to the actual needs of model construction, and preferably, the first preset number is 4.

[0071] In the above process, processing the training sample set to obtain the first sample set and the second sample set includes:

[0072] Selecting a second preset number of samples from the training sample set to generate the first sample set;

[0073] The second sample set is generated according to the user in the first sample set.

[0074] In a specific construction scenario, for each user and the items clicked by the user, the Cartesian product is taken. Then, four items that the user did not click or purchase are randomly selected, along with the items that the user clicked and the user, to form a sample. These samples together form a training sample set. A second preset number of samples are selected from the training sample set to generate a first sample set. Compared to the training sample set, the first sample set is a small batch sample set. The data related to the items in the first sample set is removed, that is, the items clicked or purchased by the user and the four items that the user did not click or purchase are removed to generate a second sample set. This second sample set can also be called a one-sided small batch sample set.

[0075] More specifically, calculating the auxiliary party representation gradient according to the product representation and the user representation and sending the auxiliary party representation gradient to the auxiliary party includes:

[0076] Calculating the product representation and the user representation according to a preset calculation formula to obtain an initial auxiliary party representation gradient;

[0077] Noise obfuscation is performed on the initial auxiliary representation gradient to obtain the auxiliary representation gradient.

[0078] Among them, the preset calculation formula is:

[0079]

[0080] Q j is the user in the sample, j is a positive integer, D + The product clicked or purchased by the user, a i is the product representation corresponding to all products in the sample, a0 is product D + The corresponding product representation, b is the user representation corresponding to the user, and γ is the adjustment parameter.

[0081] Furthermore, performing noise obfuscation on the initial auxiliary side representation gradient to obtain the auxiliary side representation gradient includes:

[0082] The target noise is determined based on the initial auxiliary representation gradient, and the target noise is superimposed on the initial auxiliary representation gradient to perform noise obfuscation to obtain the auxiliary representation gradient. Specifically, the product recommender calculates noise of the same shape as the initial auxiliary representation gradient, namely the target noise, and superimposes it on the initial auxiliary representation gradient. The noise intensity of this noise can be adjusted to balance gradient security and computational accuracy. This approach can improve the accuracy of the constructed model.

[0083] In parallel with the above manner, performing noise obfuscation on the initial auxiliary side representation gradient to obtain the auxiliary side representation gradient may also include:

[0084] A noise with the same intensities as the initial auxiliary representation gradient is selected from a plurality of pre-stored noises and superimposed onto the initial auxiliary representation gradient to obtain the auxiliary representation gradient. The product recommender pre-stores multiple noise intensities. During the initial stages of model building, the product recommender pre-selects a noise that is substantially identical to the initial auxiliary representation gradient. This approach can accelerate model building.

[0085] After the above model is built, the product recommender saves the product representations corresponding to all products, and the auxiliary party saves the user representations corresponding to all users for use in product recommendations.

[0086] Specifically, data transmission is achieved between the product recommender and the auxiliary party through a communication connection.

[0087] Example 2

[0088] Figure 3 This is a structural diagram of a product recommendation system for a simple hybrid privacy protection scenario provided by an embodiment of the present invention. Figure 3 As shown, corresponding to the above-mentioned product recommendation method for a simple hybrid privacy protection scenario, this embodiment 2 provides a product recommendation system for a simple hybrid privacy protection scenario, which generally includes an acquisition module 10, a calculation module 20 and a determination module 30. The acquisition module 10 is used to obtain the user ID corresponding to the target user and send the user ID to the auxiliary party after receiving the recommendation request of the target user. The calculation module 20 is connected to the acquisition module 10, and is used to receive the user representation fed back by the auxiliary party, and obtain candidate products based on the user representation and all product representations calculated by the preset model. The user representation is calculated by the auxiliary party based on the user attributes obtained using the user ID and the preset model. The determination module 30 is connected to the calculation module 20, and is used to determine the target product from the candidate products according to the preset recommendation rules and send it to the target user.

[0089] The product recommendation system for a simple hybrid privacy protection scenario provided by the present invention involves two participants, namely a recommender and an auxiliary party. The recommender stores the interaction information between users and products, as well as product attributes, and the auxiliary party stores user attributes. In order to improve data privacy, the recommender does not send product attributes directly to the auxiliary party, and the auxiliary party does not send user attributes directly to the recommender. In order to realize model construction, complete data is required. Therefore, the auxiliary party converts user attributes into user representations and sends them to the recommender. The recommender converts product attributes into product representations and further obtains the auxiliary party representation gradient based on the product representation before sending it to the auxiliary party. In this way, the recommender obtains the user representation data required to update the model, and the auxiliary party obtains the product representation-related data required to update the model. That is, the recommender and the auxiliary party both obtain the data required to modify their respective models, and protect their respective data, thereby meeting the privacy requirements of simple hybrid scenarios.

[0090] For the parts not described in detail in the second embodiment, please refer to the description in the first embodiment, which will not be repeated here.

[0091] Example 3

[0092] Corresponding to the above method, the present invention further provides a computer device, comprising:

[0093] A processor and a memory, wherein the memory stores a computer program that can be run on the processor. When the computer program is executed by the processor, the product recommendation method for a simple hybrid privacy protection scenario provided by any one of the above embodiments is executed.

[0094] in, Figure 4 The computer device is shown as an example, and may include a processor 1510, a video display adapter 1511, a disk drive 1512, an input / output interface 1513, a network interface 1514, and a memory 1520. The processor 1510, the video display adapter 1511, the disk drive 1512, the input / output interface 1513, the network interface 1514, and the memory 1520 may be communicatively connected via a communication bus 1530.

[0095] Among them, the processor 1510 can be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, etc., to execute relevant programs to implement the technical solutions provided by the present invention.

[0096] The memory 1520 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1520 can store an operating system 1521 for controlling the operation of the electronic device and a basic input and output system (BIOS) for controlling the low-level operation of the electronic device. In addition, a web browser 1523, a data storage management system 1524, and a device identification information processing system 1525, etc. can also be stored. The above-mentioned device identification information processing system 1525 can be an application program that specifically implements the operations of the aforementioned steps in the embodiment of the present invention. In short, when the technical solution provided by the present invention is implemented by software or firmware, the relevant program code is stored in the memory 1520 and is called and executed by the processor 1510.

[0097] The input / output interface 1513 is used to connect input / output modules to implement information input and output. The input / output modules can be configured as components in the device (not shown in the figure) or can be externally connected to the device to provide corresponding functions. Input devices may include a keyboard, mouse, touch screen, microphone, various sensors, etc., and output devices may include a display, speaker, vibrator, indicator light, etc.

[0098] The network interface 1514 is used to connect to a communication module (not shown) to enable communication between the device and other devices. The communication module can communicate via a wired method (such as USB, network cable, etc.) or a wireless method (such as mobile network, WIFI, Bluetooth, etc.).

[0099] The bus comprises a pathway that transmits information between the various components of the device (eg, processor 1510 , video display adapter 1511 , disk drive 1512 , input / output interface 1513 , network interface 1514 , and memory 1520 ).

[0100] In addition, the electronic device can also obtain information on specific collection conditions from the virtual resource object collection condition information database for use in condition judgment, etc.

[0101] It should be noted that although the above device only shows a processor 1510, a video display adapter 1511, a disk drive 1512, an input / output interface 1513, a network interface 1514, a memory 1520, a bus, etc., in a specific implementation, the device may also include other components necessary for normal operation. In addition, those skilled in the art will understand that the above device may only include components necessary to implement the solution of the present invention, and does not necessarily include all the components shown in the figure.

[0102] Example 4

[0103] The present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed, it implements the product recommendation method for a simple hybrid privacy protection scenario provided by any one of the above embodiments.

[0104] From the above description of the embodiments, it can be seen that those skilled in the art can clearly understand that the present invention can be implemented by means of software plus the necessary general-purpose hardware platform. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in various embodiments of the present invention or certain parts of the embodiments.

[0105] Each embodiment in this specification is described in a progressive manner. The same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments. In particular, for system or system embodiments, since they are basically similar to method embodiments, the description is relatively simple. For relevant parts, refer to the partial description of the method embodiment. The system and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without expending creative work.

[0106] The technical solutions provided by the present invention have been described in detail above. Specific examples have been used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is intended only to help understand the method and core concept of the present invention. At the same time, those skilled in the art will appreciate that the specific implementation methods and application scopes may vary based on the concepts of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A product recommendation method for a simple hybrid privacy protection scenario, wherein the product recommendation method is applied to a product recommendation party, and is characterized in that: The product recommendation method includes: After receiving the recommendation request from the target user, obtaining the user ID corresponding to the target user and sending the user ID to the auxiliary party; receiving a user representation fed back by the auxiliary party, and obtaining candidate products based on the user representation and all product representations calculated using product attributes and a preset model, wherein the user representation is calculated by the auxiliary party based on the user attributes obtained using the user ID and the preset model, wherein the preset model includes a twin-tower model; A target product is determined from the candidate products according to a preset recommendation rule and sent to the target user.

2. The product recommendation method according to claim 1, characterized in that: The acquiring of candidate products according to the user representation and all product representations calculated through product attributes and a preset model includes: The distance between the user representation and all the product representations is calculated according to a preset algorithm, and the candidate products are obtained according to the distance.

3. The product recommendation method according to claim 2, characterized in that: Determining the target product from the candidate products according to the preset recommendation rules and sending the target product to the target user includes: All the candidate commodities are sorted according to the distance, and the candidate commodities at a preset position are selected as the target commodities and sent to the target user.

4. The product recommendation method according to any one of claims 1 to 3, characterized in that: The construction process of the preset model includes: Generate training sample sets based on the interaction information between users and products; Processing the training sample set to obtain a first sample set and a second sample set, and sending the second sample set to the auxiliary party; Extracting product attributes corresponding to the products in the first sample set, and calculating corresponding product representations based on the product attributes; receiving a user representation fed back by the auxiliary party, the user representation being obtained by the auxiliary party extracting user attributes corresponding to the user ID in the second sample set and calculating based on the user attributes; Calculate an auxiliary party representation gradient based on the product representation and the user representation and send the auxiliary party representation gradient to the auxiliary party, the auxiliary party calculates an auxiliary party neural network representation gradient based on the auxiliary party representation gradient and updates the first part of the preset model based on the auxiliary party neural network representation gradient; A recommender neural network representation gradient is calculated based on the product representation and the user representation, and a second part of the preset model is updated based on the recommender neural network representation gradient.

5. The product recommendation method according to claim 4, characterized in that: The training sample set includes a plurality of samples consisting of users, commodities clicked or purchased by the users, and a first preset number of commodities not clicked or purchased by the users, and processing the training sample set to obtain the first sample set and the second sample set includes: Selecting a second preset number of samples from the training sample set to generate the first sample set; The second sample set is generated according to the user in the first sample set.

6. The product recommendation method according to claim 4, characterized in that: The calculating the auxiliary party representation gradient according to the product representation and the user representation and sending the auxiliary party representation gradient to the auxiliary party includes: Calculating the product representation and the user representation according to a preset calculation formula to obtain an initial auxiliary party representation gradient; Noise obfuscation is performed on the initial auxiliary representation gradient to obtain the auxiliary representation gradient.

7. The product recommendation method according to claim 6, characterized in that: The performing noise obfuscation on the initial auxiliary side representation gradient to obtain the auxiliary side representation gradient comprises: Determine target noise according to the initial auxiliary representation gradient, and superimpose the target noise on the initial auxiliary representation gradient to perform noise confusion to obtain the auxiliary representation gradient; or A noise having the same intensity as the initial auxiliary side representation gradient is selected from a plurality of pre-stored noises, and the noise is superimposed on the initial auxiliary side representation gradient to obtain the auxiliary side representation gradient.

8. A product recommendation system for simple hybrid privacy protection scenarios, characterized by: include: An acquisition module, configured to, upon receiving a recommendation request from a target user, acquire a user ID corresponding to the target user and send the user ID to an auxiliary party; a calculation module, connected to the acquisition module, configured to receive the user representation fed back by the auxiliary party and acquire candidate products based on the user representation and all product representations calculated using product attributes and a preset model, wherein the user representation is calculated by the auxiliary party based on the user attributes obtained using the user ID and the preset model, wherein the preset model includes a twin-tower model; The determination module is connected to the calculation module and is used to determine the target product from the candidate products according to a preset recommendation rule and send the target product to the target user.

9. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the computer program is executed by the processor, the product recommendation method for a simple hybrid privacy protection scenario according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored therein, characterized in that: When the computer program is executed, the product recommendation method for a simple hybrid privacy protection scenario according to any one of claims 1 to 7 is implemented.

Images